[PATCH v6] OMAP2+: PM: omap device: API's for handling mstandby mode

[PATCH v6] OMAP2+: PM: omap device: API's for handling mstandby mode

[Adrian Hunter]

On 01/12/10 20:47, ext G, Manjunath Kondaiah wrote:
> Certain errata in OMAP2+ processors will require forcing
> master standby to "no standby" mode before completing on going
> operation. Without this, the results will be unpredictable.
>
> Since current implementation of PM run time framework does not support
> changing sysconfig settings during middle of the on going operation,
> these API's will support the same. One API will force the device's
> sysconfig mstandby mode settings to "no standby" and other API will
> release "no standby" mode and sets it to "smart standby" or "no
> standby? depending on HWMOD_SWSUP_MSTANDBY value.
>
> These API's should be used by device drivers only incase of
> erratum applicable to their modules if there is no other methods
> to resolve.
>
> These API's are required for multiple DMA errata which require
> putting DMA controller in no mstandby mode before stopping dma.
>
> The applicable errata:
> 1. Erratum ID: i557(Applicable for omap36xx all ES versions)
> The channel hangs when the Pause bit (DMA4_CDPi [7] ) is cleared
> through config port while in Standby.
>
> 2. Erratum ID: i541
> sDMA FIFO draining does not finish. Applicable to all omap2+ except
> omap4.
>
> 3. Erratum ID:i88
> The sDMA to be put in no mstandby mode before disabling the channel
> after completing the data transfer operation.
> Applicable only for OMAP3430 ES1.0
>
> Also fixes typo HWMOD_SWSUP_MSTDBY to HWMOD_SWSUP_MSTANDBY in
> omap_hwmod.h
>
> Signed-off-by: G, Manjunath Kondaiah<manjugk@ti.com>
> Cc: Kevin Hilman<khilman@deeprootsystems.com>
> Cc: Paul Walmsley<paul@pwsan.com>
> Cc: linux-arm-kernel at lists.infradead.org
>
> ---
> Change summary:
> v3: Review comments incorporated for:
> https://patchwork.kernel.org/patch/282212/
> v4: added mutex changes
> v5: typo fixes for errata and erratum
> v6: fixed oh increment bug and also mutex(missing in v5)
>
>   arch/arm/mach-omap2/omap_hwmod.c              |   47 +++++++++++++++++++-
>   arch/arm/plat-omap/include/plat/omap_device.h |    2 +
>   arch/arm/plat-omap/include/plat/omap_hwmod.h  |    4 +-
>   arch/arm/plat-omap/omap_device.c              |   60 +++++++++++++++++++++++++
>   4 files changed, 111 insertions(+), 2 deletions(-)
>

Does not seem to handle overlapping requests to set/clear
midle mode.

Also, there is a mutex but don't these functions get called
sometimes in interrupt context?


> diff --git a/arch/arm/mach-omap2/omap_hwmod.c b/arch/arm/mach-omap2/omap_hwmod.c
> index 5a30658..023bca9 100644
> --- a/arch/arm/mach-omap2/omap_hwmod.c
> +++ b/arch/arm/mach-omap2/omap_hwmod.c
> @@ -1427,6 +1427,52 @@ int omap_hwmod_set_slave_idlemode(struct omap_hwmod *oh, u8 idlemode)
>   }
>
>   /**
> + * omap_hwmod_set_master_standbymode - set the hwmod's OCP mstandby mode
> + * @oh: struct omap_hwmod *
> + * @midlemode: flag to set mstandby to either "no standby" or "smart standby"
> + *
> + * Sets the IP block's OCP mstandby mode in hardware, and updates our
> + * local copy.  Intended to be used by drivers that have some erratum
> + * that requires direct manipulation of the MIDLEMODE bits.  Returns
> + * -EINVAL if @oh is null, or passes along the return value from
> + * _set_master_standbymode().
> + *
> + * Any users of this function should be scrutinized carefully.
> + */
> +int omap_hwmod_set_master_standbymode(struct omap_hwmod *oh, u8 idlemode)
> +{
> +	u32 v;
> +	u8 sf;
> +	int retval = 0;
> +
> +	if (!oh)
> +		return -EINVAL;
> +
> +	if (!oh->class->sysc)
> +		return -EINVAL;
> +
> +	mutex_lock(&omap_hwmod_mutex);
> +
> +	v = oh->_sysc_cache;
> +	sf = oh->class->sysc->sysc_flags;
> +
> +	if (sf&  SYSC_HAS_MIDLEMODE) {
> +		if (idlemode)
> +			idlemode = HWMOD_IDLEMODE_NO;
> +		else
> +			idlemode = (oh->flags&  HWMOD_SWSUP_MSTANDBY) ?
> +				HWMOD_IDLEMODE_NO : HWMOD_IDLEMODE_SMART;
> +	}
> +	retval = _set_master_standbymode(oh, idlemode,&v);
> +	if (!retval)
> +		_write_sysconfig(v, oh);
> +
> +	mutex_unlock(&omap_hwmod_mutex);
> +
> +	return retval;
> +}
> +
> +/**
>    * omap_hwmod_register - register a struct omap_hwmod
>    * @oh: struct omap_hwmod *
>    *
> @@ -2116,4 +2162,3 @@ int omap_hwmod_for_each_by_class(const char *classname,
>
>   	return ret;
>   }
> -
> diff --git a/arch/arm/plat-omap/include/plat/omap_device.h b/arch/arm/plat-omap/include/plat/omap_device.h
> index 28e2d1a..4fbf7c1 100644
> --- a/arch/arm/plat-omap/include/plat/omap_device.h
> +++ b/arch/arm/plat-omap/include/plat/omap_device.h
> @@ -116,6 +116,8 @@ int omap_device_enable_hwmods(struct omap_device *od);
>   int omap_device_disable_clocks(struct omap_device *od);
>   int omap_device_enable_clocks(struct omap_device *od);
>
> +int omap_device_require_no_mstandby(struct platform_device *pdev);
> +int omap_device_release_no_mstandby(struct platform_device *pdev);
>
>   /*
>    * Entries should be kept in latency order ascending
> diff --git a/arch/arm/plat-omap/include/plat/omap_hwmod.h b/arch/arm/plat-omap/include/plat/omap_hwmod.h
> index 7eaa8ed..c7ff65a 100644
> --- a/arch/arm/plat-omap/include/plat/omap_hwmod.h
> +++ b/arch/arm/plat-omap/include/plat/omap_hwmod.h
> @@ -354,7 +354,7 @@ struct omap_hwmod_omap4_prcm {
>    *
>    * HWMOD_SWSUP_SIDLE: omap_hwmod code should manually bring module in and out
>    *     of idle, rather than relying on module smart-idle
> - * HWMOD_SWSUP_MSTDBY: omap_hwmod code should manually bring module in and out
> + * HWMOD_SWSUP_MSTANDBY: omap_hwmod code should manually bring module in and out
>    *     of standby, rather than relying on module smart-standby
>    * HWMOD_INIT_NO_RESET: don't reset this module at boot - important for
>    *     SDRAM controller, etc.
> @@ -526,6 +526,8 @@ int omap_hwmod_disable_clocks(struct omap_hwmod *oh);
>
>   int omap_hwmod_set_slave_idlemode(struct omap_hwmod *oh, u8 idlemode);
>
> +int omap_hwmod_set_master_standbymode(struct omap_hwmod *oh, u8 idlemode);
> +
>   int omap_hwmod_reset(struct omap_hwmod *oh);
>   void omap_hwmod_ocp_barrier(struct omap_hwmod *oh);
>
> diff --git a/arch/arm/plat-omap/omap_device.c b/arch/arm/plat-omap/omap_device.c
> index abe933c..2d42ab2 100644
> --- a/arch/arm/plat-omap/omap_device.c
> +++ b/arch/arm/plat-omap/omap_device.c
> @@ -584,6 +584,66 @@ int omap_device_idle(struct platform_device *pdev)
>   }
>
>   /**
> + * omap_device_require_no_mstandby - set no mstandby mode of an omap_device
> + * @od: struct omap_device * to idle
> + *
> + * Sets the IP block's OCP master standby to no mstandby mode in hardware.
> + *
> + * Intended to be used by drivers that have some erratum that requires direct
> + * manipulation of the MSTANDBYMODE bits. Returns -EINVAL if the
> + * omap_device is not currently enabled or passes along the return value
> + * of omap_hwmod_set_master_standbymode().
> + */
> +int omap_device_require_no_mstandby(struct platform_device *pdev)
> +{
> +	int ret  = 0, i;
> +	struct omap_device *od;
> +
> +	od = _find_by_pdev(pdev);
> +	if (od->_state != OMAP_DEVICE_STATE_ENABLED) {
> +		WARN(1, "omap_device: %s.%d: %s() called from invalid state %d\n",
> +		     od->pdev.name, od->pdev.id, __func__, od->_state);
> +		return -EINVAL;
> +	}
> +
> +	for (i = 0; i<  od->hwmods_cnt; i++)
> +		ret = omap_hwmod_set_master_standbymode(od->hwmods[i], true);
> +
> +	return ret;
> +}
> +
> +/**
> + * omap_device_release_no_mstandby - releases no mstandby mode of an omap_device
> + * @od: struct omap_device * to idle
> + *
> + * Release no mstandby mode and sets the master standby to either no standby or
> + * smart standby in IP block's OCP in hardware depending on status of the flag
> + * HWMOD_SWSUP_MSTANDBY.
> + *
> + * Intended to be used by drivers that have some erratum that requires direct
> + * manipulation of the MSTANDBYMODE bits. Returns -EINVAL if the
> + * omap_device is not currently enabled or passes along the return value
> + * of omap_hwmod_set_master_standbymode().
> + */
> +int omap_device_release_no_mstandby(struct platform_device *pdev)
> +{
> +	int ret  = 0, i;
> +	struct omap_device *od;
> +
> +	od = _find_by_pdev(pdev);
> +	if (od->_state != OMAP_DEVICE_STATE_ENABLED) {
> +		WARN(1, "omap_device: %s.%d: %s() called from invalid state %d\n",
> +		     od->pdev.name, od->pdev.id, __func__, od->_state);
> +		return -EINVAL;
> +	}
> +
> +	for (i = 0; i<  od->hwmods_cnt; i++)
> +		ret = omap_hwmod_set_master_standbymode(od->hwmods[i], false);
> +
> +	return ret;
> +}
> +
> +/**
>    * omap_device_shutdown - shut down an omap_device
>    * @od: struct omap_device * to shut down
>    *

[PATCH v6] OMAP2+: PM: omap device: API's for handling mstandby mode

[G, Manjunath Kondaiah]

* Adrian Hunter <adrian.hunter@nokia.com> [2010-12-01 22:16:44 +0200]:

> On 01/12/10 20:47, ext G, Manjunath Kondaiah wrote:
> >Certain errata in OMAP2+ processors will require forcing
> >master standby to "no standby" mode before completing on going
> >operation. Without this, the results will be unpredictable.
> >
> >Since current implementation of PM run time framework does not support
> >changing sysconfig settings during middle of the on going operation,
> >these API's will support the same. One API will force the device's
> >sysconfig mstandby mode settings to "no standby" and other API will
> >release "no standby" mode and sets it to "smart standby" or "no
> >standby? depending on HWMOD_SWSUP_MSTANDBY value.
> >
> >These API's should be used by device drivers only incase of
> >erratum applicable to their modules if there is no other methods
> >to resolve.
> >
> >These API's are required for multiple DMA errata which require
> >putting DMA controller in no mstandby mode before stopping dma.
> >
> >The applicable errata:
> >1. Erratum ID: i557(Applicable for omap36xx all ES versions)
> >The channel hangs when the Pause bit (DMA4_CDPi [7] ) is cleared
> >through config port while in Standby.
> >
> >2. Erratum ID: i541
> >sDMA FIFO draining does not finish. Applicable to all omap2+ except
> >omap4.
> >
> >3. Erratum ID:i88
> >The sDMA to be put in no mstandby mode before disabling the channel
> >after completing the data transfer operation.
> >Applicable only for OMAP3430 ES1.0
> >
> >Also fixes typo HWMOD_SWSUP_MSTDBY to HWMOD_SWSUP_MSTANDBY in
> >omap_hwmod.h
> >
> >Signed-off-by: G, Manjunath Kondaiah<manjugk@ti.com>
> >Cc: Kevin Hilman<khilman@deeprootsystems.com>
> >Cc: Paul Walmsley<paul@pwsan.com>
> >Cc: linux-arm-kernel at lists.infradead.org
> >
> >---
> >Change summary:
> >v3: Review comments incorporated for:
> >https://patchwork.kernel.org/patch/282212/
> >v4: added mutex changes
> >v5: typo fixes for errata and erratum
> >v6: fixed oh increment bug and also mutex(missing in v5)
> >
> >  arch/arm/mach-omap2/omap_hwmod.c              |   47 +++++++++++++++++++-
> >  arch/arm/plat-omap/include/plat/omap_device.h |    2 +
> >  arch/arm/plat-omap/include/plat/omap_hwmod.h  |    4 +-
> >  arch/arm/plat-omap/omap_device.c              |   60 +++++++++++++++++++++++++
> >  4 files changed, 111 insertions(+), 2 deletions(-)
> >
> 
> Does not seem to handle overlapping requests to set/clear
> midle mode.

We can have use count so that we can avoid register access during
middle of operation if mstandby mode is already changed to "no standby"
mode.

> 
> Also, there is a mutex but don't these functions get called
> sometimes in interrupt context?

I feel we should leave this to caller of these API's. The caller should 
make sure that, he disables interrupts if it is used in interrupt context.

-Manjunath

[PATCH v6] OMAP2+: PM: omap device: API's for handling mstandby mode

[Adrian Hunter]

On 02/12/10 08:12, ext G, Manjunath Kondaiah wrote:
> * Adrian Hunter<adrian.hunter@nokia.com>  [2010-12-01 22:16:44 +0200]:
>
>> On 01/12/10 20:47, ext G, Manjunath Kondaiah wrote:
>>> Certain errata in OMAP2+ processors will require forcing
>>> master standby to "no standby" mode before completing on going
>>> operation. Without this, the results will be unpredictable.
>>>
>>> Since current implementation of PM run time framework does not support
>>> changing sysconfig settings during middle of the on going operation,
>>> these API's will support the same. One API will force the device's
>>> sysconfig mstandby mode settings to "no standby" and other API will
>>> release "no standby" mode and sets it to "smart standby" or "no
>>> standby? depending on HWMOD_SWSUP_MSTANDBY value.
>>>
>>> These API's should be used by device drivers only incase of
>>> erratum applicable to their modules if there is no other methods
>>> to resolve.
>>>
>>> These API's are required for multiple DMA errata which require
>>> putting DMA controller in no mstandby mode before stopping dma.
>>>
>>> The applicable errata:
>>> 1. Erratum ID: i557(Applicable for omap36xx all ES versions)
>>> The channel hangs when the Pause bit (DMA4_CDPi [7] ) is cleared
>>> through config port while in Standby.
>>>
>>> 2. Erratum ID: i541
>>> sDMA FIFO draining does not finish. Applicable to all omap2+ except
>>> omap4.
>>>
>>> 3. Erratum ID:i88
>>> The sDMA to be put in no mstandby mode before disabling the channel
>>> after completing the data transfer operation.
>>> Applicable only for OMAP3430 ES1.0
>>>
>>> Also fixes typo HWMOD_SWSUP_MSTDBY to HWMOD_SWSUP_MSTANDBY in
>>> omap_hwmod.h
>>>
>>> Signed-off-by: G, Manjunath Kondaiah<manjugk@ti.com>
>>> Cc: Kevin Hilman<khilman@deeprootsystems.com>
>>> Cc: Paul Walmsley<paul@pwsan.com>
>>> Cc: linux-arm-kernel at lists.infradead.org
>>>
>>> ---
>>> Change summary:
>>> v3: Review comments incorporated for:
>>> https://patchwork.kernel.org/patch/282212/
>>> v4: added mutex changes
>>> v5: typo fixes for errata and erratum
>>> v6: fixed oh increment bug and also mutex(missing in v5)
>>>
>>>   arch/arm/mach-omap2/omap_hwmod.c              |   47 +++++++++++++++++++-
>>>   arch/arm/plat-omap/include/plat/omap_device.h |    2 +
>>>   arch/arm/plat-omap/include/plat/omap_hwmod.h  |    4 +-
>>>   arch/arm/plat-omap/omap_device.c              |   60 +++++++++++++++++++++++++
>>>   4 files changed, 111 insertions(+), 2 deletions(-)
>>>
>>
>> Does not seem to handle overlapping requests to set/clear
>> midle mode.
>
> We can have use count so that we can avoid register access during
> middle of operation if mstandby mode is already changed to "no standby"
> mode.

Exactly

>
>>
>> Also, there is a mutex but don't these functions get called
>> sometimes in interrupt context?
>
> I feel we should leave this to caller of these API's. The caller should
> make sure that, he disables interrupts if it is used in interrupt context.

How does that help?  It seems to me that a mutex cannot be used for this.

>
> -Manjunath
> --
> To unsubscribe from this list: send the line "unsubscribe linux-omap" in
> the body of a message to majordomo at vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html
>

[PATCH v6] OMAP2+: PM: omap device: API's for handling mstandby mode

[Cousson, Benoit]

Hi Adrian and Manju,

On 12/2/2010 8:58 AM, Adrian Hunter wrote:
> On 02/12/10 08:12, ext G, Manjunath Kondaiah wrote:
>> * Adrian Hunter<adrian.hunter@nokia.com>   [2010-12-01 22:16:44 +0200]:
>>> On 01/12/10 20:47, ext G, Manjunath Kondaiah wrote:

[...]

>>> Also, there is a mutex but don't these functions get called
>>> sometimes in interrupt context?
>>
>> I feel we should leave this to caller of these API's. The caller should
>> make sure that, he disables interrupts if it is used in interrupt context.
>
> How does that help?  It seems to me that a mutex cannot be used for this.

That mutex is the one that was protecting the hwmod list and should not 
be used, we do have a per-hwmod mutex.
But it will disappear soon, Paul already submitted a patch to replace it 
by a spinlock (8297db801c424f95ef3ebfbd28e62b94798d35dd)

Please note that the goal of this API is to not be used at all :-)
It is there because of dirty HW bugs in several IPs.
So I'm fine to limit its usability and to prevent its usage everywhere.
Thanks to the spinlock it will anyway no longer be an issue.

Thanks,
Benoit