On 9/13/22 22:34, Vinod Koul wrote:
> On 06-09-22, 10:43, JC Kuo wrote:
>> Hi Vinod,
>
> Please _do_ _not_ _top_ _post_
>
>> Before the device or host is being attached, we can keep most of the
>> transceivers powered down (PD=1/PD_DR=1) to minimize power consumption. At this
>> stage, in .phy_power_on(), we enable only the single-ended receiver (PD_ZI=0)
>> for detecting connection. Upon detecting device's or host's connection, host or
>> controller driver will invoke tegra_phy_xusb_utmi_pad_power_on() to power on all
>> of the transceivers (PD=0/PD_DR=0) to equip full link functionality.
>
> Thanks for this explanation... It helps!
>
> Just a suggestion, can this be moved into phy_init() you have detected
> connection in phy_power_on(), the transceiver can be enabled in
> phy_int... Would that work?
> That would work, too. However, because Tegra USB has separate phys for USB3 SS
and USB2, I'd like to keep the USB2 phy operations as they are now, so that USB
host and device controller drivers do not have to distinguish the phy type and
invoke different phy stubs. Furthermore, PD_ZI=0 does really power on the USB2
phy, partially.
For example:
1. in .probe(),
for_each_usb_phy {
phy_init(phy);
}
for_each_usb3_phy {
phy_power_on(phy);
};
2. upon detecting connection,
phy_power_on(the_target_usb2_phy);
Thanks,
JC
