Hi,
On Wed, Sep 22, 2021, at 22:54, Marc Zyngier wrote:
From: Alyssa Rosenzweig <redacted>
[...]
quoted hunk ↗ jump to hunk
+
+ /* Use the first reg entry to work out the port index */
+ port->idx = idx >> 11;
+ port->pcie = pcie;
+ port->np = np;
+
+ port->base = devm_platform_ioremap_resource(platform, port->idx + 2);
+ if (IS_ERR(port->base))
+ return -ENODEV;
+
+ rmw_set(PORT_APPCLK_EN, port + PORT_APPCLK);
I think this should be
rmw_set(PORT_APPCLK_EN, port->base + PORT_APPCLK);
quoted hunk ↗ jump to hunk
+
+ rmw_set(PORT_PERST_OFF, port->base + PORT_PERST);
+ gpiod_set_value(reset, 1);
+
+ ret = readl_relaxed_poll_timeout(port->base + PORT_STATUS, stat,
+ stat & PORT_STATUS_READY, 100, 250000);
+ if (ret < 0) {
+ dev_err(pcie->dev, "port %pOF ready wait timeout\n", np);
+ return ret;
+ }
+
+ /* Flush writes and enable the link */
+ dma_wmb();
I don't think this barrier is required.
quoted hunk ↗ jump to hunk
+
+ writel_relaxed(PORT_LTSSMCTL_START, port->base + PORT_LTSSMCTL);
+
+ return 0;
+}
+
[...]
Looks good to me otherwise,
Reviewed-by: Sven Peter <redacted>
Thanks,
Sven