Hi Shimoda-san, On Fri, Aug 31, 2018 at 8:50 AM Yoshihiro Shimoda <yoshihiro.shimoda.uh@xxxxxxxxxxx> wrote: > R-Car Gen3 needs to enable clocks of both host and peripheral. > Since [eo]hci-platform disables the reset(s) when the drivers are > removed, renesas_usbhs driver doesn't work correctly. To fix this > issue, this patch adds multiple clocks management on this > renesas_usbhs driver. > > Signed-off-by: Yoshihiro Shimoda <yoshihiro.shimoda.uh@xxxxxxxxxxx> Thanks for your patch! > --- a/drivers/usb/renesas_usbhs/common.c > +++ b/drivers/usb/renesas_usbhs/common.c > @@ -5,6 +5,7 @@ > * Copyright (C) 2011 Renesas Solutions Corp. > * Kuninori Morimoto <kuninori.morimoto.gx@xxxxxxxxxxx> > */ > +#include <linux/clk.h> > #include <linux/err.h> > #include <linux/gpio.h> > #include <linux/io.h> > @@ -336,11 +337,23 @@ static void usbhsc_power_ctrl(struct usbhs_priv *priv, int enable) > { > struct platform_device *pdev = usbhs_priv_to_pdev(priv); > struct device *dev = usbhs_priv_to_dev(priv); > + int ret; > > if (enable) { > /* enable PM */ > pm_runtime_get_sync(dev); > > + /* enable clks if exist */ > + if (priv->clks) { > + ret = clk_bulk_prepare(priv->num_clks, priv->clks); > + if (!ret) { > + ret = clk_bulk_enable(priv->num_clks, > + priv->clks); > + if (ret) > + return; clk_bulk_prepare_enable(), which also takes care of unpreparing all clocks if any clock fails to enabled. > + } > + } > + > /* enable platform power */ > usbhs_platform_call(priv, power_ctrl, pdev, priv->base, enable); > > @@ -353,6 +366,12 @@ static void usbhsc_power_ctrl(struct usbhs_priv *priv, int enable) > /* disable platform power */ > usbhs_platform_call(priv, power_ctrl, pdev, priv->base, enable); > > + /* disable clks if exist */ > + if (priv->clks) { > + clk_bulk_disable(priv->num_clks, priv->clks); > + clk_bulk_unprepare(priv->num_clks, priv->clks); clk_bulk_disable_unprepare() > + } > + > /* disable PM */ > pm_runtime_put_sync(dev); > } > @@ -534,6 +553,11 @@ static struct renesas_usbhs_platform_info *usbhs_parse_dt(struct device *dev) > return info; > } > > +static const struct clk_bulk_data usbhs_rcar3_clks[] = { > + { .id = "hsusb" }, > + { .id = "ehci/ohci" }, > +}; This structure contains space for the clock pointers, which are not really needed here (the struct is copied). Perhaps you can replace "struct clk_bulk_data *clks" in struct usbhs_priv replace by "struct clk_bulk_data clks[2]", ... > + > static int usbhs_probe(struct platform_device *pdev) > { > struct renesas_usbhs_platform_info *info = renesas_usbhs_get_info(pdev); > @@ -620,6 +644,15 @@ static int usbhs_probe(struct platform_device *pdev) > break; > } > > + if (priv->dparam.type == USBHS_TYPE_RCAR_GEN3 || > + priv->dparam.type == USBHS_TYPE_RCAR_GEN3_WITH_PLL) { > + priv->clks = devm_kmemdup(&pdev->dev, usbhs_rcar3_clks, > + sizeof(usbhs_rcar3_clks), GFP_KERNEL); > + if (!priv->clks) > + return -ENOMEM; ... and instead of allocating/copying, fill in the clock names here: priv->clks[0].id = "hsusb"; priv->clks[1].id = "ehci/ohci"; As priv->clks will no longer be a pointer, all checks for it should be replaced by checks for priv->num_clks. What do you think? Gr{oetje,eeting}s, Geert -- Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- geert@xxxxxxxxxxxxxx In personal conversations with technical people, I call myself a hacker. But when I'm talking to journalists I just say "programmer" or something like that. -- Linus Torvalds