[PATCH 0/1] MHI fix for v5.16 - Take 2

[PATCH 0/1] MHI fix for v5.16 - Take 2

[Manivannan Sadhasivam]

Hi Greg,

Here is the final MHI fix for the v5.16 cycle. Kalle reported a regression with
ath11k stack due to 020d3b26c07a, and that's been affecting multiple WLAN
products like QCA6390 and WCN6855. The regression doesn't let the device to
resume from suspend and that has been due to a firmware bug. The device goes
to a bad MHI state during suspend and that can be seen in the MHI state during
resume. But that doesn't seem to affect the functionality of the device as it
continues to work fine if we allow it to resume.

Ideally, this should be fixed in the firmware but as like most of the products
out in market, getting a firmware fix for this issue not doable. So we have to
workaround for it in the MHI stack by allowing it to resume irrespective of the
MHI state.

Since the fix patch also touches ath11k stack, Kalle has given his Ack so that
the patch can be taken via mhi/char-misc tree.

Thanks,
Mani

Loic Poulain (1):
  bus: mhi: core: Add support for forced PM resume

 drivers/bus/mhi/core/pm.c             | 10 +++++++---
 drivers/bus/mhi/pci_generic.c         |  2 +-
 drivers/net/wireless/ath/ath11k/mhi.c |  6 +++++-
 include/linux/mhi.h                   |  3 ++-
 4 files changed, 15 insertions(+), 6 deletions(-)

-- 
2.25.1

[PATCH 1/1] bus: mhi: core: Add support for forced PM resume

[Manivannan Sadhasivam]

From: Loic Poulain <loic.poulain@linaro.org>

For whatever reason, some devices like QCA6390, WCN6855 using ath11k
are not in M3 state during PM resume, but still functional. The
mhi_pm_resume should then not fail in those cases, and let the higher
level device specific stack continue resuming process.

Add a new parameter to mhi_pm_resume, to force resuming, whatever the
current MHI state is. This fixes a regression with non functional
ath11k WiFi after suspend/resume cycle on some machines.

Bug report: https://bugzilla.kernel.org/show_bug.cgi?id=214179

Fixes: 020d3b26c07a ("bus: mhi: Early MHI resume failure in non M3 state")
Cc: stable@vger.kernel.org #5.13
Link: https://lore.kernel.org/regressions/871r5p0x2u.fsf@codeaurora.org/
Reported-by: Kalle Valo <kvalo@codeaurora.org>
Reported-by: Pengyu Ma <mapengyu@gmail.com>
Tested-by: Kalle Valo <kvalo@kernel.org>
Acked-by: Kalle Valo <kvalo@kernel.org>
Signed-off-by: Loic Poulain <loic.poulain@linaro.org>
[mani: Added comment, bug report, reported-by tags and CCed stable]
Link: https://lore.kernel.org/r/20211206161059.107007-1-manivannan.sadhasivam@linaro.org
Signed-off-by: Manivannan Sadhasivam <manivannan.sadhasivam@linaro.org>
---
 drivers/bus/mhi/core/pm.c             | 10 +++++++---
 drivers/bus/mhi/pci_generic.c         |  2 +-
 drivers/net/wireless/ath/ath11k/mhi.c |  6 +++++-
 include/linux/mhi.h                   |  3 ++-
 4 files changed, 15 insertions(+), 6 deletions(-)

diff --git a/drivers/bus/mhi/core/pm.c b/drivers/bus/mhi/core/pm.c
index fb99e3727155..8a486374d57a 100644
--- a/drivers/bus/mhi/core/pm.c
+++ b/drivers/bus/mhi/core/pm.c
@@ -881,7 +881,7 @@ int mhi_pm_suspend(struct mhi_controller *mhi_cntrl)
 }
 EXPORT_SYMBOL_GPL(mhi_pm_suspend);
 
-int mhi_pm_resume(struct mhi_controller *mhi_cntrl)
+int mhi_pm_resume(struct mhi_controller *mhi_cntrl, bool force)
 {
 	struct mhi_chan *itr, *tmp;
 	struct device *dev = &mhi_cntrl->mhi_dev->dev;
@@ -898,8 +898,12 @@ int mhi_pm_resume(struct mhi_controller *mhi_cntrl)
 	if (MHI_PM_IN_ERROR_STATE(mhi_cntrl->pm_state))
 		return -EIO;
 
-	if (mhi_get_mhi_state(mhi_cntrl) != MHI_STATE_M3)
-		return -EINVAL;
+	if (mhi_get_mhi_state(mhi_cntrl) != MHI_STATE_M3) {
+		dev_warn(dev, "Resuming from non M3 state (%s)\n",
+			 TO_MHI_STATE_STR(mhi_get_mhi_state(mhi_cntrl)));
+		if (!force)
+			return -EINVAL;
+	}
 
 	/* Notify clients about exiting LPM */
 	list_for_each_entry_safe(itr, tmp, &mhi_cntrl->lpm_chans, node) {
diff --git a/drivers/bus/mhi/pci_generic.c b/drivers/bus/mhi/pci_generic.c
index 4c577a731709..3394819e8115 100644
--- a/drivers/bus/mhi/pci_generic.c
+++ b/drivers/bus/mhi/pci_generic.c
@@ -962,7 +962,7 @@ static int __maybe_unused mhi_pci_runtime_resume(struct device *dev)
 		return 0; /* Nothing to do at MHI level */
 
 	/* Exit M3, transition to M0 state */
-	err = mhi_pm_resume(mhi_cntrl);
+	err = mhi_pm_resume(mhi_cntrl, false);
 	if (err) {
 		dev_err(&pdev->dev, "failed to resume device: %d\n", err);
 		goto err_recovery;
diff --git a/drivers/net/wireless/ath/ath11k/mhi.c b/drivers/net/wireless/ath/ath11k/mhi.c
index 26c7ae242db6..f1f2fa2d690d 100644
--- a/drivers/net/wireless/ath/ath11k/mhi.c
+++ b/drivers/net/wireless/ath/ath11k/mhi.c
@@ -533,7 +533,11 @@ static int ath11k_mhi_set_state(struct ath11k_pci *ab_pci,
 		ret = mhi_pm_suspend(ab_pci->mhi_ctrl);
 		break;
 	case ATH11K_MHI_RESUME:
-		ret = mhi_pm_resume(ab_pci->mhi_ctrl);
+		/* Do force MHI resume as some devices like QCA6390, WCN6855
+		 * are not in M3 state but they are functional. So just ignore
+		 * the MHI state while resuming.
+		 */
+		ret = mhi_pm_resume(ab_pci->mhi_ctrl, true);
 		break;
 	case ATH11K_MHI_TRIGGER_RDDM:
 		ret = mhi_force_rddm_mode(ab_pci->mhi_ctrl);
diff --git a/include/linux/mhi.h b/include/linux/mhi.h
index 723985879035..102303288cee 100644
--- a/include/linux/mhi.h
+++ b/include/linux/mhi.h
@@ -660,8 +660,9 @@ int mhi_pm_suspend(struct mhi_controller *mhi_cntrl);
 /**
  * mhi_pm_resume - Resume MHI from suspended state
  * @mhi_cntrl: MHI controller
+ * @force: Force resuming to M0 irrespective of the device MHI state
  */
-int mhi_pm_resume(struct mhi_controller *mhi_cntrl);
+int mhi_pm_resume(struct mhi_controller *mhi_cntrl, bool force);
 
 /**
  * mhi_download_rddm_image - Download ramdump image from device for
-- 
2.25.1

Re: [PATCH 1/1] bus: mhi: core: Add support for forced PM resume

[Greg KH]

On Wed, Dec 08, 2021 at 02:27:35PM +0530, Manivannan Sadhasivam wrote:
> From: Loic Poulain <loic.poulain@linaro.org>
> 
> For whatever reason, some devices like QCA6390, WCN6855 using ath11k
> are not in M3 state during PM resume, but still functional. The
> mhi_pm_resume should then not fail in those cases, and let the higher
> level device specific stack continue resuming process.
> 
> Add a new parameter to mhi_pm_resume, to force resuming, whatever the
> current MHI state is. This fixes a regression with non functional
> ath11k WiFi after suspend/resume cycle on some machines.
> 
> Bug report: https://bugzilla.kernel.org/show_bug.cgi?id=214179
> 
> Fixes: 020d3b26c07a ("bus: mhi: Early MHI resume failure in non M3 state")
> Cc: stable@vger.kernel.org #5.13
> Link: https://lore.kernel.org/regressions/871r5p0x2u.fsf@codeaurora.org/
> Reported-by: Kalle Valo <kvalo@codeaurora.org>
> Reported-by: Pengyu Ma <mapengyu@gmail.com>
> Tested-by: Kalle Valo <kvalo@kernel.org>
> Acked-by: Kalle Valo <kvalo@kernel.org>
> Signed-off-by: Loic Poulain <loic.poulain@linaro.org>
> [mani: Added comment, bug report, reported-by tags and CCed stable]
> Link: https://lore.kernel.org/r/20211206161059.107007-1-manivannan.sadhasivam@linaro.org
> Signed-off-by: Manivannan Sadhasivam <manivannan.sadhasivam@linaro.org>
> ---
>  drivers/bus/mhi/core/pm.c             | 10 +++++++---
>  drivers/bus/mhi/pci_generic.c         |  2 +-
>  drivers/net/wireless/ath/ath11k/mhi.c |  6 +++++-
>  include/linux/mhi.h                   |  3 ++-
>  4 files changed, 15 insertions(+), 6 deletions(-)
> 
> diff --git a/drivers/bus/mhi/core/pm.c b/drivers/bus/mhi/core/pm.c
> index fb99e3727155..8a486374d57a 100644
> --- a/drivers/bus/mhi/core/pm.c
> +++ b/drivers/bus/mhi/core/pm.c
> @@ -881,7 +881,7 @@ int mhi_pm_suspend(struct mhi_controller *mhi_cntrl)
>  }
>  EXPORT_SYMBOL_GPL(mhi_pm_suspend);
>  
> -int mhi_pm_resume(struct mhi_controller *mhi_cntrl)
> +int mhi_pm_resume(struct mhi_controller *mhi_cntrl, bool force)
>  {
>  	struct mhi_chan *itr, *tmp;
>  	struct device *dev = &mhi_cntrl->mhi_dev->dev;
> @@ -898,8 +898,12 @@ int mhi_pm_resume(struct mhi_controller *mhi_cntrl)
>  	if (MHI_PM_IN_ERROR_STATE(mhi_cntrl->pm_state))
>  		return -EIO;
>  
> -	if (mhi_get_mhi_state(mhi_cntrl) != MHI_STATE_M3)
> -		return -EINVAL;
> +	if (mhi_get_mhi_state(mhi_cntrl) != MHI_STATE_M3) {
> +		dev_warn(dev, "Resuming from non M3 state (%s)\n",
> +			 TO_MHI_STATE_STR(mhi_get_mhi_state(mhi_cntrl)));
> +		if (!force)
> +			return -EINVAL;
> +	}
>  
>  	/* Notify clients about exiting LPM */
>  	list_for_each_entry_safe(itr, tmp, &mhi_cntrl->lpm_chans, node) {
> diff --git a/drivers/bus/mhi/pci_generic.c b/drivers/bus/mhi/pci_generic.c
> index 4c577a731709..3394819e8115 100644
> --- a/drivers/bus/mhi/pci_generic.c
> +++ b/drivers/bus/mhi/pci_generic.c
> @@ -962,7 +962,7 @@ static int __maybe_unused mhi_pci_runtime_resume(struct device *dev)
>  		return 0; /* Nothing to do at MHI level */
>  
>  	/* Exit M3, transition to M0 state */
> -	err = mhi_pm_resume(mhi_cntrl);
> +	err = mhi_pm_resume(mhi_cntrl, false);
>  	if (err) {
>  		dev_err(&pdev->dev, "failed to resume device: %d\n", err);
>  		goto err_recovery;
> diff --git a/drivers/net/wireless/ath/ath11k/mhi.c b/drivers/net/wireless/ath/ath11k/mhi.c
> index 26c7ae242db6..f1f2fa2d690d 100644
> --- a/drivers/net/wireless/ath/ath11k/mhi.c
> +++ b/drivers/net/wireless/ath/ath11k/mhi.c
> @@ -533,7 +533,11 @@ static int ath11k_mhi_set_state(struct ath11k_pci *ab_pci,
>  		ret = mhi_pm_suspend(ab_pci->mhi_ctrl);
>  		break;
>  	case ATH11K_MHI_RESUME:
> -		ret = mhi_pm_resume(ab_pci->mhi_ctrl);
> +		/* Do force MHI resume as some devices like QCA6390, WCN6855
> +		 * are not in M3 state but they are functional. So just ignore
> +		 * the MHI state while resuming.
> +		 */
> +		ret = mhi_pm_resume(ab_pci->mhi_ctrl, true);
>  		break;
>  	case ATH11K_MHI_TRIGGER_RDDM:
>  		ret = mhi_force_rddm_mode(ab_pci->mhi_ctrl);
> diff --git a/include/linux/mhi.h b/include/linux/mhi.h
> index 723985879035..102303288cee 100644
> --- a/include/linux/mhi.h
> +++ b/include/linux/mhi.h
> @@ -660,8 +660,9 @@ int mhi_pm_suspend(struct mhi_controller *mhi_cntrl);
>  /**
>   * mhi_pm_resume - Resume MHI from suspended state
>   * @mhi_cntrl: MHI controller
> + * @force: Force resuming to M0 irrespective of the device MHI state
>   */
> -int mhi_pm_resume(struct mhi_controller *mhi_cntrl);
> +int mhi_pm_resume(struct mhi_controller *mhi_cntrl, bool force);

apis like this are horrid to work with over time.

Why not just have:
	mhi_pm_resume_force()
which then internally can set a flag that does this?  That way the
driver author does not have to stop every time they see this call and
look up exactly what the true/false field means in the function call in
their driver.

It also lets you leave alone the existing calls to mhi_pm_suspend() that
do not want to "force" anything.

self-documenting code is good, this is not self-documenting at all.

Also, is "force" really what you are doing here?  This is a "normal"
resume call, which should always work.  The "force" option here really
is just "ignore the current state of suspend for the device".  So
perhaps mhi_pm_resume_ignore_current_state() might be better?  Or
something shorter?

Naming is hard, sorry.

thanks,

greg k-h

Re: [PATCH 1/1] bus: mhi: core: Add support for forced PM resume

[Manivannan Sadhasivam]

On Wed, Dec 08, 2021 at 10:07:32AM +0100, Greg KH wrote:

[...]

> > diff --git a/include/linux/mhi.h b/include/linux/mhi.h
> > index 723985879035..102303288cee 100644
> > --- a/include/linux/mhi.h
> > +++ b/include/linux/mhi.h
> > @@ -660,8 +660,9 @@ int mhi_pm_suspend(struct mhi_controller *mhi_cntrl);
> >  /**
> >   * mhi_pm_resume - Resume MHI from suspended state
> >   * @mhi_cntrl: MHI controller
> > + * @force: Force resuming to M0 irrespective of the device MHI state
> >   */
> > -int mhi_pm_resume(struct mhi_controller *mhi_cntrl);
> > +int mhi_pm_resume(struct mhi_controller *mhi_cntrl, bool force);
> 
> apis like this are horrid to work with over time.
> 
> Why not just have:
> 	mhi_pm_resume_force()
> which then internally can set a flag that does this?  That way the
> driver author does not have to stop every time they see this call and
> look up exactly what the true/false field means in the function call in
> their driver.
> 

Okay.

> It also lets you leave alone the existing calls to mhi_pm_suspend() that
> do not want to "force" anything.
> 
> self-documenting code is good, this is not self-documenting at all.
> 
> Also, is "force" really what you are doing here?  This is a "normal"
> resume call, which should always work.

The normal resume here is resuming with M3 state only.

> The "force" option here really
> is just "ignore the current state of suspend for the device".  So
> perhaps mhi_pm_resume_ignore_current_state() might be better?  Or
> something shorter?
> 

And we are actually forcing here. As per the MHI spec, the devices has to be in
M3 state during resume. So if we allow any device to go through resume without
being in M3, that implies we are doing a force resume.

I'll use the mhi_pm_resume_force() API as you suggested.

Thanks,
Mani

> Naming is hard, sorry.
> 
> thanks,
> 
> greg k-h