On 23-07-04, Andrej Picej wrote:
> If regulator is not marked as always-on the regulator gets disabled on
> reboot breaking the next boot.
>
> Signed-off-by: Andrej Picej <[email protected]>
> ---
> arch/arm/boot/dts/imx6qdl-phytec-pfla02.dtsi | 1 +
> 1 file changed, 1 insertion(+)
>
> diff --git a/arch/arm/boot/dts/imx6qdl-phytec-pfla02.dtsi b/arch/arm/boot/dts/imx6qdl-phytec-pfla02.dtsi
> index 80adb2a02cc9..25d6a036d5b8 100644
> --- a/arch/arm/boot/dts/imx6qdl-phytec-pfla02.dtsi
> +++ b/arch/arm/boot/dts/imx6qdl-phytec-pfla02.dtsi
> @@ -192,6 +192,7 @@ vdd_3v3_pmic_io_reg: ldo6 {
> vdd_sd0_reg: ldo9 {
> regulator-min-microvolt = <3300000>;
> regulator-max-microvolt = <3300000>;
> + regulator-always-on;
I think this is the supply for the sd-card, so you can make use of
'vmmc-supply'.
Regards,
Marco
> };
>
> vdd_sd1_reg: ldo10 {
> --
> 2.25.1
>
>
>
Hi Marco,
On 4. 07. 23 10:17, Marco Felsch wrote:
> On 23-07-04, Andrej Picej wrote:
>> If regulator is not marked as always-on the regulator gets disabled on
>> reboot breaking the next boot.
>>
>> Signed-off-by: Andrej Picej <[email protected]>
>> ---
>> arch/arm/boot/dts/imx6qdl-phytec-pfla02.dtsi | 1 +
>> 1 file changed, 1 insertion(+)
>>
>> diff --git a/arch/arm/boot/dts/imx6qdl-phytec-pfla02.dtsi b/arch/arm/boot/dts/imx6qdl-phytec-pfla02.dtsi
>> index 80adb2a02cc9..25d6a036d5b8 100644
>> --- a/arch/arm/boot/dts/imx6qdl-phytec-pfla02.dtsi
>> +++ b/arch/arm/boot/dts/imx6qdl-phytec-pfla02.dtsi
>> @@ -192,6 +192,7 @@ vdd_3v3_pmic_io_reg: ldo6 {
>> vdd_sd0_reg: ldo9 {
>> regulator-min-microvolt = <3300000>;
>> regulator-max-microvolt = <3300000>;
>> + regulator-always-on;
>
> I think this is the supply for the sd-card, so you can make use of
> 'vmmc-supply'.
This is already the case:
> &usdhc3 {
> pinctrl-names = "default";
> pinctrl-0 = <&pinctrl_usdhc3
> &pinctrl_usdhc3_cdwp>;
> cd-gpios = <&gpio1 27 GPIO_ACTIVE_LOW>;
> wp-gpios = <&gpio1 29 GPIO_ACTIVE_HIGH>;
> vmmc-supply = <&vdd_sd0_reg>;
> status = "disabled";
> };
I think the main reason for a failed boot is that the PMIC doesn't get
reset and that the bootloader doesn't specifically enable the SD card
regulator.
Could this patch still be applied or should we put the fix in reset
routine/bootloader?
Best regards,
Andrej
>
> Regards,
> Marco
>
>> };
>>
>> vdd_sd1_reg: ldo10 {
>> --
>> 2.25.1
>>
>>
>>
On 05.07.23 10:28, Andrej Picej wrote:
> Hi Marco,
>
> On 4. 07. 23 10:17, Marco Felsch wrote:
>> On 23-07-04, Andrej Picej wrote:
>>> If regulator is not marked as always-on the regulator gets disabled on
>>> reboot breaking the next boot.
>>>
>>> Signed-off-by: Andrej Picej <[email protected]>
>>> ---
>>> arch/arm/boot/dts/imx6qdl-phytec-pfla02.dtsi | 1 +
>>> 1 file changed, 1 insertion(+)
>>>
>>> diff --git a/arch/arm/boot/dts/imx6qdl-phytec-pfla02.dtsi b/arch/arm/boot/dts/imx6qdl-phytec-pfla02.dtsi
>>> index 80adb2a02cc9..25d6a036d5b8 100644
>>> --- a/arch/arm/boot/dts/imx6qdl-phytec-pfla02.dtsi
>>> +++ b/arch/arm/boot/dts/imx6qdl-phytec-pfla02.dtsi
>>> @@ -192,6 +192,7 @@ vdd_3v3_pmic_io_reg: ldo6 {
>>> vdd_sd0_reg: ldo9 {
>>> regulator-min-microvolt = <3300000>;
>>> regulator-max-microvolt = <3300000>;
>>> + regulator-always-on;
>>
>> I think this is the supply for the sd-card, so you can make use of
>> 'vmmc-supply'.
>
> This is already the case:
>
>> &usdhc3 {
>> pinctrl-names = "default";
>> pinctrl-0 = <&pinctrl_usdhc3
>> &pinctrl_usdhc3_cdwp>;
>> cd-gpios = <&gpio1 27 GPIO_ACTIVE_LOW>;
>> wp-gpios = <&gpio1 29 GPIO_ACTIVE_HIGH>;
>> vmmc-supply = <&vdd_sd0_reg>;
>> status = "disabled";
>> };
>
> I think the main reason for a failed boot is that the PMIC doesn't get reset and that the bootloader doesn't specifically enable the SD card regulator.
>
> Could this patch still be applied or should we put the fix in reset routine/bootloader?
Is SD-Card not main boot medium? From your description, I thought BootROM
will fail to boot before bootloader has a chance to do anything about it.
>
> Best regards,
> Andrej
>
>>
>> Regards,
>> Marco
>>
>>> };
>>> vdd_sd1_reg: ldo10 {
>>> --
>>> 2.25.1
>>>
>>>
>>>
>
>
--
Pengutronix e.K. | |
Steuerwalder Str. 21 | http://www.pengutronix.de/ |
31137 Hildesheim, Germany | Phone: +49-5121-206917-0 |
Amtsgericht Hildesheim, HRA 2686 | Fax: +49-5121-206917-5555 |
On 5. 07. 23 10:30, Ahmad Fatoum wrote:
> On 05.07.23 10:28, Andrej Picej wrote:
>> Hi Marco,
>>
>> On 4. 07. 23 10:17, Marco Felsch wrote:
>>> On 23-07-04, Andrej Picej wrote:
>>>> If regulator is not marked as always-on the regulator gets disabled on
>>>> reboot breaking the next boot.
>>>>
>>>> Signed-off-by: Andrej Picej <[email protected]>
>>>> ---
>>>> arch/arm/boot/dts/imx6qdl-phytec-pfla02.dtsi | 1 +
>>>> 1 file changed, 1 insertion(+)
>>>>
>>>> diff --git a/arch/arm/boot/dts/imx6qdl-phytec-pfla02.dtsi b/arch/arm/boot/dts/imx6qdl-phytec-pfla02.dtsi
>>>> index 80adb2a02cc9..25d6a036d5b8 100644
>>>> --- a/arch/arm/boot/dts/imx6qdl-phytec-pfla02.dtsi
>>>> +++ b/arch/arm/boot/dts/imx6qdl-phytec-pfla02.dtsi
>>>> @@ -192,6 +192,7 @@ vdd_3v3_pmic_io_reg: ldo6 {
>>>> vdd_sd0_reg: ldo9 {
>>>> regulator-min-microvolt = <3300000>;
>>>> regulator-max-microvolt = <3300000>;
>>>> + regulator-always-on;
>>>
>>> I think this is the supply for the sd-card, so you can make use of
>>> 'vmmc-supply'.
>>
>> This is already the case:
>>
>>> &usdhc3 {
>>> pinctrl-names = "default";
>>> pinctrl-0 = <&pinctrl_usdhc3
>>> &pinctrl_usdhc3_cdwp>;
>>> cd-gpios = <&gpio1 27 GPIO_ACTIVE_LOW>;
>>> wp-gpios = <&gpio1 29 GPIO_ACTIVE_HIGH>;
>>> vmmc-supply = <&vdd_sd0_reg>;
>>> status = "disabled";
>>> };
>>
>> I think the main reason for a failed boot is that the PMIC doesn't get reset and that the bootloader doesn't specifically enable the SD card regulator.
>>
>> Could this patch still be applied or should we put the fix in reset routine/bootloader?
>
> Is SD-Card not main boot medium? From your description, I thought BootROM
> will fail to boot before bootloader has a chance to do anything about it.
>
Yes sorry, you are absolutly right, the BootROM fails. It confused me
because I could see the booloader booting, but it was from one of the
fallback mediums. So I guess fixing the bootloader is not really an option.
Sorry for the confusion.
>>
>> Best regards,
>> Andrej
>>
>>>
>>> Regards,
>>> Marco
>>>
>>>> };
>>>> vdd_sd1_reg: ldo10 {
>>>> --
>>>> 2.25.1
>>>>
>>>>
>>>>
>>
>>
>
On 5. 07. 23 10:40, Andrej Picej wrote:
>
>
> On 5. 07. 23 10:30, Ahmad Fatoum wrote:
>> On 05.07.23 10:28, Andrej Picej wrote:
>>> Hi Marco,
>>>
>>> On 4. 07. 23 10:17, Marco Felsch wrote:
>>>> On 23-07-04, Andrej Picej wrote:
>>>>> If regulator is not marked as always-on the regulator gets disabled on
>>>>> reboot breaking the next boot.
>>>>>
>>>>> Signed-off-by: Andrej Picej <[email protected]>
>>>>> ---
>>>>> arch/arm/boot/dts/imx6qdl-phytec-pfla02.dtsi | 1 +
>>>>> 1 file changed, 1 insertion(+)
>>>>>
>>>>> diff --git a/arch/arm/boot/dts/imx6qdl-phytec-pfla02.dtsi
>>>>> b/arch/arm/boot/dts/imx6qdl-phytec-pfla02.dtsi
>>>>> index 80adb2a02cc9..25d6a036d5b8 100644
>>>>> --- a/arch/arm/boot/dts/imx6qdl-phytec-pfla02.dtsi
>>>>> +++ b/arch/arm/boot/dts/imx6qdl-phytec-pfla02.dtsi
>>>>> @@ -192,6 +192,7 @@ vdd_3v3_pmic_io_reg: ldo6 {
>>>>> vdd_sd0_reg: ldo9 {
>>>>> regulator-min-microvolt = <3300000>;
>>>>> regulator-max-microvolt = <3300000>;
>>>>> + regulator-always-on;
>>>>
>>>> I think this is the supply for the sd-card, so you can make use of
>>>> 'vmmc-supply'.
>>>
>>> This is already the case:
>>>
>>>> &usdhc3 {
>>>> pinctrl-names = "default";
>>>> pinctrl-0 = <&pinctrl_usdhc3
>>>> &pinctrl_usdhc3_cdwp>;
>>>> cd-gpios = <&gpio1 27 GPIO_ACTIVE_LOW>;
>>>> wp-gpios = <&gpio1 29 GPIO_ACTIVE_HIGH>;
>>>> vmmc-supply = <&vdd_sd0_reg>;
>>>> status = "disabled";
>>>> };
>>>
>>> I think the main reason for a failed boot is that the PMIC doesn't
>>> get reset and that the bootloader doesn't specifically enable the SD
>>> card regulator.
>>>
>>> Could this patch still be applied or should we put the fix in reset
>>> routine/bootloader?
>>
>> Is SD-Card not main boot medium? From your description, I thought BootROM
>> will fail to boot before bootloader has a chance to do anything about it.
>>
>
> Yes sorry, you are absolutly right, the BootROM fails. It confused me
> because I could see the booloader booting, but it was from one of the
> fallback mediums. So I guess fixing the bootloader is not really an option.
> Sorry for the confusion.
>
Ok, the main problem is well known, that's why PHYTEC disables the imx
watchdog and uses a PMIC one for the reboot handler. This one resets the
board completely. The SD card regulator problem is really just the
manifestation of that bug. Unfortunately I didn't noticed that. :(
I will create a v2 with a proper fix, where imx watchdog gets disabled.
Thanks for your help,
Andrej
>
>>>
>>> Best regards,
>>> Andrej
>>>
>>>>
>>>> Regards,
>>>> Marco
>>>>
>>>>> };
>>>>> vdd_sd1_reg: ldo10 {
>>>>> --
>>>>> 2.25.1
>>>>>
>>>>>
>>>>>
>>>
>>>
>>
On 05.07.23 12:39, Andrej Picej wrote:> On 5. 07. 23 10:40, Andrej Picej wrote:
>> On 5. 07. 23 10:30, Ahmad Fatoum wrote:
>>> On 05.07.23 10:28, Andrej Picej wrote:
>>>> I think the main reason for a failed boot is that the PMIC doesn't get reset and that the bootloader doesn't specifically enable the SD card regulator.
>>>>
>>>> Could this patch still be applied or should we put the fix in reset routine/bootloader?
>>>
>>> Is SD-Card not main boot medium? From your description, I thought BootROM
>>> will fail to boot before bootloader has a chance to do anything about it.
>>>
>>
>> Yes sorry, you are absolutly right, the BootROM fails. It confused me because I could see the booloader booting, but it was from one of the fallback mediums. So I guess fixing the bootloader is not really an option.
>> Sorry for the confusion.
>>
>
> Ok, the main problem is well known, that's why PHYTEC disables the imx watchdog and uses a PMIC one for the reboot handler. This one resets the board completely. The SD card regulator problem is really just the manifestation of that bug. Unfortunately I didn't noticed that. :(
>
> I will create a v2 with a proper fix, where imx watchdog gets disabled.
I'd be wary about solving it this way at the DTSI level, because it can
break existing users:
- Boot flow depends on reading boot reason, but with PMIC reset, everything
is power-on reset
- Bootloader starts i.MX watchdog, but new kernel will service only
PMIC watchdog leading to system reset
- Even if updating bootloader and kernel together, fallback of kernel
may end up that bootloader uses PMIC watchdog, but kernel uses i.MX
watchdog
- There can be valid reasons to use both watchdogs and disabling
one at the SoM level breaks that
I had a similar issue once (Board controller reset to be used instead of SoC
reset) and settled on using the barebox watchdog-priority/restart-priority[1]
binding to select the "better" watchdog and then fixed up this choice into
the kernel command line (barebox CONFIG_SYSTEMD_OF_WATCHDOG).
If you decide to fix it for the evaluation kits, please add some text
into the commit message that this fix should not be backported to older kernels.
While it's ultimately the correct thing to do, changing this is IMO not stable
backport material.
[1]: FWIW, there was past discussion about adding restart priorities to Linux, e.g.
https://lore.kernel.org/all/[email protected]/
Cheers,
Ahmad
>
> Thanks for your help,
> Andrej
>
>
>>
>>>>
>>>> Best regards,
>>>> Andrej
>>>>
>>>>>
>>>>> Regards,
>>>>> Marco
>>>>>
>>>>>> };
>>>>>> vdd_sd1_reg: ldo10 {
>>>>>> --
>>>>>> 2.25.1
>>>>>>
>>>>>>
>>>>>>
>>>>
>>>>
>>>
>
--
Pengutronix e.K. | |
Steuerwalder Str. 21 | http://www.pengutronix.de/ |
31137 Hildesheim, Germany | Phone: +49-5121-206917-0 |
Amtsgericht Hildesheim, HRA 2686 | Fax: +49-5121-206917-5555 |
On 5. 07. 23 14:06, Ahmad Fatoum wrote:
> On 05.07.23 12:39, Andrej Picej wrote:> On 5. 07. 23 10:40, Andrej Picej wrote:
>>> On 5. 07. 23 10:30, Ahmad Fatoum wrote:
>>>> On 05.07.23 10:28, Andrej Picej wrote:
>>>>> I think the main reason for a failed boot is that the PMIC doesn't get reset and that the bootloader doesn't specifically enable the SD card regulator.
>>>>>
>>>>> Could this patch still be applied or should we put the fix in reset routine/bootloader?
>>>>
>>>> Is SD-Card not main boot medium? From your description, I thought BootROM
>>>> will fail to boot before bootloader has a chance to do anything about it.
>>>>
>>>
>>> Yes sorry, you are absolutly right, the BootROM fails. It confused me because I could see the booloader booting, but it was from one of the fallback mediums. So I guess fixing the bootloader is not really an option.
>>> Sorry for the confusion.
>>>
>>
>> Ok, the main problem is well known, that's why PHYTEC disables the imx watchdog and uses a PMIC one for the reboot handler. This one resets the board completely. The SD card regulator problem is really just the manifestation of that bug. Unfortunately I didn't noticed that. :(
>>
>> I will create a v2 with a proper fix, where imx watchdog gets disabled.
>
> I'd be wary about solving it this way at the DTSI level, because it can
> break existing users:
>
> - Boot flow depends on reading boot reason, but with PMIC reset, everything
> is power-on reset
>
> - Bootloader starts i.MX watchdog, but new kernel will service only
> PMIC watchdog leading to system reset
>
> - Even if updating bootloader and kernel together, fallback of kernel
> may end up that bootloader uses PMIC watchdog, but kernel uses i.MX
> watchdog
>
> - There can be valid reasons to use both watchdogs and disabling
> one at the SoM level breaks that
> > I had a similar issue once (Board controller reset to be used instead
of SoC
> reset) and settled on using the barebox watchdog-priority/restart-priority[1]
> binding to select the "better" watchdog and then fixed up this choice into
> the kernel command line (barebox CONFIG_SYSTEMD_OF_WATCHDOG).
>
> If you decide to fix it for the evaluation kits, please add some text
> into the commit message that this fix should not be backported to older kernels.
> While it's ultimately the correct thing to do, changing this is IMO not stable
> backport material.
>
> [1]: FWIW, there was past discussion about adding restart priorities to Linux, e.g.
> https://lore.kernel.org/all/[email protected]/
Ok I see the problems this might raise. Nevertheless I think it would be
good to sync upstream and PHYTEC downstream dts with this fix. Since by
default imx watchdog reset is really a no-go with phyFLEX.
We could add the aliases for the watchdogs, marking the PMIC watchdog as
watchdog0, and imx watchdog as watchdog1 which I think should tell the
kernel to use the PMIC reboot handler, but not really sure how is that
any better then disabling the imx watchdog.
Why do you think this shouldn't be backported to older kernels? Because
someone might use this with "not compatible" bootloader?
Thanks for your help,
Andrej
>
> Cheers,
> Ahmad
>
>>
>> Thanks for your help,
>> Andrej
>>
>>
>>>
>>>>>
>>>>> Best regards,
>>>>> Andrej
>>>>>
>>>>>>
>>>>>> Regards,
>>>>>> Marco
>>>>>>
>>>>>>> };
>>>>>>> vdd_sd1_reg: ldo10 {
>>>>>>> --
>>>>>>> 2.25.1
>>>>>>>
>>>>>>>
>>>>>>>
>>>>>
>>>>>
>>>>
>>
>
Hello Andrej,
On 07.07.23 11:28, Andrej Picej wrote:
>
>
> On 5. 07. 23 14:06, Ahmad Fatoum wrote:
>> On 05.07.23 12:39, Andrej Picej wrote:> On 5. 07. 23 10:40, Andrej Picej wrote:
>>>> On 5. 07. 23 10:30, Ahmad Fatoum wrote:
>>>>> On 05.07.23 10:28, Andrej Picej wrote:
>>>>>> I think the main reason for a failed boot is that the PMIC doesn't get reset and that the bootloader doesn't specifically enable the SD card regulator.
>>>>>>
>>>>>> Could this patch still be applied or should we put the fix in reset routine/bootloader?
>>>>>
>>>>> Is SD-Card not main boot medium? From your description, I thought BootROM
>>>>> will fail to boot before bootloader has a chance to do anything about it.
>>>>>
>>>>
>>>> Yes sorry, you are absolutly right, the BootROM fails. It confused me because I could see the booloader booting, but it was from one of the fallback mediums. So I guess fixing the bootloader is not really an option.
>>>> Sorry for the confusion.
>>>>
>>>
>>> Ok, the main problem is well known, that's why PHYTEC disables the imx watchdog and uses a PMIC one for the reboot handler. This one resets the board completely. The SD card regulator problem is really just the manifestation of that bug. Unfortunately I didn't noticed that. :(
>>>
>>> I will create a v2 with a proper fix, where imx watchdog gets disabled.
>>
>> I'd be wary about solving it this way at the DTSI level, because it can
>> break existing users:
>>
>> - Boot flow depends on reading boot reason, but with PMIC reset, everything
>> is power-on reset
>>
>> - Bootloader starts i.MX watchdog, but new kernel will service only
>> PMIC watchdog leading to system reset
>>
>> - Even if updating bootloader and kernel together, fallback of kernel
>> may end up that bootloader uses PMIC watchdog, but kernel uses i.MX
>> watchdog
>>
>> - There can be valid reasons to use both watchdogs and disabling
>> one at the SoM level breaks that
>> > I had a similar issue once (Board controller reset to be used instead
> of SoC
>> reset) and settled on using the barebox watchdog-priority/restart-priority[1]
>> binding to select the "better" watchdog and then fixed up this choice into
>> the kernel command line (barebox CONFIG_SYSTEMD_OF_WATCHDOG).
>>
>> If you decide to fix it for the evaluation kits, please add some text
>> into the commit message that this fix should not be backported to older kernels.
>> While it's ultimately the correct thing to do, changing this is IMO not stable
>> backport material.
>>
>> [1]: FWIW, there was past discussion about adding restart priorities to Linux, e.g.
>> https://lore.kernel.org/all/[email protected]/
>
> Ok I see the problems this might raise. Nevertheless I think it would be good to sync upstream and PHYTEC downstream dts with this fix. Since by default imx watchdog reset is really a no-go with phyFLEX.
What does downstream do? Might've been useful to note in the commit message,
especially if there's an erratum.
> We could add the aliases for the watchdogs, marking the PMIC watchdog as watchdog0, and imx watchdog as watchdog1 which I think should tell the kernel to use the PMIC reboot handler, but not really sure how is that any better then disabling the imx watchdog.
Restart priority is set by watchdog_set_restart_priority and both
da906[23] and i.MX watchdog set the same value of 128 here. Might
be worthwhile to increase PMIC over i.MX watchdog priority to ensure
it's taken instead.
> Why do you think this shouldn't be backported to older kernels? Because
> someone might use this with "not compatible" bootloader?
Yes. Selecting a different watchdog is IMO not something that should
come in via a stable update.
Cheers,
Ahmad
>
> Thanks for your help,
> Andrej
>
>>
>> Cheers,
>> Ahmad
>>
>>>
>>> Thanks for your help,
>>> Andrej
>>>
>>>
>>>>
>>>>>>
>>>>>> Best regards,
>>>>>> Andrej
>>>>>>
>>>>>>>
>>>>>>> Regards,
>>>>>>> Marco
>>>>>>>
>>>>>>>> };
>>>>>>>> vdd_sd1_reg: ldo10 {
>>>>>>>> --
>>>>>>>> 2.25.1
>>>>>>>>
>>>>>>>>
>>>>>>>>
>>>>>>
>>>>>>
>>>>>
>>>
>>
>
--
Pengutronix e.K. | |
Steuerwalder Str. 21 | http://www.pengutronix.de/ |
31137 Hildesheim, Germany | Phone: +49-5121-206917-0 |
Amtsgericht Hildesheim, HRA 2686 | Fax: +49-5121-206917-5555 |