At the moment, enabling this driver without the SNVS RTC driver
being active will hang the kernel as soon as the power button
is pressed.
The reason is that in that case the SNVS isn't enabled, and
any attempt to read the SNVS registers will simply hang forever.
Ensure the clock is enabled (during the interrupt handler) to
make this driver work.
Also see commit 7f8993995410 ("drivers/rtc/rtc-snvs: add clock support")
and commit edb190cb1734
("rtc: snvs: make sure clock is enabled for interrupt handle")
for similar updates to the snvs rtc driver.
Signed-off-by: André Draszik <[email protected]>
Cc: Anson Huang <[email protected]>
Cc: Dmitry Torokhov <[email protected]>
Cc: "Horia Geantă" <[email protected]>
Cc: Aymen Sghaier <[email protected]>
Cc: Herbert Xu <[email protected]>
Cc: "David S. Miller" <[email protected]>
Cc: Rob Herring <[email protected]>
Cc: Mark Rutland <[email protected]>
Cc: [email protected]
Cc: [email protected]
Cc: [email protected]
---
drivers/input/keyboard/snvs_pwrkey.c | 27 +++++++++++++++++++++++++++
1 file changed, 27 insertions(+)
diff --git a/drivers/input/keyboard/snvs_pwrkey.c b/drivers/input/keyboard/snvs_pwrkey.c
index 2f5e3ab5ed63..c29711d8735c 100644
--- a/drivers/input/keyboard/snvs_pwrkey.c
+++ b/drivers/input/keyboard/snvs_pwrkey.c
@@ -16,6 +16,7 @@
#include <linux/of_address.h>
#include <linux/platform_device.h>
#include <linux/pm_wakeirq.h>
+#include <linux/clk.h>
#include <linux/mfd/syscon.h>
#include <linux/regmap.h>
@@ -38,6 +39,7 @@ struct pwrkey_drv_data {
int wakeup;
struct timer_list check_timer;
struct input_dev *input;
+ struct clk *clk;
u8 minor_rev;
};
@@ -72,6 +74,9 @@ static irqreturn_t imx_snvs_pwrkey_interrupt(int irq, void *dev_id)
struct input_dev *input = pdata->input;
u32 lp_status;
+ if (pdata->clk)
+ clk_enable(pdata->clk);
+
pm_wakeup_event(input->dev.parent, 0);
regmap_read(pdata->snvs, SNVS_LPSR_REG, &lp_status);
@@ -96,6 +101,9 @@ static irqreturn_t imx_snvs_pwrkey_interrupt(int irq, void *dev_id)
/* clear SPO status */
regmap_write(pdata->snvs, SNVS_LPSR_REG, SNVS_LPSR_SPO);
+ if (pdata->clk)
+ clk_disable(pdata->clk);
+
return IRQ_HANDLED;
}
@@ -140,6 +148,25 @@ static int imx_snvs_pwrkey_probe(struct platform_device *pdev)
if (pdata->irq < 0)
return -EINVAL;
+ pdata->clk = devm_clk_get(&pdev->dev, "snvs-pwrkey");
+ if (IS_ERR(pdata->clk)) {
+ pdata->clk = NULL;
+ } else {
+ error = clk_prepare_enable(pdata->clk);
+ if (error) {
+ dev_err(&pdev->dev,
+ "Could not prepare or enable the snvs clock\n");
+ return error;
+ }
+ error = devm_add_action_or_reset(&pdev->dev,
+ (void(*)(void *))clk_disable_unprepare,
+ pdata->clk);
+ if (error) {
+ dev_err(&pdev->dev, "failed to add reset action on 'snvs-pwrkey'");
+ return error;
+ }
+ }
+
regmap_read(pdata->snvs, SNVS_HPVIDR1_REG, &vid);
pdata->minor_rev = vid & 0xff;
--
2.23.0.rc1
On Fri, Feb 07, 2020 at 08:10:22AM +0000, Horia Geanta wrote:
> On 1/30/2020 10:45 PM, André Draszik wrote:
> > At the moment, enabling this driver without the SNVS RTC driver
> > being active will hang the kernel as soon as the power button
> > is pressed.
> >
> > The reason is that in that case the SNVS isn't enabled, and
> > any attempt to read the SNVS registers will simply hang forever.
> >
> > Ensure the clock is enabled (during the interrupt handler) to
> > make this driver work.
> >
> > Also see commit 7f8993995410 ("drivers/rtc/rtc-snvs: add clock support")
> > and commit edb190cb1734
> > ("rtc: snvs: make sure clock is enabled for interrupt handle")
> > for similar updates to the snvs rtc driver.
> >
> > Signed-off-by: André Draszik <[email protected]>
> > Cc: Anson Huang <[email protected]>
> > Cc: Dmitry Torokhov <[email protected]>
> > Cc: "Horia Geantă" <[email protected]>
> > Cc: Aymen Sghaier <[email protected]>
> > Cc: Herbert Xu <[email protected]>
> > Cc: "David S. Miller" <[email protected]>
> > Cc: Rob Herring <[email protected]>
> > Cc: Mark Rutland <[email protected]>
> > Cc: [email protected]
> > Cc: [email protected]
> > Cc: [email protected]
> > ---
> > drivers/input/keyboard/snvs_pwrkey.c | 27 +++++++++++++++++++++++++++
> > 1 file changed, 27 insertions(+)
> >
> > diff --git a/drivers/input/keyboard/snvs_pwrkey.c b/drivers/input/keyboard/snvs_pwrkey.c
> > index 2f5e3ab5ed63..c29711d8735c 100644
> > --- a/drivers/input/keyboard/snvs_pwrkey.c
> > +++ b/drivers/input/keyboard/snvs_pwrkey.c
> > @@ -16,6 +16,7 @@
> > #include <linux/of_address.h>
> > #include <linux/platform_device.h>
> > #include <linux/pm_wakeirq.h>
> > +#include <linux/clk.h>
> > #include <linux/mfd/syscon.h>
> > #include <linux/regmap.h>
> >
> > @@ -38,6 +39,7 @@ struct pwrkey_drv_data {
> > int wakeup;
> > struct timer_list check_timer;
> > struct input_dev *input;
> > + struct clk *clk;
> > u8 minor_rev;
> > };
> >
> > @@ -72,6 +74,9 @@ static irqreturn_t imx_snvs_pwrkey_interrupt(int irq, void *dev_id)
> > struct input_dev *input = pdata->input;
> > u32 lp_status;
> >
> > + if (pdata->clk)
> > + clk_enable(pdata->clk);
> > +
> clk framework handles NULL pointers internally, the check is redundant.
>
> > pm_wakeup_event(input->dev.parent, 0);
> >
> > regmap_read(pdata->snvs, SNVS_LPSR_REG, &lp_status);
> > @@ -96,6 +101,9 @@ static irqreturn_t imx_snvs_pwrkey_interrupt(int irq, void *dev_id)
> > /* clear SPO status */
> > regmap_write(pdata->snvs, SNVS_LPSR_REG, SNVS_LPSR_SPO);
> >
> > + if (pdata->clk)
> > + clk_disable(pdata->clk);
> > +
> > return IRQ_HANDLED;
> > }
> >
> > @@ -140,6 +148,25 @@ static int imx_snvs_pwrkey_probe(struct platform_device *pdev)
> > if (pdata->irq < 0)
> > return -EINVAL;
> >
> > + pdata->clk = devm_clk_get(&pdev->dev, "snvs-pwrkey");
> > + if (IS_ERR(pdata->clk)) {
> > + pdata->clk = NULL;
> Using devm_clk_get_optional() would simplify error handling.
It sounds to me that this clock is not at all optional and the driver
currently "works" only by accident and therefore optional is not
suitable here.
>
> > + } else {
> > + error = clk_prepare_enable(pdata->clk);
So if you enable clock here and do not disable it, why do you need to
enable it again in interrupt?
Thanks.
--
Dmitry
On 2/10/2020 4:03 AM, Robin Gong wrote:
> On 2020/02/10 Dmitry Torokhov <[email protected]> wrote:
>> On Fri, Feb 07, 2020 at 08:10:22AM +0000, Horia Geanta wrote:
>>> On 1/30/2020 10:45 PM, Andr? Draszik wrote:
>>>> @@ -140,6 +148,25 @@ static int imx_snvs_pwrkey_probe(struct
>> platform_device *pdev)
>>>> if (pdata->irq < 0)
>>>> return -EINVAL;
>>>>
>>>> + pdata->clk = devm_clk_get(&pdev->dev, "snvs-pwrkey");
>>>> + if (IS_ERR(pdata->clk)) {
>>>> + pdata->clk = NULL;
>>> Using devm_clk_get_optional() would simplify error handling.
>>
>> It sounds to me that this clock is not at all optional and the driver currently
>> "works" only by accident and therefore optional is not suitable here.
> Yes, then we need to add all snvs clk in dts for on legacy i.MX chips in this patchset
> to avoid any potential function broken.
In that case the DT binding should be updated too,
to make the clock mandatory.
But before doing this all i.MX SoCs should be checked.
Thanks,
Horia
On Mon, Feb 10, 2020 at 06:33:30AM +0000, Horia Geanta wrote:
> On 2/10/2020 4:03 AM, Robin Gong wrote:
> > On 2020/02/10 Dmitry Torokhov <[email protected]> wrote:
> >> On Fri, Feb 07, 2020 at 08:10:22AM +0000, Horia Geanta wrote:
> >>> On 1/30/2020 10:45 PM, Andr? Draszik wrote:
> >>>> @@ -140,6 +148,25 @@ static int imx_snvs_pwrkey_probe(struct
> >> platform_device *pdev)
> >>>> if (pdata->irq < 0)
> >>>> return -EINVAL;
> >>>>
> >>>> + pdata->clk = devm_clk_get(&pdev->dev, "snvs-pwrkey");
> >>>> + if (IS_ERR(pdata->clk)) {
> >>>> + pdata->clk = NULL;
> >>> Using devm_clk_get_optional() would simplify error handling.
> >>
> >> It sounds to me that this clock is not at all optional and the driver currently
> >> "works" only by accident and therefore optional is not suitable here.
> > Yes, then we need to add all snvs clk in dts for on legacy i.MX chips in this patchset
> > to avoid any potential function broken.
How many are there? I am not too terribly opposed of having the driver
handle missing clk if there are very many legacy DTSes out there. But
then we need to handle it properly (i.e. current iteration does not
handle referral properly for example).
> In that case the DT binding should be updated too,
> to make the clock mandatory.
I think this should be done in either case, as as far I understand the
part can not function without the clock and it worked purely by chance
on some systems as something else was turning the clock on.
Thanks.
--
Dmitry
Hi,
On Tue, 2020-02-11 at 01:54 +0000, Robin Gong wrote:
> On 2020/02/11 Dmitry Torokhov <[email protected]> wrote:
> > On Mon, Feb 10, 2020 at 06:33:30AM +0000, Horia Geanta wrote:
> > > On 2/10/2020 4:03 AM, Robin Gong wrote:
> > > > On 2020/02/10 Dmitry Torokhov <[email protected]> wrote:
> > > > > On Fri, Feb 07, 2020 at 08:10:22AM +0000, Horia Geanta wrote:
> > > > > > On 1/30/2020 10:45 PM, André Draszik wrote:
> > > > > > > @@ -140,6 +148,25 @@ static int imx_snvs_pwrkey_probe(struct
> > > > > platform_device *pdev)
> > > > > > > if (pdata->irq < 0)
> > > > > > > return -EINVAL;
> > > > > > >
> > > > > > > + pdata->clk = devm_clk_get(&pdev->dev, "snvs-pwrkey");
> > > > > > > + if (IS_ERR(pdata->clk)) {
> > > > > > > + pdata->clk = NULL;
> > > > > > Using devm_clk_get_optional() would simplify error handling.
> > > > >
> > > > > It sounds to me that this clock is not at all optional and the
> > > > > driver currently "works" only by accident and therefore optional is not
> > suitable here.
> > > > Yes, then we need to add all snvs clk in dts for on legacy i.MX
> > > > chips in this patchset to avoid any potential function broken.
> >
> > How many are there? I am not too terribly opposed of having the driver handle
> > missing clk if there are very many legacy DTSes out there. But then we need to
> > handle it properly (i.e. current iteration does not handle referral properly for
> > example).
> There are four dtsi which have clock support in snvs-rtc including i.mx7s/i.mx8mq/8mm/8mn. So for this patch set,
> it's better update
> i.mx8mX dtsi except i.mx7s.
> > > In that case the DT binding should be updated too, to make the clock
> > > mandatory.
> >
> > I think this should be done in either case, as as far I understand the part can not
> > function without the clock and it worked purely by chance on some systems as
> > something else was turning the clock on.
> Yes, for all chips snvs clk management added, snvs clock also has to been add
> in snvs_pwrkey dts, but for others legacy chips like i.mx6X which have no snvs
> clk management, snvs clock is always on, so no need such clk in snvs_pwrkey
> dts either, optional is better.
> > Thanks.
It seems to me though that the clock should really be moved into the (parent) SNVS
node itself, rather than duplicating
the clock in the the power key node and in
the RTC node. Is that possible? (I don't know)...
To summarise, I'll post an updated patchset within the next couple days so to:
* keep the clock optional (for i.MX6 platforms)
* convert to devm_clk_get_optional()
* only enable the clock in interrupt handler and imx_imx_snvs_check_for_events()
but not during driver loading
* update all four DTSIs: imx8mm.dtsi imx8mn.dtsi imx8mq.dtsi imx7s.dtsi
Note that I'll only be able to test onn the i.MX7
Cheers,
Andre'