12.06.2021 10:34, Peter Chen пишет:
> On 21-06-09 15:04:04, Dmitry Osipenko wrote:
>> The ChipIdea driver enters into suspend immediately after seeing a
>> VBUS disconnection. Some devices need an extra delay after losing
>> VBUS, otherwise VBUS may be floating, preventing the PHY's suspending
>> by the VBUS detection sensors. This problem was found on Tegra30 Asus
>> Transformer TF700T tablet device, where the USB PHY wakes up immediately
>> from suspend because VBUS sensor continues to detect VBUS as active after
>> disconnection. A minimum delay of 20ms is needed in order to fix this
>> issue, hence add 25ms delay before suspending the PHY.
>>
>> Reported-by: Maxim Schwalm <maxim.schwalm@xxxxxxxxx> # Asus TF700T
>> Tested-by: Maxim Schwalm <maxim.schwalm@xxxxxxxxx> # Asus TF700T
>> Signed-off-by: Dmitry Osipenko <digetx@xxxxxxxxx>
>> ---
>> drivers/usb/chipidea/ci_hdrc_tegra.c | 8 ++++++++
>> 1 file changed, 8 insertions(+)
>>
>> diff --git a/drivers/usb/chipidea/ci_hdrc_tegra.c b/drivers/usb/chipidea/ci_hdrc_tegra.c
>> index 60361141ac04..d1359b76a0e8 100644
>> --- a/drivers/usb/chipidea/ci_hdrc_tegra.c
>> +++ b/drivers/usb/chipidea/ci_hdrc_tegra.c
>> @@ -4,6 +4,7 @@
>> */
>>
>> #include <linux/clk.h>
>> +#include <linux/delay.h>
>> #include <linux/io.h>
>> #include <linux/module.h>
>> #include <linux/of_device.h>
>> @@ -255,6 +256,13 @@ static int tegra_ehci_hub_control(struct ci_hdrc *ci, u16 typeReq, u16 wValue,
>>
>> static void tegra_usb_enter_lpm(struct ci_hdrc *ci, bool enable)
>> {
>> + /*
>> + * Give hardware time to settle down after VBUS disconnection,
>> + * otherwise PHY may wake up from suspend immediately.
>> + */
>> + if (enable)
>> + msleep(25);
>> +
>
> How could you know 25ms is enough for other Tegra designs?
I don't know what is the maximum timeout could be, but it shouldn't be a
problem to bump the timeout if somebody will report the need to do so.
> Could you poll VBUS wakeup threshold register to ensure the
> wakeup will not occur?
We indeed can poll the wakeup threshold status in the PHY driver, it
works too. I'll make the patch for the PHY driver, thank you for the
suggestion.
> The similar design exists at function:
> hw_wait_vbus_lower_bsv.
The hw_wait_vbus_lower_bsv uses 5sec timeout, which should be too much.
I'll set the polling timeout to 100ms.
