Serialize execution environment changes for MHI

Currently, client devices are created in SBL or AMSS (mission mode) and only destroyed after power down or SYS ERROR. When moving between certain execution environments, such as from SBL to AMSS, no clean-up is required. This presents an issue where SBL-specific channels are left open and client drivers now run in an execution environment where they cannot operate. Fix this by expanding the mhi_destroy_device() to do an execution environment specific clean-up if one is requested. Close the gap and destroy devices in such scenarios that allow SBL client drivers to clean up once device enters mission mode.

Signed-off-by Bhaumik Bhatt <[email protected]>

Currently, client devices are created in SBL or AMSS (mission mode) and only destroyed after power down or SYS ERROR. When moving between certain execution environments, such as from SBL to AMSS, no clean-up is required. This presents an issue where SBL-specific channels are left open and client drivers now run in an execution environment where they cannot operate. Fix this by expanding the mhi_destroy_device() to do an execution environment specific clean-up if one is requested. Close the gap and destroy devices in such scenarios that allow SBL client drivers to clean up once device enters mission mode.

Signed-off-by Bhaumik Bhatt <[email protected]>

Currently, client devices are created in SBL or AMSS (mission mode) and only destroyed after power down or SYS ERROR. When moving between certain execution environments, such as from SBL to AMSS, no clean-up is required. This presents an issue where SBL-specific channels are left open and client drivers now run in an execution environment where they cannot operate. Fix this by expanding the mhi_destroy_device() to do an execution environment specific clean-up if one is requested. Close the gap and destroy devices in such scenarios that allow SBL client drivers to clean up once device enters mission mode.

Signed-off-by Bhaumik Bhatt <[email protected]>

---
 drivers/bus/mhi/core/main.c | 29 +++++++++++++++++++++++++----
 drivers/bus/mhi/core/pm.c   |  3 +++
 2 files changed, 28 insertions(+), 4 deletions(-)

drivers/bus/mhi/core/main.c +25 -4

index 4e0131b..7a2e98c 100644
@@ -244,8 +244,10 @@ static void mhi_del_ring_element(struct mhi_controller *mhi_cntrl,
 
 int mhi_destroy_device(struct device *dev, void *data)
 {
+	struct mhi_chan *ul_chan, *dl_chan;
 	struct mhi_device *mhi_dev;
 	struct mhi_controller *mhi_cntrl;
+	enum mhi_ee_type ee = MHI_EE_MAX;
 
 	if (dev->bus != &mhi_bus_type)
 		return 0;

@@ -257,6 +259,17 @@ int mhi_destroy_device(struct device *dev, void *data)
 	if (mhi_dev->dev_type == MHI_DEVICE_CONTROLLER)
 		return 0;
 
+	ul_chan = mhi_dev->ul_chan;
+	dl_chan = mhi_dev->dl_chan;
+
+	/*
+	 * If execution environment is specified, remove only those devices that
+	 * started in them based on ee_mask for the channels as we move on to a
+	 * different execution environment
+	 */
+	if (data)
+		ee = *(enum mhi_ee_type *)data;
+
 	/*
 	 * For the suspend and resume case, this function will get called
 	 * without mhi_unregister_controller(). Hence, we need to drop the

@@ -264,11 +277,19 @@ int mhi_destroy_device(struct device *dev, void *data)
 	 * be sure that there will be no instances of mhi_dev left after
 	 * this.
 	 */
-	if (mhi_dev->ul_chan)
-		put_device(&mhi_dev->ul_chan->mhi_dev->dev);
+	if (ul_chan) {
+		if (ee != MHI_EE_MAX && !(ul_chan->ee_mask & BIT(ee)))
+			return 0;
 
-	if (mhi_dev->dl_chan)
-		put_device(&mhi_dev->dl_chan->mhi_dev->dev);
+		put_device(&ul_chan->mhi_dev->dev);
+	}
+
+	if (dl_chan) {
+		if (ee != MHI_EE_MAX && !(dl_chan->ee_mask & BIT(ee)))
+			return 0;
+
+		put_device(&dl_chan->mhi_dev->dev);
+	}
 
 	dev_dbg(&mhi_cntrl->mhi_dev->dev, "destroy device for chan:%s\n",
 		 mhi_dev->name);

drivers/bus/mhi/core/pm.c +3 -0

index 681960c..3bd81d0 100644
@@ -377,6 +377,7 @@ static int mhi_pm_mission_mode_transition(struct mhi_controller *mhi_cntrl)
 {
 	struct mhi_event *mhi_event;
 	struct device *dev = &mhi_cntrl->mhi_dev->dev;
+	enum mhi_ee_type current_ee = mhi_cntrl->ee;
 	int i, ret;
 
 	dev_dbg(dev, "Processing Mission Mode transition\n");

@@ -395,6 +396,8 @@ static int mhi_pm_mission_mode_transition(struct mhi_controller *mhi_cntrl)
 
 	wake_up_all(&mhi_cntrl->state_event);
 
+	device_for_each_child(&mhi_cntrl->mhi_dev->dev, &current_ee,
+			      mhi_destroy_device);
 	mhi_cntrl->status_cb(mhi_cntrl, MHI_CB_EE_MISSION_MODE);
 
 	/* Force MHI to be in M0 state before continuing */

Signature · 2 lines

The Qualcomm Innovation Center, Inc. is a member of the Code Aurora Forum,
a Linux Foundation Collaborative Project

During full boot chain firmware download, the PM state worker downloads the AMSS image after a blocking wait for the SBL execution environment change when running in PBL transition itself. Improve this design by having the host download the AMSS image from the SBL transition of PM state worker thread when a DEV_ST_TRANSITION_SBL is queued instead of the blocking wait.

Signed-off-by Bhaumik Bhatt <[email protected]>

Reviewed-by Manivannan Sadhasivam <[email protected]>

In current design, whenever the BHI interrupt is fired, the execution environment is updated. This can cause race conditions and impede ongoing power up/down processing. For example, if a power down is in progress, MHI host updates to a local "disabled" execution environment. If a BHI interrupt fires later, that value gets replaced with one from the BHI EE register. This impacts the controller as it does not expect multiple RDDM execution environment change status callbacks as an example. Another issue would be that the device can enter mission mode and the execution environment is updated, while device creation for SBL channels is still going on due to slower PM state worker thread run, leading to multiple attempts at opening the same channel.

Ensure that EE changes are handled only from appropriate places and occur one after another and handle only PBL modes or RDDM EE changes as critical events directly from the interrupt handler. Simplify handling by waiting for SYS ERROR before handling RDDM. This also makes sure that we use the correct execution environment to notify the controller driver when the device resets to one of the PBL execution environments.

Signed-off-by Bhaumik Bhatt <[email protected]>

In current design, whenever the BHI interrupt is fired, the execution environment is updated. This can cause race conditions and impede ongoing power up/down processing. For example, if a power down is in progress, MHI host updates to a local "disabled" execution environment. If a BHI interrupt fires later, that value gets replaced with one from the BHI EE register. This impacts the controller as it does not expect multiple RDDM execution environment change status callbacks as an example. Another issue would be that the device can enter mission mode and the execution environment is updated, while device creation for SBL channels is still going on due to slower PM state worker thread run, leading to multiple attempts at opening the same channel.

Ensure that EE changes are handled only from appropriate places and occur one after another and handle only PBL modes or RDDM EE changes as critical events directly from the interrupt handler. Simplify handling by waiting for SYS ERROR before handling RDDM. This also makes sure that we use the correct execution environment to notify the controller driver when the device resets to one of the PBL execution environments.

Signed-off-by Bhaumik Bhatt <[email protected]>

---
 drivers/bus/mhi/core/main.c | 40 +++++++++++++++++++++-------------------
 drivers/bus/mhi/core/pm.c   |  7 ++++---
 2 files changed, 25 insertions(+), 22 deletions(-)

drivers/bus/mhi/core/main.c +21 -19

index 7a2e98c..9715f51 100644
@@ -430,7 +430,7 @@ irqreturn_t mhi_intvec_threaded_handler(int irq_number, void *priv)
 	struct device *dev = &mhi_cntrl->mhi_dev->dev;
 	enum mhi_state state = MHI_STATE_MAX;
 	enum mhi_pm_state pm_state = 0;
-	enum mhi_ee_type ee = 0;
+	enum mhi_ee_type ee = MHI_EE_MAX;
 
 	write_lock_irq(&mhi_cntrl->pm_lock);
 	if (!MHI_REG_ACCESS_VALID(mhi_cntrl->pm_state)) {

@@ -439,8 +439,7 @@ irqreturn_t mhi_intvec_threaded_handler(int irq_number, void *priv)
 	}
 
 	state = mhi_get_mhi_state(mhi_cntrl);
-	ee = mhi_cntrl->ee;
-	mhi_cntrl->ee = mhi_get_exec_env(mhi_cntrl);
+	ee = mhi_get_exec_env(mhi_cntrl);
 	dev_dbg(dev, "local ee:%s device ee:%s dev_state:%s\n",
 		TO_MHI_EXEC_STR(mhi_cntrl->ee), TO_MHI_EXEC_STR(ee),
 		TO_MHI_STATE_STR(state));

@@ -452,27 +451,30 @@ irqreturn_t mhi_intvec_threaded_handler(int irq_number, void *priv)
 	}
 	write_unlock_irq(&mhi_cntrl->pm_lock);
 
-	 /* If device supports RDDM don't bother processing SYS error */
-	if (mhi_cntrl->rddm_image) {
-		/* host may be performing a device power down already */
-		if (!mhi_is_active(mhi_cntrl))
-			goto exit_intvec;
+	if (pm_state != MHI_PM_SYS_ERR_DETECT || ee == mhi_cntrl->ee)
+		goto exit_intvec;
 
-		if (mhi_cntrl->ee == MHI_EE_RDDM && mhi_cntrl->ee != ee) {
+	switch (ee) {
+	case MHI_EE_RDDM:
+		/* proceed if power down is not already in progress */
+		if (mhi_cntrl->rddm_image && mhi_is_active(mhi_cntrl)) {
 			mhi_cntrl->status_cb(mhi_cntrl, MHI_CB_EE_RDDM);
+			mhi_cntrl->ee = ee;
 			wake_up_all(&mhi_cntrl->state_event);
 		}
-		goto exit_intvec;
-	}
-
-	if (pm_state == MHI_PM_SYS_ERR_DETECT) {
+		break;
+	case MHI_EE_PBL:
+	case MHI_EE_EDL:
+	case MHI_EE_PTHRU:
+		mhi_cntrl->status_cb(mhi_cntrl, MHI_CB_FATAL_ERROR);
+		mhi_cntrl->ee = ee;
 		wake_up_all(&mhi_cntrl->state_event);
-
-		/* For fatal errors, we let controller decide next step */
-		if (MHI_IN_PBL(ee))
-			mhi_cntrl->status_cb(mhi_cntrl, MHI_CB_FATAL_ERROR);
-		else
-			mhi_pm_sys_err_handler(mhi_cntrl);
+		mhi_pm_sys_err_handler(mhi_cntrl);
+		break;
+	default:
+		wake_up_all(&mhi_cntrl->state_event);
+		mhi_pm_sys_err_handler(mhi_cntrl);
+		break;
 	}
 
 exit_intvec:

drivers/bus/mhi/core/pm.c +4 -3

index c09ec13..c870fa8 100644
@@ -377,21 +377,22 @@ static int mhi_pm_mission_mode_transition(struct mhi_controller *mhi_cntrl)
 {
 	struct mhi_event *mhi_event;
 	struct device *dev = &mhi_cntrl->mhi_dev->dev;
-	enum mhi_ee_type current_ee = mhi_cntrl->ee;
+	enum mhi_ee_type ee = MHI_EE_MAX, current_ee = mhi_cntrl->ee;
 	int i, ret;
 
 	dev_dbg(dev, "Processing Mission Mode transition\n");
 
 	write_lock_irq(&mhi_cntrl->pm_lock);
 	if (MHI_REG_ACCESS_VALID(mhi_cntrl->pm_state))
-		mhi_cntrl->ee = mhi_get_exec_env(mhi_cntrl);
+		ee = mhi_get_exec_env(mhi_cntrl);
 
-	if (!MHI_IN_MISSION_MODE(mhi_cntrl->ee)) {
+	if (!MHI_IN_MISSION_MODE(ee)) {
 		mhi_cntrl->pm_state = MHI_PM_LD_ERR_FATAL_DETECT;
 		write_unlock_irq(&mhi_cntrl->pm_lock);
 		wake_up_all(&mhi_cntrl->state_event);
 		return -EIO;
 	}
+	mhi_cntrl->ee = ee;
 	write_unlock_irq(&mhi_cntrl->pm_lock);
 
 	wake_up_all(&mhi_cntrl->state_event);

Signature · 2 lines

The Qualcomm Innovation Center, Inc. is a member of the Code Aurora Forum,
a Linux Foundation Collaborative Project

In current design, whenever the BHI interrupt is fired, the execution environment is updated. This can cause race conditions and impede ongoing power up/down processing. For example, if a power down is in progress, MHI host updates to a local "disabled" execution environment. If a BHI interrupt fires later, that value gets replaced with one from the BHI EE register. This impacts the controller as it does not expect multiple RDDM execution environment change status callbacks as an example. Another issue would be that the device can enter mission mode and the execution environment is updated, while device creation for SBL channels is still going on due to slower PM state worker thread run, leading to multiple attempts at opening the same channel.

Ensure that EE changes are handled only from appropriate places and occur one after another and handle only PBL modes or RDDM EE changes as critical events directly from the interrupt handler. Simplify handling by waiting for SYS ERROR before handling RDDM. This also makes sure that we use the correct execution environment to notify the controller driver when the device resets to one of the PBL execution environments.

Signed-off-by Bhaumik Bhatt <[email protected]>

drivers/bus/mhi/core/main.c +13 -13

@@ -452,27 +451,30 @@ irqreturn_t mhi_intvec_threaded_handler(int irq_number, void *priv)
  	}
  	write_unlock_irq(&mhi_cntrl->pm_lock);
 
-	 /* If device supports RDDM don't bother processing SYS error */
-	if (mhi_cntrl->rddm_image) {
-		/* host may be performing a device power down already */
-		if (!mhi_is_active(mhi_cntrl))
-			goto exit_intvec;
+	if (pm_state != MHI_PM_SYS_ERR_DETECT || ee == mhi_cntrl->ee)
+		goto exit_intvec;
 
-		if (mhi_cntrl->ee == MHI_EE_RDDM && mhi_cntrl->ee != ee) {
+	switch (ee) {
+	case MHI_EE_RDDM:
+		/* proceed if power down is not already in progress */
+		if (mhi_cntrl->rddm_image && mhi_is_active(mhi_cntrl)) {
  			mhi_cntrl->status_cb(mhi_cntrl, MHI_CB_EE_RDDM);
+			mhi_cntrl->ee = ee;
  			wake_up_all(&mhi_cntrl->state_event);
  		}
-		goto exit_intvec;
-	}
-
-	if (pm_state == MHI_PM_SYS_ERR_DETECT) {
+		break;
+	case MHI_EE_PBL:
+	case MHI_EE_EDL:
+	case MHI_EE_PTHRU:
+		mhi_cntrl->status_cb(mhi_cntrl, MHI_CB_FATAL_ERROR);
+		mhi_cntrl->ee = ee;
  		wake_up_all(&mhi_cntrl->state_event);
-
-		/* For fatal errors, we let controller decide next step */
-		if (MHI_IN_PBL(ee))

Quoted 22 lines · Bhaumik Bhatt

In current design, whenever the BHI interrupt is fired, the execution environment is updated. This can cause race conditions and impede ongoing power up/down processing. For example, if a power down is in progress, MHI host updates to a local "disabled" execution environment. If a BHI interrupt fires later, that value gets replaced with one from the BHI EE register. This impacts the controller as it does not expect multiple RDDM execution environment change status callbacks as an example. Another issue would be that the device can enter mission mode and the execution environment is updated, while device creation for SBL channels is still going on due to slower PM state worker thread run, leading to multiple attempts at opening the same channel.

Ensure that EE changes are handled only from appropriate places and occur one after another and handle only PBL modes or RDDM EE changes as critical events directly from the interrupt handler. Simplify handling by waiting for SYS ERROR before handling RDDM. This also makes sure that we use the correct execution environment to notify the controller driver when the device resets to one of the PBL execution environments.

Signed-off-by Bhaumik Bhatt <[email protected]>

<snip>

Quoted 34 lines

drivers/bus/mhi/core/main.c +13 -11

@@ -452,27 +451,30 @@ irqreturn_t mhi_intvec_threaded_handler(int 
 irq_number, void *priv)
  	}
  	write_unlock_irq(&mhi_cntrl->pm_lock);
  -	 /* If device supports RDDM don't bother processing SYS error */
-	if (mhi_cntrl->rddm_image) {
-		/* host may be performing a device power down already */
-		if (!mhi_is_active(mhi_cntrl))
-			goto exit_intvec;
+	if (pm_state != MHI_PM_SYS_ERR_DETECT || ee == mhi_cntrl->ee)
+		goto exit_intvec;
  -		if (mhi_cntrl->ee == MHI_EE_RDDM && mhi_cntrl->ee != ee) {
+	switch (ee) {
+	case MHI_EE_RDDM:
+		/* proceed if power down is not already in progress */
+		if (mhi_cntrl->rddm_image && mhi_is_active(mhi_cntrl)) {
  			mhi_cntrl->status_cb(mhi_cntrl, MHI_CB_EE_RDDM);
+			mhi_cntrl->ee = ee;
  			wake_up_all(&mhi_cntrl->state_event);
  		}
-		goto exit_intvec;
-	}
-
-	if (pm_state == MHI_PM_SYS_ERR_DETECT) {
+		break;
+	case MHI_EE_PBL:
+	case MHI_EE_EDL:
+	case MHI_EE_PTHRU:
+		mhi_cntrl->status_cb(mhi_cntrl, MHI_CB_FATAL_ERROR);
+		mhi_cntrl->ee = ee;
  		wake_up_all(&mhi_cntrl->state_event);
-
-		/* For fatal errors, we let controller decide next step */
-		if (MHI_IN_PBL(ee))

drivers/bus/mhi/core/main.c irqreturn_t mhi_intvec_threaded_handler(int · line 479

Quoted 10 lines

drivers/bus/mhi/core/main.c +6 -3

-			mhi_cntrl->status_cb(mhi_cntrl, MHI_CB_FATAL_ERROR);
-		else
-			mhi_pm_sys_err_handler(mhi_cntrl);
+		mhi_pm_sys_err_handler(mhi_cntrl);
+		break;
+	default:
+		wake_up_all(&mhi_cntrl->state_event);
+		mhi_pm_sys_err_handler(mhi_cntrl);
+		break;
  	}

Bhaumik, can you explain the above change? Before this patch (which is now committed), if there was a fatal error, the controller was notified (MHI_CB_FATAL_ERROR) and it decided all action. After this patch, the controller is notified, but also the core attempts to handle the syserr.

This is a change in behavior, and seems to make a mess of the controller, and possibly the core fighting each other.

Specifically, I'm rebasing the AIC100 driver onto 5.13, which has this change, and I'm seeing a serious regression. I'm thinking that for the PBL/EDL/PTHRU case, mhi_pm_sys_err_handler() should not be called.

Thoughts?

Quoted 77 lines · Jeffrey Hugo

Quoted 22 lines · Bhaumik Bhatt

In current design, whenever the BHI interrupt is fired, the execution environment is updated. This can cause race conditions and impede ongoing power up/down processing. For example, if a power down is in progress, MHI host updates to a local "disabled" execution environment. If a BHI interrupt fires later, that value gets replaced with one from the BHI EE register. This impacts the controller as it does not expect multiple RDDM execution environment change status callbacks as an example. Another issue would be that the device can enter mission mode and the execution environment is updated, while device creation for SBL channels is still going on due to slower PM state worker thread run, leading to multiple attempts at opening the same channel.

Ensure that EE changes are handled only from appropriate places and occur one after another and handle only PBL modes or RDDM EE changes as critical events directly from the interrupt handler. Simplify handling by waiting for SYS ERROR before handling RDDM. This also makes sure that we use the correct execution environment to notify the controller driver when the device resets to one of the PBL execution environments.

Signed-off-by Bhaumik Bhatt <[email protected]>

<snip>

Quoted 36 lines

drivers/bus/mhi/core/main.c +0 -0

@@ -452,27 +451,30 @@ irqreturn_t mhi_intvec_threaded_handler(int 
 irq_number, void *priv)
       }
       write_unlock_irq(&mhi_cntrl->pm_lock);

  -     /* If device supports RDDM don't bother processing SYS error */ -    if (mhi_cntrl->rddm_image) {
-        /* host may be performing a device power down already */ -        if (!mhi_is_active(mhi_cntrl))
-            goto exit_intvec;
+    if (pm_state != MHI_PM_SYS_ERR_DETECT || ee == mhi_cntrl->ee) +        goto exit_intvec;
  -        if (mhi_cntrl->ee == MHI_EE_RDDM && mhi_cntrl->ee != ee) { +    switch (ee) {
+    case MHI_EE_RDDM:
+        /* proceed if power down is not already in progress */ +        if (mhi_cntrl->rddm_image && mhi_is_active(mhi_cntrl)) { mhi_cntrl->status_cb(mhi_cntrl, MHI_CB_EE_RDDM); +            mhi_cntrl->ee = ee; wake_up_all(&mhi_cntrl->state_event);
          }
-        goto exit_intvec;
-    }
-
-    if (pm_state == MHI_PM_SYS_ERR_DETECT) {
+        break;
+    case MHI_EE_PBL:
+    case MHI_EE_EDL:
+    case MHI_EE_PTHRU:
+        mhi_cntrl->status_cb(mhi_cntrl, MHI_CB_FATAL_ERROR); +        mhi_cntrl->ee = ee; wake_up_all(&mhi_cntrl->state_event);
-
-        /* For fatal errors, we let controller decide next step */ -        if (MHI_IN_PBL(ee))
-            mhi_cntrl->status_cb(mhi_cntrl, MHI_CB_FATAL_ERROR); -        else

Quoted 8 lines

-            mhi_pm_sys_err_handler(mhi_cntrl);
+        mhi_pm_sys_err_handler(mhi_cntrl);
+        break;
+    default:
+        wake_up_all(&mhi_cntrl->state_event);
+        mhi_pm_sys_err_handler(mhi_cntrl);
+        break;
      }

Bhaumik, can you explain the above change?  Before this patch (which is now committed), if there was a fatal error, the controller was notified (MHI_CB_FATAL_ERROR) and it decided all action.  After this patch, the controller is notified, but also the core attempts to handle the syserr.

Quoted 8 lines

This is a change in behavior, and seems to make a mess of the controller, and possibly the core fighting each other.

Specifically, I'm rebasing the AIC100 driver onto 5.13, which has this change, and I'm seeing a serious regression.  I'm thinking that for the PBL/EDL/PTHRU case, mhi_pm_sys_err_handler() should not be called.

Thoughts?

I see. We use this heavily for entry to EDL use-cases.

We do require the mhi_pm_sys_err_handler() to be called in those cases as any entry to PBL/EDL means there is need to clean-up MHI host and notify all its clients.

We include PTHRU in here because its a SYS ERROR in PBL modes.

Premise being the controller should not be in that business of doing any of the clean-up that is responsibility of the core driver. We're using this feature set to ensure controller is only notified.

What does AIC100 do on fatal error that you run in to issues? I don't think any of the other controllers do anything other than disabling runtime PM since device is down. Maybe there's some room for improvement.

Quoted 95 lines · Bhaumik Bhatt

Quoted 77 lines · Jeffrey Hugo

Quoted 22 lines · Bhaumik Bhatt

In current design, whenever the BHI interrupt is fired, the execution environment is updated. This can cause race conditions and impede ongoing power up/down processing. For example, if a power down is in progress, MHI host updates to a local "disabled" execution environment. If a BHI interrupt fires later, that value gets replaced with one from the BHI EE register. This impacts the controller as it does not expect multiple RDDM execution environment change status callbacks as an example. Another issue would be that the device can enter mission mode and the execution environment is updated, while device creation for SBL channels is still going on due to slower PM state worker thread run, leading to multiple attempts at opening the same channel.

Ensure that EE changes are handled only from appropriate places and occur one after another and handle only PBL modes or RDDM EE changes as critical events directly from the interrupt handler. Simplify handling by waiting for SYS ERROR before handling RDDM. This also makes sure that we use the correct execution environment to notify the controller driver when the device resets to one of the PBL execution environments.

Signed-off-by Bhaumik Bhatt <[email protected]>

<snip>

Quoted 36 lines

drivers/bus/mhi/core/main.c +0 -0

@@ -452,27 +451,30 @@ irqreturn_t mhi_intvec_threaded_handler(int 
 irq_number, void *priv)
       }
       write_unlock_irq(&mhi_cntrl->pm_lock);

  -     /* If device supports RDDM don't bother processing SYS error */ -    if (mhi_cntrl->rddm_image) {
-        /* host may be performing a device power down already */ -        if (!mhi_is_active(mhi_cntrl))
-            goto exit_intvec;
+    if (pm_state != MHI_PM_SYS_ERR_DETECT || ee == mhi_cntrl->ee) +        goto exit_intvec;
  -        if (mhi_cntrl->ee == MHI_EE_RDDM && mhi_cntrl->ee != ee) { +    switch (ee) {
+    case MHI_EE_RDDM:
+        /* proceed if power down is not already in progress */ +        if (mhi_cntrl->rddm_image && mhi_is_active(mhi_cntrl)) { mhi_cntrl->status_cb(mhi_cntrl, MHI_CB_EE_RDDM); +            mhi_cntrl->ee = ee; wake_up_all(&mhi_cntrl->state_event);
          }
-        goto exit_intvec;
-    }
-
-    if (pm_state == MHI_PM_SYS_ERR_DETECT) {
+        break;
+    case MHI_EE_PBL:
+    case MHI_EE_EDL:
+    case MHI_EE_PTHRU:
+        mhi_cntrl->status_cb(mhi_cntrl, MHI_CB_FATAL_ERROR); +        mhi_cntrl->ee = ee; wake_up_all(&mhi_cntrl->state_event);
-
-        /* For fatal errors, we let controller decide next step */ -        if (MHI_IN_PBL(ee))
-            mhi_cntrl->status_cb(mhi_cntrl, MHI_CB_FATAL_ERROR); -        else

Quoted 8 lines

-            mhi_pm_sys_err_handler(mhi_cntrl);
+        mhi_pm_sys_err_handler(mhi_cntrl);
+        break;
+    default:
+        wake_up_all(&mhi_cntrl->state_event);
+        mhi_pm_sys_err_handler(mhi_cntrl);
+        break;
      }

Bhaumik, can you explain the above change?  Before this patch (which is now committed), if there was a fatal error, the controller was notified (MHI_CB_FATAL_ERROR) and it decided all action.  After this patch, the controller is notified, but also the core attempts to handle the syserr.

Quoted 8 lines

This is a change in behavior, and seems to make a mess of the controller, and possibly the core fighting each other.

Specifically, I'm rebasing the AIC100 driver onto 5.13, which has this change, and I'm seeing a serious regression.  I'm thinking that for the PBL/EDL/PTHRU case, mhi_pm_sys_err_handler() should not be called.

Thoughts?

I see. We use this heavily for entry to EDL use-cases.

We do require the mhi_pm_sys_err_handler() to be called in those cases as any entry to PBL/EDL means there is need to clean-up MHI host and notify all its clients.

We include PTHRU in here because its a SYS ERROR in PBL modes.

Quoted 7 lines

Premise being the controller should not be in that business of doing any of the clean-up that is responsibility of the core driver. We're using this feature set to ensure controller is only notified.

What does AIC100 do on fatal error that you run in to issues? I don't think any of the other controllers do anything other than disabling runtime PM since device is down. Maybe there's some room for improvement.

Our usecase is PBL as a result of a full device crash. AIC100 doesn't exercise the EDL/PTHRU usecases. (Just giving you some context, not trying to imply EDL is not valuable to others for example).

In that case (FATAL_ERROR), our controller schedules a work item, and then returnsas we assume FATAL_ERROR is notified in atomic context. We need to do non-mhi cleanup first. Then we powerdown the MHI to cleanup the MHI core, and kick off all the clients (its been a long while, but initially, we were seeing the syserr handling in the core not sufficiently kicking off the clients). Then we power up MHI. MHI will init in PBL, still in syserr, and handle it.

I haven't fully traced everything, but we were getting into some really bad states with the core triggering mhi_pm_sys_err_handler() per this patch.

Its important that we have control over the ordering of our cleanup, vs the MHI cleanup. Sadly, non-atomic context (sleeping) is a requirement of our cleanup.

Seems like we have differing requirements here. Hmm. What about an API the controller can call, that does mhi_pm_sys_err_handler() (or equivalent) so that the controller can control when the MHI core cleanup is done, but doesn't need to be concerned with the details? Or, do you have a suggestion?

Update debug prints to include local device in the BHI interrupt handler. This helps show transitions better between MHI states.

Signed-off-by Bhaumik Bhatt <[email protected]>

Reviewed-by Manivannan Sadhasivam <[email protected]>

v6:

(quoted diff) +0 -3

-Add patch to improve debug message
-Fix switch-case fall through warning for EE serialization patch
-Address review comments and update commit text

v5:

(quoted diff) +0 -3

-Update commit text for "clear devices when moving execution environments" patch
-Added test platform details that were missed out in the cover letter
-Merged two if checks in to a single one for EE serialization patch

v4:

(quoted diff) +0 -2

-Addressed review comments for additional info logging for EE movements
-Updated switch case for EE handling in mhi_intvec_threaded_handler()

v3:
-Update commit text to accurately reflect changes and reasoning based on reviews

v2:
-Add patch to clear devices when moving execution environments

Note: This patch is first in series of execution environment related changes.

During full boot chain firmware download, the PM state worker downloads the AMSS image after waiting for the SBL execution environment change in PBL mode itself. Since getting rid of the firmware load worker thread, this design needs to change and MHI host must download the AMSS image from the SBL mode of PM state worker thread instead of blocking waits for SBL EE in PBL transition processing.

Ensure that EE changes are handled only from appropriate places and occur one after another and handle only PBL or RDDM EE changes as critical events directly from the interrupt handler and the status callback is given to the controller drivers promptly.

When moving from SBL to AMSS EE, clear SBL specific client devices by calling remove callbacks for them so they are not left opened in a different execution environment.

This patchset was tested on ARM64.

Bhaumik Bhatt (4):
bus: mhi: core: Destroy SBL devices when moving to mission mode
bus: mhi: core: Download AMSS image from appropriate function
bus: mhi: core: Process execution environment changes serially
bus: mhi: core: Update debug prints to include local device state

 drivers/bus/mhi/core/boot.c     | 51 +++++++++++++--------------
 drivers/bus/mhi/core/internal.h |  1 +
 drivers/bus/mhi/core/main.c     | 76 +++++++++++++++++++++++++++--------------
 drivers/bus/mhi/core/pm.c       | 10 ++++--
 4 files changed, 83 insertions(+), 55 deletions(-)

Signature · 2 lines

The Qualcomm Innovation Center, Inc. is a member of the Code Aurora Forum,
a Linux Foundation Collaborative Project