The 11/17/2021 10:52, Philipp Zabel wrote:

Hi Horatio,

Hi Phillip,

On Wed, 2021-11-17 at 10:18 +0100, Horatiu Vultur wrote:

quoted

+static int lan966x_reset_switch(struct lan966x *lan966x)
+{
+     struct reset_control *reset;
+     int val = 0;
+     int ret;
+
+     reset = devm_reset_control_get_shared(lan966x->dev, "switch");
+     if (IS_ERR(reset))
+             dev_warn(lan966x->dev, "Could not obtain switch reset: %ld\n",
+                      PTR_ERR(reset));
+     else
+             reset_control_reset(reset);

According to the device tree bindings, both resets are required.
I'd expect this to return on error.
Is there any chance of the device working with out the switch reset
being triggered?

The only case that I see is if the bootloader triggers this switch
reset and then when bootloader starts the kernel and doesn't set back
the switch in reset. Is this a valid scenario or is a bug in the
bootloader?

quoted

+
+     reset = devm_reset_control_get_shared(lan966x->dev, "phy");
+     if (IS_ERR(reset)) {
+             dev_warn(lan966x->dev, "Could not obtain phy reset: %ld\n",
+                      PTR_ERR(reset));
+     } else {
+             reset_control_reset(reset);
+     }

Same as above.
Consider printing errors with %pe or dev_err_probe().

quoted

+     lan_wr(SYS_RESET_CFG_CORE_ENA_SET(0), lan966x, SYS_RESET_CFG);
+     lan_wr(SYS_RAM_INIT_RAM_INIT_SET(1), lan966x, SYS_RAM_INIT);
+     ret = readx_poll_timeout(lan966x_ram_init, lan966x,
+                              val, (val & BIT(1)) == 0, READL_SLEEP_US,
+                              READL_TIMEOUT_US);
+     if (ret)
+             return ret;
+
+     lan_wr(SYS_RESET_CFG_CORE_ENA_SET(1), lan966x, SYS_RESET_CFG);
+
+     return 0;
+}
+
+static int lan966x_probe(struct platform_device *pdev)
+{
+     struct fwnode_handle *ports, *portnp;
+     struct lan966x *lan966x;
+     int err, i;
+
+     lan966x = devm_kzalloc(&pdev->dev, sizeof(*lan966x), GFP_KERNEL);
+     if (!lan966x)
+             return -ENOMEM;
+
+     platform_set_drvdata(pdev, lan966x);
+     lan966x->dev = &pdev->dev;
+
+     ports = device_get_named_child_node(&pdev->dev, "ethernet-ports");
+     if (!ports) {
+             dev_err(&pdev->dev, "no ethernet-ports child not found\n");
+             err = -ENODEV;
+             goto out;

No need to goto as long as there's just a "return err;" after the out:
label.

True, I will udate this.

quoted

+     }
+
+     err = lan966x_create_targets(pdev, lan966x);
+     if (err)
+             goto out;
+
+     if (lan966x_reset_switch(lan966x)) {
+             err = -EINVAL;

This should propagate the error returned from lan966x_reset_switch()
instead.

I will fix it in the next version.

quoted

+             goto out;
+     }
+
+     i = 0;
+     fwnode_for_each_available_child_node(ports, portnp)
+             ++i;
+
+     lan966x->num_phys_ports = i;
+     lan966x->ports = devm_kcalloc(&pdev->dev, lan966x->num_phys_ports,
+                                   sizeof(struct lan966x_port *),
+                                   GFP_KERNEL);

        if (!lan966x->ports)
                return -ENOMEM;

Good catch.

regards
Philipp

-- 
/Horatiu