[PATCH 0/2] i.MX6QP GPC fixes

[PATCH 0/2] i.MX6QP GPC fixes

[Lucas Stach]

Hi Shawn,

those 2 patches implement the necessary workaround for the PU power
domain on the i.MX6QP. Both are on top of Dong Aishengs GPC fixes
series.

I'll submit the DT change to make use of the new compatible when
the GPC driver branch has been merged upstream.

Regards,
Lucas

Lucas Stach (2):
  dt-bindings: imx-gpc: add i.MX6 QuadPlus compatible
  soc/imx: add workaround for i.MX6QP to the GPC PD driver

 .../devicetree/bindings/power/fsl,imx-gpc.txt         |  5 ++++-
 drivers/soc/imx/gpc.c                                 | 19 +++++++++++++++++++
 2 files changed, 23 insertions(+), 1 deletion(-)

-- 
2.11.0

[PATCH 1/2] dt-bindings: imx-gpc: add i.MX6 QuadPlus compatible

[Lucas Stach]

While the GPC on i.MX6QP is mostly comptible to the i.MX6Q one,
the QuadPlus requires special workarounds for hardware erratum
ERR009619.

Signed-off-by: Lucas Stach <l.stach@pengutronix.de>
---
 Documentation/devicetree/bindings/power/fsl,imx-gpc.txt | 5 ++++-
 1 file changed, 4 insertions(+), 1 deletion(-)

diff --git a/Documentation/devicetree/bindings/power/fsl,imx-gpc.txt b/Documentation/devicetree/bindings/power/fsl,imx-gpc.txt
index 416e12c6755a..181d2cc6f054 100644
--- a/Documentation/devicetree/bindings/power/fsl,imx-gpc.txt
+++ b/Documentation/devicetree/bindings/power/fsl,imx-gpc.txt
@@ -5,7 +5,10 @@ The i.MX6 General Power Control (GPC) block contains DVFS load tracking
 counters and Power Gating Control (PGC).
 
 Required properties:
-- compatible: Should be "fsl,imx6q-gpc" or "fsl,imx6sl-gpc"
+- compatible: Should be one of the following:
+  - fsl,imx6q-gpc
+  - fsl,imx6qp-gpc
+  - fsl,imx6sl-gpc
 - reg: should be register base and length as documented in the
   datasheet
 - interrupts: Should contain one interrupt specifier for the GPC interrupt
-- 
2.11.0

[PATCH 1/2] dt-bindings: imx-gpc: add i.MX6 QuadPlus compatible

[Rob Herring]

On Thu, Mar 23, 2017 at 03:44:17PM +0100, Lucas Stach wrote:
> While the GPC on i.MX6QP is mostly comptible to the i.MX6Q one,
> the QuadPlus requires special workarounds for hardware erratum
> ERR009619.
> 
> Signed-off-by: Lucas Stach <l.stach@pengutronix.de>
> ---
>  Documentation/devicetree/bindings/power/fsl,imx-gpc.txt | 5 ++++-
>  1 file changed, 4 insertions(+), 1 deletion(-)

Acked-by: Rob Herring <robh@kernel.org>

[PATCH 2/2] soc/imx: add workaround for i.MX6QP to the GPC PD driver

[Lucas Stach]

On i.MX6QP, due to hardware erratum ERR009619, the PRE clocks may be
stalled during the power up sequencing of the PU power domain. As this
may lead to a complete loss of display output, the recommended
workaround is to keep the PU domain enabled during normal system
operation.

Implement this by rejecting the domain power down request on the
affected SoC.

Signed-off-by: Lucas Stach <l.stach@pengutronix.de>
---
 drivers/soc/imx/gpc.c | 19 +++++++++++++++++++
 1 file changed, 19 insertions(+)

diff --git a/drivers/soc/imx/gpc.c b/drivers/soc/imx/gpc.c
index 4294287e5f6c..599e1e46f694 100644
--- a/drivers/soc/imx/gpc.c
+++ b/drivers/soc/imx/gpc.c
@@ -45,6 +45,7 @@ struct imx_pm_domain {
 	unsigned int reg_offs;
 	signed char cntr_pdn_bit;
 	unsigned int ipg_rate_mhz;
+	bool allow_dynamic_pd;
 };
 
 static inline struct imx_pm_domain *
@@ -59,6 +60,9 @@ static int imx6_pm_domain_power_off(struct generic_pm_domain *genpd)
 	int iso, iso2sw;
 	u32 val;
 
+	if (!pd->allow_dynamic_pd)
+		return -EBUSY;
+
 	/* Read ISO and ISO2SW power down delays */
 	regmap_read(pd->regmap, pd->reg_offs + GPC_PGC_PUPSCR_OFFS, &val);
 	iso = val & 0x3f;
@@ -255,6 +259,7 @@ static struct imx_pm_domain imx_gpc_domains[] = {
 		},
 		.reg_offs = 0x260,
 		.cntr_pdn_bit = 0,
+		.allow_dynamic_pd = true,
 	}, {
 		.base = {
 			.name = "DISPLAY",
@@ -263,23 +268,33 @@ static struct imx_pm_domain imx_gpc_domains[] = {
 		},
 		.reg_offs = 0x240,
 		.cntr_pdn_bit = 4,
+		.allow_dynamic_pd = true,
 	}
 };
 
 struct imx_gpc_dt_data {
 	int num_domains;
+	bool err009619_present;
 };
 
 static const struct imx_gpc_dt_data imx6q_dt_data = {
 	.num_domains = 2,
+	.err009619_present = false,
+};
+
+static const struct imx_gpc_dt_data imx6qp_dt_data = {
+	.num_domains = 2,
+	.err009619_present = true,
 };
 
 static const struct imx_gpc_dt_data imx6sl_dt_data = {
 	.num_domains = 3,
+	.err009619_present = false,
 };
 
 static const struct of_device_id imx_gpc_dt_ids[] = {
 	{ .compatible = "fsl,imx6q-gpc", .data = &imx6q_dt_data },
+	{ .compatible = "fsl,imx6qp-gpc", .data = &imx6qp_dt_data },
 	{ .compatible = "fsl,imx6sl-gpc", .data = &imx6sl_dt_data },
 	{ }
 };
@@ -377,6 +392,10 @@ static int imx_gpc_probe(struct platform_device *pdev)
 		return ret;
 	}
 
+	/* Disable PU power down in normal operation if ERR009619 is present */
+	if (of_id_data->err009619_present)
+		imx_gpc_domains[1].allow_dynamic_pd = false;
+
 	if (!pgc_node) {
 		ret = imx_gpc_old_dt_init(&pdev->dev, regmap,
 					  of_id_data->num_domains);
-- 
2.11.0

[PATCH 2/2] soc/imx: add workaround for i.MX6QP to the GPC PD driver

[Shawn Guo]

On Thu, Mar 23, 2017 at 03:44:18PM +0100, Lucas Stach wrote:
> On i.MX6QP, due to hardware erratum ERR009619, the PRE clocks may be
> stalled during the power up sequencing of the PU power domain. As this
> may lead to a complete loss of display output, the recommended
> workaround is to keep the PU domain enabled during normal system
> operation.
> 
> Implement this by rejecting the domain power down request on the
> affected SoC.
> 
> Signed-off-by: Lucas Stach <l.stach@pengutronix.de>
> ---
>  drivers/soc/imx/gpc.c | 19 +++++++++++++++++++
>  1 file changed, 19 insertions(+)
> 
> diff --git a/drivers/soc/imx/gpc.c b/drivers/soc/imx/gpc.c
> index 4294287e5f6c..599e1e46f694 100644
> --- a/drivers/soc/imx/gpc.c
> +++ b/drivers/soc/imx/gpc.c
> @@ -45,6 +45,7 @@ struct imx_pm_domain {
>  	unsigned int reg_offs;
>  	signed char cntr_pdn_bit;
>  	unsigned int ipg_rate_mhz;
> +	bool allow_dynamic_pd;
>  };
>  
>  static inline struct imx_pm_domain *
> @@ -59,6 +60,9 @@ static int imx6_pm_domain_power_off(struct generic_pm_domain *genpd)
>  	int iso, iso2sw;
>  	u32 val;
>  
> +	if (!pd->allow_dynamic_pd)
> +		return -EBUSY;
> +
>  	/* Read ISO and ISO2SW power down delays */
>  	regmap_read(pd->regmap, pd->reg_offs + GPC_PGC_PUPSCR_OFFS, &val);
>  	iso = val & 0x3f;
> @@ -255,6 +259,7 @@ static struct imx_pm_domain imx_gpc_domains[] = {
>  		},
>  		.reg_offs = 0x260,
>  		.cntr_pdn_bit = 0,
> +		.allow_dynamic_pd = true,
>  	}, {
>  		.base = {
>  			.name = "DISPLAY",
> @@ -263,23 +268,33 @@ static struct imx_pm_domain imx_gpc_domains[] = {
>  		},
>  		.reg_offs = 0x240,
>  		.cntr_pdn_bit = 4,
> +		.allow_dynamic_pd = true,
>  	}
>  };
>  
>  struct imx_gpc_dt_data {
>  	int num_domains;
> +	bool err009619_present;
>  };
>  
>  static const struct imx_gpc_dt_data imx6q_dt_data = {
>  	.num_domains = 2,
> +	.err009619_present = false,
> +};
> +
> +static const struct imx_gpc_dt_data imx6qp_dt_data = {
> +	.num_domains = 2,
> +	.err009619_present = true,
>  };
>  
>  static const struct imx_gpc_dt_data imx6sl_dt_data = {
>  	.num_domains = 3,
> +	.err009619_present = false,
>  };
>  
>  static const struct of_device_id imx_gpc_dt_ids[] = {
>  	{ .compatible = "fsl,imx6q-gpc", .data = &imx6q_dt_data },
> +	{ .compatible = "fsl,imx6qp-gpc", .data = &imx6qp_dt_data },
>  	{ .compatible = "fsl,imx6sl-gpc", .data = &imx6sl_dt_data },
>  	{ }
>  };
> @@ -377,6 +392,10 @@ static int imx_gpc_probe(struct platform_device *pdev)
>  		return ret;
>  	}
>  
> +	/* Disable PU power down in normal operation if ERR009619 is present */
> +	if (of_id_data->err009619_present)
> +		imx_gpc_domains[1].allow_dynamic_pd = false;

I'm uncomfortable with using hard-coded value indexing the
imx_gpc_domains array for a particular power domain.  We already have a
couple of such uses in the code, and it becomes even more.  Can we have
some defines or enumerations to help?

Shawn

> +
>  	if (!pgc_node) {
>  		ret = imx_gpc_old_dt_init(&pdev->dev, regmap,
>  					  of_id_data->num_domains);
> -- 
> 2.11.0
> 

[PATCH 2/2] soc/imx: add workaround for i.MX6QP to the GPC PD driver

[A.S. Dong]

Hi Lucas,

> -----Original Message-----
> From: Lucas Stach [mailto:l.stach at pengutronix.de]
> Sent: Thursday, March 23, 2017 10:44 PM
> To: Shawn Guo
> Cc: Fabio Estevam; A.S. Dong; devicetree at vger.kernel.org; linux-arm-
> kernel at lists.infradead.org; kernel at pengutronix.de; patchwork-
> lst at pengutronix.de
> Subject: [PATCH 2/2] soc/imx: add workaround for i.MX6QP to the GPC PD
> driver
> 
> On i.MX6QP, due to hardware erratum ERR009619, the PRE clocks may be
> stalled during the power up sequencing of the PU power domain. As this may
> lead to a complete loss of display output, the recommended workaround is
> to keep the PU domain enabled during normal system operation.
> 
> Implement this by rejecting the domain power down request on the affected
> SoC.
> 
> Signed-off-by: Lucas Stach <l.stach@pengutronix.de>
> ---
>  drivers/soc/imx/gpc.c | 19 +++++++++++++++++++
>  1 file changed, 19 insertions(+)
> 
> diff --git a/drivers/soc/imx/gpc.c b/drivers/soc/imx/gpc.c index
> 4294287e5f6c..599e1e46f694 100644
> --- a/drivers/soc/imx/gpc.c
> +++ b/drivers/soc/imx/gpc.c
> @@ -45,6 +45,7 @@ struct imx_pm_domain {
>  	unsigned int reg_offs;
>  	signed char cntr_pdn_bit;
>  	unsigned int ipg_rate_mhz;
> +	bool allow_dynamic_pd;
>  };
> 
>  static inline struct imx_pm_domain *
> @@ -59,6 +60,9 @@ static int imx6_pm_domain_power_off(struct
> generic_pm_domain *genpd)
>  	int iso, iso2sw;
>  	u32 val;
> 
> +	if (!pd->allow_dynamic_pd)
> +		return -EBUSY;
> +
>  	/* Read ISO and ISO2SW power down delays */
>  	regmap_read(pd->regmap, pd->reg_offs + GPC_PGC_PUPSCR_OFFS, &val);
>  	iso = val & 0x3f;
> @@ -255,6 +259,7 @@ static struct imx_pm_domain imx_gpc_domains[] = {
>  		},
>  		.reg_offs = 0x260,
>  		.cntr_pdn_bit = 0,
> +		.allow_dynamic_pd = true,
>  	}, {
>  		.base = {
>  			.name = "DISPLAY",
> @@ -263,23 +268,33 @@ static struct imx_pm_domain imx_gpc_domains[] = {
>  		},
>  		.reg_offs = 0x240,
>  		.cntr_pdn_bit = 4,
> +		.allow_dynamic_pd = true,

Just a Nitpick, besides shawn's comment in another mail,
if we use a domain flag like IMX_PD_ALWAYS_ON set by SoC errata flag,
then probably we can save all the default allow_dynamic_pd lines.
This also release the life when adding new domains.

Otherwise, the two patch seems good to me.

Regards
Dong Aisheng

>  	}
>  };
> 
>  struct imx_gpc_dt_data {
>  	int num_domains;
> +	bool err009619_present;
>  };
> 
>  static const struct imx_gpc_dt_data imx6q_dt_data = {
>  	.num_domains = 2,
> +	.err009619_present = false,
> +};
> +
> +static const struct imx_gpc_dt_data imx6qp_dt_data = {
> +	.num_domains = 2,
> +	.err009619_present = true,
>  };
> 
>  static const struct imx_gpc_dt_data imx6sl_dt_data = {
>  	.num_domains = 3,
> +	.err009619_present = false,
>  };
> 
>  static const struct of_device_id imx_gpc_dt_ids[] = {
>  	{ .compatible = "fsl,imx6q-gpc", .data = &imx6q_dt_data },
> +	{ .compatible = "fsl,imx6qp-gpc", .data = &imx6qp_dt_data },
>  	{ .compatible = "fsl,imx6sl-gpc", .data = &imx6sl_dt_data },
>  	{ }
>  };
> @@ -377,6 +392,10 @@ static int imx_gpc_probe(struct platform_device *pdev)
>  		return ret;
>  	}
> 
> +	/* Disable PU power down in normal operation if ERR009619 is present
> */
> +	if (of_id_data->err009619_present)
> +		imx_gpc_domains[1].allow_dynamic_pd = false;
> +
>  	if (!pgc_node) {
>  		ret = imx_gpc_old_dt_init(&pdev->dev, regmap,
>  					  of_id_data->num_domains);
> --
> 2.11.0

[PATCH 2/2] soc/imx: add workaround for i.MX6QP to the GPC PD driver

[Lucas Stach]

Am Mittwoch, den 29.03.2017, 16:08 +0000 schrieb A.S. Dong:
> Hi Lucas,
> 
> > -----Original Message-----
> > From: Lucas Stach [mailto:l.stach at pengutronix.de]
> > Sent: Thursday, March 23, 2017 10:44 PM
> > To: Shawn Guo
> > Cc: Fabio Estevam; A.S. Dong; devicetree at vger.kernel.org; linux-arm-
> > kernel at lists.infradead.org; kernel at pengutronix.de; patchwork-
> > lst at pengutronix.de
> > Subject: [PATCH 2/2] soc/imx: add workaround for i.MX6QP to the GPC PD
> > driver
> > 
> > On i.MX6QP, due to hardware erratum ERR009619, the PRE clocks may be
> > stalled during the power up sequencing of the PU power domain. As this may
> > lead to a complete loss of display output, the recommended workaround is
> > to keep the PU domain enabled during normal system operation.
> > 
> > Implement this by rejecting the domain power down request on the affected
> > SoC.
> > 
> > Signed-off-by: Lucas Stach <l.stach@pengutronix.de>
> > ---
> >  drivers/soc/imx/gpc.c | 19 +++++++++++++++++++
> >  1 file changed, 19 insertions(+)
> > 
> > diff --git a/drivers/soc/imx/gpc.c b/drivers/soc/imx/gpc.c index
> > 4294287e5f6c..599e1e46f694 100644
> > --- a/drivers/soc/imx/gpc.c
> > +++ b/drivers/soc/imx/gpc.c
> > @@ -45,6 +45,7 @@ struct imx_pm_domain {
> >  	unsigned int reg_offs;
> >  	signed char cntr_pdn_bit;
> >  	unsigned int ipg_rate_mhz;
> > +	bool allow_dynamic_pd;
> >  };
> > 
> >  static inline struct imx_pm_domain *
> > @@ -59,6 +60,9 @@ static int imx6_pm_domain_power_off(struct
> > generic_pm_domain *genpd)
> >  	int iso, iso2sw;
> >  	u32 val;
> > 
> > +	if (!pd->allow_dynamic_pd)
> > +		return -EBUSY;
> > +
> >  	/* Read ISO and ISO2SW power down delays */
> >  	regmap_read(pd->regmap, pd->reg_offs + GPC_PGC_PUPSCR_OFFS, &val);
> >  	iso = val & 0x3f;
> > @@ -255,6 +259,7 @@ static struct imx_pm_domain imx_gpc_domains[] = {
> >  		},
> >  		.reg_offs = 0x260,
> >  		.cntr_pdn_bit = 0,
> > +		.allow_dynamic_pd = true,
> >  	}, {
> >  		.base = {
> >  			.name = "DISPLAY",
> > @@ -263,23 +268,33 @@ static struct imx_pm_domain imx_gpc_domains[] = {
> >  		},
> >  		.reg_offs = 0x240,
> >  		.cntr_pdn_bit = 4,
> > +		.allow_dynamic_pd = true,
> 
> Just a Nitpick, besides shawn's comment in another mail,
> if we use a domain flag like IMX_PD_ALWAYS_ON set by SoC errata flag,
> then probably we can save all the default allow_dynamic_pd lines.
> This also release the life when adding new domains.

If other things like that show up we can certainly switch to a flags
field. In the meantime I like that the current style is more explicit
about it.

Regards,
Lucas

[PATCH 2/2] soc/imx: add workaround for i.MX6QP to the GPC PD driver

[Lothar Waßmann]

Hi,

On Wed, 29 Mar 2017 18:13:46 +0200 Lucas Stach wrote:
> Am Mittwoch, den 29.03.2017, 16:08 +0000 schrieb A.S. Dong:
> > Hi Lucas,
> > 
> > > -----Original Message-----
> > > From: Lucas Stach [mailto:l.stach at pengutronix.de]
> > > Sent: Thursday, March 23, 2017 10:44 PM
> > > To: Shawn Guo
> > > Cc: Fabio Estevam; A.S. Dong; devicetree at vger.kernel.org; linux-arm-
> > > kernel at lists.infradead.org; kernel at pengutronix.de; patchwork-
> > > lst at pengutronix.de
> > > Subject: [PATCH 2/2] soc/imx: add workaround for i.MX6QP to the GPC PD
> > > driver
> > > 
> > > On i.MX6QP, due to hardware erratum ERR009619, the PRE clocks may be
> > > stalled during the power up sequencing of the PU power domain. As this may
> > > lead to a complete loss of display output, the recommended workaround is
> > > to keep the PU domain enabled during normal system operation.
> > > 
> > > Implement this by rejecting the domain power down request on the affected
> > > SoC.
> > > 
> > > Signed-off-by: Lucas Stach <l.stach@pengutronix.de>
> > > ---
> > >  drivers/soc/imx/gpc.c | 19 +++++++++++++++++++
> > >  1 file changed, 19 insertions(+)
> > > 
> > > diff --git a/drivers/soc/imx/gpc.c b/drivers/soc/imx/gpc.c index
> > > 4294287e5f6c..599e1e46f694 100644
> > > --- a/drivers/soc/imx/gpc.c
> > > +++ b/drivers/soc/imx/gpc.c
> > > @@ -45,6 +45,7 @@ struct imx_pm_domain {
> > >  	unsigned int reg_offs;
> > >  	signed char cntr_pdn_bit;
> > >  	unsigned int ipg_rate_mhz;
> > > +	bool allow_dynamic_pd;
> > >  };
> > > 
> > >  static inline struct imx_pm_domain *
> > > @@ -59,6 +60,9 @@ static int imx6_pm_domain_power_off(struct
> > > generic_pm_domain *genpd)
> > >  	int iso, iso2sw;
> > >  	u32 val;
> > > 
> > > +	if (!pd->allow_dynamic_pd)
> > > +		return -EBUSY;
> > > +
> > >  	/* Read ISO and ISO2SW power down delays */
> > >  	regmap_read(pd->regmap, pd->reg_offs + GPC_PGC_PUPSCR_OFFS, &val);
> > >  	iso = val & 0x3f;
> > > @@ -255,6 +259,7 @@ static struct imx_pm_domain imx_gpc_domains[] = {
> > >  		},
> > >  		.reg_offs = 0x260,
> > >  		.cntr_pdn_bit = 0,
> > > +		.allow_dynamic_pd = true,
> > >  	}, {
> > >  		.base = {
> > >  			.name = "DISPLAY",
> > > @@ -263,23 +268,33 @@ static struct imx_pm_domain imx_gpc_domains[] = {
> > >  		},
> > >  		.reg_offs = 0x240,
> > >  		.cntr_pdn_bit = 4,
> > > +		.allow_dynamic_pd = true,
> > 
> > Just a Nitpick, besides shawn's comment in another mail,
> > if we use a domain flag like IMX_PD_ALWAYS_ON set by SoC errata flag,
> > then probably we can save all the default allow_dynamic_pd lines.
> > This also release the life when adding new domains.
> 
> If other things like that show up we can certainly switch to a flags
> field. In the meantime I like that the current style is more explicit
> about it.
> 
Since the purpose of the driver is all about dynamic power management,
I would rather add a 'disable_dynamic_pd' for the case(s) where it's
not allowed, rather than having to state the obvious over and over
again.


Lothar Wa?mann

[PATCH 2/2] soc/imx: add workaround for i.MX6QP to the GPC PD driver

[Lucas Stach]

Am Donnerstag, den 30.03.2017, 09:08 +0200 schrieb Lothar Wa?mann:
> Hi,
> 
> On Wed, 29 Mar 2017 18:13:46 +0200 Lucas Stach wrote:
> > Am Mittwoch, den 29.03.2017, 16:08 +0000 schrieb A.S. Dong:
> > > Hi Lucas,
> > > 
> > > > -----Original Message-----
> > > > From: Lucas Stach [mailto:l.stach at pengutronix.de]
> > > > Sent: Thursday, March 23, 2017 10:44 PM
> > > > To: Shawn Guo
> > > > Cc: Fabio Estevam; A.S. Dong; devicetree at vger.kernel.org; linux-arm-
> > > > kernel at lists.infradead.org; kernel at pengutronix.de; patchwork-
> > > > lst at pengutronix.de
> > > > Subject: [PATCH 2/2] soc/imx: add workaround for i.MX6QP to the GPC PD
> > > > driver
> > > > 
> > > > On i.MX6QP, due to hardware erratum ERR009619, the PRE clocks may be
> > > > stalled during the power up sequencing of the PU power domain. As this may
> > > > lead to a complete loss of display output, the recommended workaround is
> > > > to keep the PU domain enabled during normal system operation.
> > > > 
> > > > Implement this by rejecting the domain power down request on the affected
> > > > SoC.
> > > > 
> > > > Signed-off-by: Lucas Stach <l.stach@pengutronix.de>
> > > > ---
> > > >  drivers/soc/imx/gpc.c | 19 +++++++++++++++++++
> > > >  1 file changed, 19 insertions(+)
> > > > 
> > > > diff --git a/drivers/soc/imx/gpc.c b/drivers/soc/imx/gpc.c index
> > > > 4294287e5f6c..599e1e46f694 100644
> > > > --- a/drivers/soc/imx/gpc.c
> > > > +++ b/drivers/soc/imx/gpc.c
> > > > @@ -45,6 +45,7 @@ struct imx_pm_domain {
> > > >  	unsigned int reg_offs;
> > > >  	signed char cntr_pdn_bit;
> > > >  	unsigned int ipg_rate_mhz;
> > > > +	bool allow_dynamic_pd;
> > > >  };
> > > > 
> > > >  static inline struct imx_pm_domain *
> > > > @@ -59,6 +60,9 @@ static int imx6_pm_domain_power_off(struct
> > > > generic_pm_domain *genpd)
> > > >  	int iso, iso2sw;
> > > >  	u32 val;
> > > > 
> > > > +	if (!pd->allow_dynamic_pd)
> > > > +		return -EBUSY;
> > > > +
> > > >  	/* Read ISO and ISO2SW power down delays */
> > > >  	regmap_read(pd->regmap, pd->reg_offs + GPC_PGC_PUPSCR_OFFS, &val);
> > > >  	iso = val & 0x3f;
> > > > @@ -255,6 +259,7 @@ static struct imx_pm_domain imx_gpc_domains[] = {
> > > >  		},
> > > >  		.reg_offs = 0x260,
> > > >  		.cntr_pdn_bit = 0,
> > > > +		.allow_dynamic_pd = true,
> > > >  	}, {
> > > >  		.base = {
> > > >  			.name = "DISPLAY",
> > > > @@ -263,23 +268,33 @@ static struct imx_pm_domain imx_gpc_domains[] = {
> > > >  		},
> > > >  		.reg_offs = 0x240,
> > > >  		.cntr_pdn_bit = 4,
> > > > +		.allow_dynamic_pd = true,
> > > 
> > > Just a Nitpick, besides shawn's comment in another mail,
> > > if we use a domain flag like IMX_PD_ALWAYS_ON set by SoC errata flag,
> > > then probably we can save all the default allow_dynamic_pd lines.
> > > This also release the life when adding new domains.
> > 
> > If other things like that show up we can certainly switch to a flags
> > field. In the meantime I like that the current style is more explicit
> > about it.
> > 
> Since the purpose of the driver is all about dynamic power management,
> I would rather add a 'disable_dynamic_pd' for the case(s) where it's
> not allowed, rather than having to state the obvious over and over
> again.

Convincing argument. I'll change that in v2.

Regards,
Lucas