[U-Boot] iMX8QXP MEK: no ethernet

[U-Boot] iMX8QXP MEK: no ethernet

[Andrejs Cainikovs]

Hi,

I cant get ethernet to work on iMX8QXP MEK board with latest sources 
(tried v2019.01 as well). Is this a bug or work in progress?

---

U-Boot SPL 2019.04-rc2-00134-g783e66816d (Feb 28 2019 - 13:15:40 +0100)
Normal Boot
Trying to boot from MMC2_2


U-Boot 2019.04-rc2-00134-g783e66816d (Feb 28 2019 - 13:15:40 +0100)

CPU:   NXP i.MX8QXP RevB A35 at 147228 MHz

Model: Freescale i.MX8QXP MEK
Board: iMX8QXP MEK
Build: SCFW 3301c1a9
Boot:  SD1
DRAM:  3 GiB
MMC:   FSL_SDHC: 0, FSL_SDHC: 1
Loading Environment from MMC... OK
In:    serial at 5a060000
Out:   serial at 5a060000
Err:   serial at 5a060000
Net:   eth-1: ethernet at 5b040000
Hit any key to stop autoboot:  0
=> dhcp
No ethernet found.

---

Best regards,
Andrejs Cainikovs.

[U-Boot] iMX8QXP MEK: no ethernet

[Anatolij Gustschin]

Hi,

On Thu, 28 Feb 2019 12:27:45 +0000
Andrejs Cainikovs Andrejs.Cainikovs at netmodule.com wrote:

> Hi,
> 
> I cant get ethernet to work on iMX8QXP MEK board with latest sources 
> (tried v2019.01 as well). Is this a bug or work in progress?

This is probably due to the missing delay after PHY reset in fec_gpio_reset().

Please try with attached patch, it worked for me on MEK board with RevA CPU.

--
Anatolij
-------------- next part --------------
A non-text attachment was scrubbed...
Name: 0001-net-fec-delay-after-PHY-reset-when-on-i.MX8QXP-MEK-b.patch
Type: text/x-patch
Size: 952 bytes
Desc: not available
URL: <http://lists.denx.de/pipermail/u-boot/attachments/20190228/68f8097e/attachment.bin>

[U-Boot] iMX8QXP MEK: no ethernet

[Andrejs Cainikovs]

Hi,

On 28/02/2019 23:45, Anatolij Gustschin wrote:
> 
>> I cant get ethernet to work on iMX8QXP MEK board with latest sources
>> (tried v2019.01 as well). Is this a bug or work in progress?
> 
> This is probably due to the missing delay after PHY reset in fec_gpio_reset().
> 
> Please try with attached patch, it worked for me on MEK board with RevA CPU.

Thanks a lot! It works now.

Best regards,
Andrejs Cainikovs.

[U-Boot] iMX8QXP MEK: no ethernet

[Anatolij Gustschin]

Hi,

On Fri, 1 Mar 2019 09:50:06 +0000
Andrejs Cainikovs Andrejs.Cainikovs at netmodule.com wrote:
...
> Thanks a lot! It works now.

Great, thanks for testing! I don't like the patch however and I'm still
not sure how we should fix it in a proper way for upcoming release.
Unfortunately, currently I don't have time to work on a suitable fix.

--
Anatolij

[U-Boot] iMX8QXP MEK: no ethernet

[Stefano Babic]

On 01/03/19 11:05, Anatolij Gustschin wrote:
> Hi,
> 
> On Fri, 1 Mar 2019 09:50:06 +0000
> Andrejs Cainikovs Andrejs.Cainikovs at netmodule.com wrote:
> ...
>> Thanks a lot! It works now.
> 
> Great, thanks for testing! I don't like the patch however and I'm still
> not sure how we should fix it in a proper way for upcoming release.
> Unfortunately, currently I don't have time to work on a suitable fix.
> 

What about to change this in DT ?

diff --git a/arch/arm/dts/fsl-imx8qxp-mek.dts
b/arch/arm/dts/fsl-imx8qxp-mek.dts
index 41f7ec1763..b6c4593334 100644
--- a/arch/arm/dts/fsl-imx8qxp-mek.dts
+++ b/arch/arm/dts/fsl-imx8qxp-mek.dts
@@ -223,7 +223,7 @@
        fsl,magic-packet;
        status = "okay";
        phy-reset-gpios = <&pca9557_a 4 GPIO_ACTIVE_LOW>;
-       phy-reset-duration = <10>;
+       phy-reset-duration = <150>;

        mdio {
                #address-cells = <1>;

Best regards,
Stefano


-- 
=====================================================================
DENX Software Engineering GmbH,      Managing Director: Wolfgang Denk
HRB 165235 Munich, Office: Kirchenstr.5, D-82194 Groebenzell, Germany
Phone: +49-8142-66989-53 Fax: +49-8142-66989-80 Email: sbabic at denx.de
=====================================================================

[U-Boot] iMX8QXP MEK: no ethernet

[Anatolij Gustschin]

Hi Stefano,

On Fri, 1 Mar 2019 12:19:41 +0100
Stefano Babic sbabic at denx.de wrote:
...
> What about to change this in DT ?
> 
> diff --git a/arch/arm/dts/fsl-imx8qxp-mek.dts
> b/arch/arm/dts/fsl-imx8qxp-mek.dts
> index 41f7ec1763..b6c4593334 100644
> --- a/arch/arm/dts/fsl-imx8qxp-mek.dts
> +++ b/arch/arm/dts/fsl-imx8qxp-mek.dts
> @@ -223,7 +223,7 @@
>         fsl,magic-packet;
>         status = "okay";
>         phy-reset-gpios = <&pca9557_a 4 GPIO_ACTIVE_LOW>;
> -       phy-reset-duration = <10>;
> +       phy-reset-duration = <150>;

This will extend the time for holding the PHY in reset state, but 10ms was already enough
to reset the PHY. The problem is that after reset state is deactivated, the PHY cannot be
detected unless we wait about 150 ms in fec_gpio_reset() after removing the reset level.
I don't now the reason why this happens.

--
Anatolij

[U-Boot] iMX8QXP MEK: no ethernet

[Andrejs Cainikovs]

Hi Stefano,

On 01/03/2019 12:19, Stefano Babic wrote:
> 
> What about to change this in DT ?
> 
> diff --git a/arch/arm/dts/fsl-imx8qxp-mek.dts
> b/arch/arm/dts/fsl-imx8qxp-mek.dts
> index 41f7ec1763..b6c4593334 100644
> --- a/arch/arm/dts/fsl-imx8qxp-mek.dts
> +++ b/arch/arm/dts/fsl-imx8qxp-mek.dts
> @@ -223,7 +223,7 @@
>          fsl,magic-packet;
>          status = "okay";
>          phy-reset-gpios = <&pca9557_a 4 GPIO_ACTIVE_LOW>;
> -       phy-reset-duration = <10>;
> +       phy-reset-duration = <150>;
> 
>          mdio {
>                  #address-cells = <1>;
> 

Nope, didn't worked. This parameter is a pulse width of a reset, but the 
issue is about having a delay *after* reset. However, introducing a new 
`phy-reset-wait` is a reasonable option. Thanks for idea, I will make a 
patch, and then it's up to you and Anatolij to decide whether it is a ok 
or not.

Best regards,
Andrejs Cainikovs.

[U-Boot] iMX8QXP MEK: no ethernet

[Stefano Babic]

On 01/03/19 13:30, Andrejs Cainikovs wrote:
> Hi Stefano,
> 
> On 01/03/2019 12:19, Stefano Babic wrote:
>>
>> What about to change this in DT ?
>>
>> diff --git a/arch/arm/dts/fsl-imx8qxp-mek.dts
>> b/arch/arm/dts/fsl-imx8qxp-mek.dts
>> index 41f7ec1763..b6c4593334 100644
>> --- a/arch/arm/dts/fsl-imx8qxp-mek.dts
>> +++ b/arch/arm/dts/fsl-imx8qxp-mek.dts
>> @@ -223,7 +223,7 @@
>>          fsl,magic-packet;
>>          status = "okay";
>>          phy-reset-gpios = <&pca9557_a 4 GPIO_ACTIVE_LOW>;
>> -       phy-reset-duration = <10>;
>> +       phy-reset-duration = <150>;
>>
>>          mdio {
>>                  #address-cells = <1>;
>>
> 
> Nope, didn't worked. This parameter is a pulse width of a reset, but the 
> issue is about having a delay *after* reset. However, introducing a new 
> `phy-reset-wait` is a reasonable option. 

Right, I think this is a way to do.

> Thanks for idea, I will make a 
> patch, and then it's up to you and Anatolij to decide whether it is a ok 
> or not.

Just post the patch to Joe Hershberger, too, because he's the network
maintainer.

Regards,
Stefano


-- 
=====================================================================
DENX Software Engineering GmbH,      Managing Director: Wolfgang Denk
HRB 165235 Munich, Office: Kirchenstr.5, D-82194 Groebenzell, Germany
Phone: +49-8142-66989-53 Fax: +49-8142-66989-80 Email: sbabic at denx.de
=====================================================================

[U-Boot] iMX8QXP MEK: no ethernet

[Anatolij Gustschin]

Hi all,

On Fri, 1 Mar 2019 13:36:17 +0100
Stefano Babic sbabic at denx.de wrote:
...
> > Nope, didn't worked. This parameter is a pulse width of a reset, but the 
> > issue is about having a delay *after* reset. However, introducing a new 
> > `phy-reset-wait` is a reasonable option.   
> 
> Right, I think this is a way to do.

yes, this is much better than ifdef. I'd suggest using 'phy-reset-post-delay'
property instead of 'phy-reset-wait'. Linux has already such binding, i.e.
see Documentation/devicetree/bindings/net/fsl-fec.txt and the driver
drivers/net/ethernet/freescale/fec_main.c.

--
Anatolij

[U-Boot] iMX8QXP MEK: no ethernet

[Andrejs Cainikovs]

On 01/03/2019 14:00, Anatolij Gustschin wrote:
> Hi all,
> 
> On Fri, 1 Mar 2019 13:36:17 +0100
> Stefano Babic sbabic at denx.de wrote:
>>
>> Right, I think this is a way to do.
> 
> yes, this is much better than ifdef. I'd suggest using 'phy-reset-post-delay'
> property instead of 'phy-reset-wait'. Linux has already such binding, i.e.
> see Documentation/devicetree/bindings/net/fsl-fec.txt and the driver
> drivers/net/ethernet/freescale/fec_main.c.
> 

Done. Please review.