I found that PHY's enable refcounting was broken and after fixing it I also found that machine started to hang after EHCI driver module removal. Turned out that the teardown order is incorrect because HCD must be unregistered *before* PHY's disabling. Note that it is also not correct to assert the shared reset during of driver's removal because PHY takes care of resetting shared pads and thus it's better to remove that part from the EHCI driver. Signed-off-by: Dmitry Osipenko <digetx@xxxxxxxxx> --- drivers/usb/host/ehci-tegra.c | 8 +------- 1 file changed, 1 insertion(+), 7 deletions(-) diff --git a/drivers/usb/host/ehci-tegra.c b/drivers/usb/host/ehci-tegra.c index 4d2cdec4cb78..32483bef046b 100644 --- a/drivers/usb/host/ehci-tegra.c +++ b/drivers/usb/host/ehci-tegra.c @@ -521,16 +521,10 @@ static int tegra_ehci_remove(struct platform_device *pdev) struct tegra_ehci_hcd *tegra = (struct tegra_ehci_hcd *)hcd_to_ehci(hcd)->priv; + usb_remove_hcd(hcd); otg_set_host(hcd->usb_phy->otg, NULL); - usb_phy_shutdown(hcd->usb_phy); - usb_remove_hcd(hcd); - - reset_control_assert(tegra->rst); - udelay(1); - clk_disable_unprepare(tegra->clk); - usb_put_hcd(hcd); return 0; -- 2.24.0
