[PATCH v2] bus: mhi: core: Check state before processing power_down

[PATCH v2] bus: mhi: core: Check state before processing power_down

[Jeffrey Hugo]

We cannot process a power_down if the power state is DISABLED.  There is
no valid mhi_ctxt in that case, so attepting to process the power_down
will likely result in a null pointer dereference.  If the power state is
DISABLED, there is nothing to do anyways, so just bail early.

Signed-off-by: Jeffrey Hugo <jhugo@codeaurora.org>
---

v2: Fix subject and tweak the locking to avoid needless lock/unlock/relock

 drivers/bus/mhi/core/pm.c | 9 ++++++++-
 1 file changed, 8 insertions(+), 1 deletion(-)

diff --git a/drivers/bus/mhi/core/pm.c b/drivers/bus/mhi/core/pm.c
index 56ba3ab..47f6621 100644
--- a/drivers/bus/mhi/core/pm.c
+++ b/drivers/bus/mhi/core/pm.c
@@ -1144,6 +1144,7 @@ int mhi_async_power_up(struct mhi_controller *mhi_cntrl)
 		mhi_deinit_dev_ctxt(mhi_cntrl);
 
 error_dev_ctxt:
+	mhi_cntrl->pm_state = MHI_PM_DISABLE;
 	mutex_unlock(&mhi_cntrl->pm_mutex);
 
 	return ret;
@@ -1155,11 +1156,17 @@ void mhi_power_down(struct mhi_controller *mhi_cntrl, bool graceful)
 	enum mhi_pm_state cur_state, transition_state;
 	struct device *dev = &mhi_cntrl->mhi_dev->dev;
 
+	mutex_lock(&mhi_cntrl->pm_mutex);
+	cur_state = mhi_cntrl->pm_state;
+	if (cur_state == MHI_PM_DISABLE) {
+		mutex_unlock(&mhi_cntrl->pm_mutex);
+		return; /* Already powered down */
+	}
+
 	/* If it's not a graceful shutdown, force MHI to linkdown state */
 	transition_state = (graceful) ? MHI_PM_SHUTDOWN_PROCESS :
 			   MHI_PM_LD_ERR_FATAL_DETECT;
 
-	mutex_lock(&mhi_cntrl->pm_mutex);
 	write_lock_irq(&mhi_cntrl->pm_lock);
 	cur_state = mhi_tryset_pm_state(mhi_cntrl, transition_state);
 	if (cur_state != transition_state) {
-- 
Qualcomm Technologies, Inc. is a member of the
Code Aurora Forum, a Linux Foundation Collaborative Project.

Re: [PATCH v2] bus: mhi: core: Check state before processing power_down

[Manivannan Sadhasivam]

On Wed, Feb 17, 2021 at 09:43:31AM -0700, Jeffrey Hugo wrote:
> We cannot process a power_down if the power state is DISABLED.  There is
> no valid mhi_ctxt in that case, so attepting to process the power_down
> will likely result in a null pointer dereference.  If the power state is
> DISABLED, there is nothing to do anyways, so just bail early.
> 
> Signed-off-by: Jeffrey Hugo <jhugo@codeaurora.org>
> ---
> 
> v2: Fix subject and tweak the locking to avoid needless lock/unlock/relock
> 
>  drivers/bus/mhi/core/pm.c | 9 ++++++++-
>  1 file changed, 8 insertions(+), 1 deletion(-)
> 
> diff --git a/drivers/bus/mhi/core/pm.c b/drivers/bus/mhi/core/pm.c
> index 56ba3ab..47f6621 100644
> --- a/drivers/bus/mhi/core/pm.c
> +++ b/drivers/bus/mhi/core/pm.c
> @@ -1144,6 +1144,7 @@ int mhi_async_power_up(struct mhi_controller *mhi_cntrl)
>  		mhi_deinit_dev_ctxt(mhi_cntrl);
>  
>  error_dev_ctxt:
> +	mhi_cntrl->pm_state = MHI_PM_DISABLE;
>  	mutex_unlock(&mhi_cntrl->pm_mutex);
>  
>  	return ret;
> @@ -1155,11 +1156,17 @@ void mhi_power_down(struct mhi_controller *mhi_cntrl, bool graceful)
>  	enum mhi_pm_state cur_state, transition_state;
>  	struct device *dev = &mhi_cntrl->mhi_dev->dev;
>  
> +	mutex_lock(&mhi_cntrl->pm_mutex);
> +	cur_state = mhi_cntrl->pm_state;

As I said in my previous review, you should use pm_lock for reading the
pm_state.

Thanks,
Mani

> +	if (cur_state == MHI_PM_DISABLE) {
> +		mutex_unlock(&mhi_cntrl->pm_mutex);
> +		return; /* Already powered down */
> +	}
> +
>  	/* If it's not a graceful shutdown, force MHI to linkdown state */
>  	transition_state = (graceful) ? MHI_PM_SHUTDOWN_PROCESS :
>  			   MHI_PM_LD_ERR_FATAL_DETECT;
>  
> -	mutex_lock(&mhi_cntrl->pm_mutex);
>  	write_lock_irq(&mhi_cntrl->pm_lock);
>  	cur_state = mhi_tryset_pm_state(mhi_cntrl, transition_state);
>  	if (cur_state != transition_state) {
> -- 
> Qualcomm Technologies, Inc. is a member of the
> Code Aurora Forum, a Linux Foundation Collaborative Project.
> 

Re: [PATCH v2] bus: mhi: core: Check state before processing power_down

[Jeffrey Hugo]

On 2/24/2021 2:55 AM, Manivannan Sadhasivam wrote:
> On Wed, Feb 17, 2021 at 09:43:31AM -0700, Jeffrey Hugo wrote:
>> We cannot process a power_down if the power state is DISABLED.  There is
>> no valid mhi_ctxt in that case, so attepting to process the power_down
>> will likely result in a null pointer dereference.  If the power state is
>> DISABLED, there is nothing to do anyways, so just bail early.
>>
>> Signed-off-by: Jeffrey Hugo <jhugo@codeaurora.org>
>> ---
>>
>> v2: Fix subject and tweak the locking to avoid needless lock/unlock/relock
>>
>>   drivers/bus/mhi/core/pm.c | 9 ++++++++-
>>   1 file changed, 8 insertions(+), 1 deletion(-)
>>
>> diff --git a/drivers/bus/mhi/core/pm.c b/drivers/bus/mhi/core/pm.c
>> index 56ba3ab..47f6621 100644
>> --- a/drivers/bus/mhi/core/pm.c
>> +++ b/drivers/bus/mhi/core/pm.c
>> @@ -1144,6 +1144,7 @@ int mhi_async_power_up(struct mhi_controller *mhi_cntrl)
>>   		mhi_deinit_dev_ctxt(mhi_cntrl);
>>   
>>   error_dev_ctxt:
>> +	mhi_cntrl->pm_state = MHI_PM_DISABLE;
>>   	mutex_unlock(&mhi_cntrl->pm_mutex);
>>   
>>   	return ret;
>> @@ -1155,11 +1156,17 @@ void mhi_power_down(struct mhi_controller *mhi_cntrl, bool graceful)
>>   	enum mhi_pm_state cur_state, transition_state;
>>   	struct device *dev = &mhi_cntrl->mhi_dev->dev;
>>   
>> +	mutex_lock(&mhi_cntrl->pm_mutex);
>> +	cur_state = mhi_cntrl->pm_state;
> 
> As I said in my previous review, you should use pm_lock for reading the
> pm_state.

You also said on IRC that is a refactor of the entire driver, and not a 
blocker for this fix.  Based on that, I intrepreted your comments as 
nothing needed to be done regarding the locking, so I'm confused by this 
comment.

I'm not entirely sure what you want since I feel like you are giving 
conflicting direction, but I'm guessing you want the lock of the pm_lock 
to be moved up to just under the mutex lock then?

-- 
Jeffrey Hugo
Qualcomm Technologies, Inc. is a member of the
Code Aurora Forum, a Linux Foundation Collaborative Project.

Re: [PATCH v2] bus: mhi: core: Check state before processing power_down

[Manivannan Sadhasivam]

On Wed, Feb 24, 2021 at 08:14:53AM -0700, Jeffrey Hugo wrote:
> On 2/24/2021 2:55 AM, Manivannan Sadhasivam wrote:
> > On Wed, Feb 17, 2021 at 09:43:31AM -0700, Jeffrey Hugo wrote:
> > > We cannot process a power_down if the power state is DISABLED.  There is
> > > no valid mhi_ctxt in that case, so attepting to process the power_down
> > > will likely result in a null pointer dereference.  If the power state is
> > > DISABLED, there is nothing to do anyways, so just bail early.
> > > 
> > > Signed-off-by: Jeffrey Hugo <jhugo@codeaurora.org>
> > > ---
> > > 
> > > v2: Fix subject and tweak the locking to avoid needless lock/unlock/relock
> > > 
> > >   drivers/bus/mhi/core/pm.c | 9 ++++++++-
> > >   1 file changed, 8 insertions(+), 1 deletion(-)
> > > 
> > > diff --git a/drivers/bus/mhi/core/pm.c b/drivers/bus/mhi/core/pm.c
> > > index 56ba3ab..47f6621 100644
> > > --- a/drivers/bus/mhi/core/pm.c
> > > +++ b/drivers/bus/mhi/core/pm.c
> > > @@ -1144,6 +1144,7 @@ int mhi_async_power_up(struct mhi_controller *mhi_cntrl)
> > >   		mhi_deinit_dev_ctxt(mhi_cntrl);
> > >   error_dev_ctxt:
> > > +	mhi_cntrl->pm_state = MHI_PM_DISABLE;
> > >   	mutex_unlock(&mhi_cntrl->pm_mutex);
> > >   	return ret;
> > > @@ -1155,11 +1156,17 @@ void mhi_power_down(struct mhi_controller *mhi_cntrl, bool graceful)
> > >   	enum mhi_pm_state cur_state, transition_state;
> > >   	struct device *dev = &mhi_cntrl->mhi_dev->dev;
> > > +	mutex_lock(&mhi_cntrl->pm_mutex);
> > > +	cur_state = mhi_cntrl->pm_state;
> > 
> > As I said in my previous review, you should use pm_lock for reading the
> > pm_state.
> 
> You also said on IRC that is a refactor of the entire driver, and not a
> blocker for this fix.  Based on that, I intrepreted your comments as nothing
> needed to be done regarding the locking, so I'm confused by this comment.
> 
> I'm not entirely sure what you want since I feel like you are giving
> conflicting direction, but I'm guessing you want the lock of the pm_lock to
> be moved up to just under the mutex lock then?
> 

Yes, that's what my intention is. Sorry if I confused you.

Thanks,
Mani

> -- 
> Jeffrey Hugo
> Qualcomm Technologies, Inc. is a member of the
> Code Aurora Forum, a Linux Foundation Collaborative Project.