Hi Stefan,
On Tue, Dec 4, 2018 at 10:51 PM Stefan Roese <s...@denx.de> wrote:
>
> On 04.12.18 16:37, Aditya Prayoga wrote:
>
> <snip>
>
> >>>> Yes. As mentioned above, please look at the mvpp2 driver. And e.g.
> >>>> at the armada-8040-clearfog-gt-8k.dts DT file:
> >>>>
> >>>> /* 1G SGMII */
> >>>> &cps_eth1 {
> >>>> status = "okay";
> >>>> phy-mode = "sgmii";
> >>>> phy = <&phy0>;
> >>>> phy-reset-gpios = <&cpm_gpio1 11 GPIO_ACTIVE_LOW>;
> >>>> };
> >>>>
> >>>> Here you see, how the GPIO is defined in the DT. It should not
> >>>> be too hard to get this implemented in the mvneta driver as well.
> >>>>
> >>> Yes, what I meant is to put "phy-reset-gpios" in the device tree but it
> >>> would still part of the system reset so the reset would be asserted in
> >>> board_init.
> >>>
> >>> If the reset implemented in driver (mvneta), the reset would only
> >>> asserted when u-boot going to use the network interface. This is not
> >>> desirable.
> >>
> >> Network device probe routine is always called even when not used. See this
>
> This is not correct. The probe function is *not* always called - at least
> not in newer DM drivers.
>
> >> comment in eth_initialize():
> >>
> >> /*
> >> * Devices need to write the hwaddr even if not started so that
> >> Linux
> >> * will have access to the hwaddr that u-boot stored for the
> >> device.
> >> * This is accomplished by attempting to probe each device and
> >> calling
> >> * their write_hwaddr() operation.
> >> */
> >>
>
> These are special hooks that are always called to write the MAC address,
> even when the device is not used in U-Boot. IIRC, the bind() function
> is also always called.
>
> > That is what I thought but that is not what I observed.
> > I tried to port that mvpp2 changes to mvneta
> > ---
> > diff --git a/drivers/net/mvneta.c b/drivers/net/mvneta.c
> > index 8cb04b5..34f191d 100644
> > --- a/drivers/net/mvneta.c
> > +++ b/drivers/net/mvneta.c
> > @@ -27,6 +27,7 @@
> > #include <asm/arch/soc.h>
> > #include <linux/compat.h>
> > #include <linux/mbus.h>
> > +#include <asm-generic/gpio.h>
> >
> > DECLARE_GLOBAL_DATA_PTR;
> >
> > @@ -274,6 +275,9 @@ struct mvneta_port {
> > int init;
> > int phyaddr;
> > struct phy_device *phydev;
> > +#ifdef CONFIG_DM_GPIO
> > + struct gpio_desc phy_reset_gpio;
> > +#endif
> > struct mii_dev *bus;
> > };
> >
> > @@ -1733,6 +1737,17 @@ static int mvneta_probe(struct udevice *dev)
> > pp->phyaddr = fdtdec_get_int(blob, addr, "reg", 0);
> > }
> >
> > +#ifdef CONFIG_DM_GPIO
> > + gpio_request_by_name(dev, "phy-reset-gpios", 0,
> > + &pp->phy_reset_gpio, GPIOD_IS_OUT);
> > +
> > + if (dm_gpio_is_valid(&pp->phy_reset_gpio)) {
> > + dm_gpio_set_value(&pp->phy_reset_gpio, 1);
> > + mdelay(3000);
> > + dm_gpio_set_value(&pp->phy_reset_gpio, 0);
> > + }
> > +#endif
> > +
> > bus = mdio_alloc();
> > if (!bus) {
> > printf("Failed to allocate MDIO bus\n");
> > ---
> > I intentionally put that 3 seconds delay so i can observe the
> > RJ45 LED. After pressing the reset button, the LED never turned off.
> > It turned off only when i access the network, for example tftpboot.
> >
> > here are snippet from the serial log
> > ----
> > Net:
> > Warning: ethernet@70000 (eth1) using random MAC address - ae:c1:5a:4e:ba:00
> > eth1: ethernet@70000
> > Hit any key to stop autoboot: 0
> > => tftpboot
> > ethernet@70000 Waiting for PHY auto negotiation to complete....... done
> > *** ERROR: `serverip' not set
> > =>
> > ----
> > The LED turned off during "Waiting for PHY auto negotiation"
>
> This should work. Did you correctly add the GPIO to your DT? Is the
> GPIO correctly referenced? Did you check this in the GPIO driver?
>
You're right. My fault, I put the gpio under phy node not the ethernet node.
After move the gpio into ethernet node, it works.
I will update this series and send the mvneta changes as separate patch.
Regards,
Aditya
> Thanks,
> Stefan
_______________________________________________
U-Boot mailing list
U-Boot@lists.denx.de
https://lists.denx.de/listinfo/u-boot
