On 6/7/2023 5:14 PM, Johan Hovold wrote:
On Fri, May 26, 2023 at 08:55:22PM +0530, Krishna Kurapati PSSNV wrote:
On 5/26/2023 8:25 AM, Bjorn Andersson wrote:
We need to fix the dwc3 glue design, so that the glue and the core can
cooperate - and we have a few other use cases where this is needed (e.g.
usb_role_switch propagation to the glue code).
Thanks for the comments on this patch. I had some suggestions come in
from the team internally:
1. To use the notifier call available in drivers/usb/core/notify.c and
make sure that host mode is enabled. That way we can access dwc or xhci
without any issue.
I don't think this is a good idea and instead the callbacks should be
dedicated for the xhci and dwc3 drivers. A struct with callbacks can be
passed down to the child devices, which call back into the drivers of
their parents for notifications and when they need services from them
(e.g. during suspend or on role changes).
Hi Johan,
While I agree with you that these notifications are to be used during
role switch or suspend/resume, there is no restriction on using them for
checking whether we are in host mode or not. IMO, it would be cleaner as
we won't be dereferencing dwc driver data at all to check if we are in
host mode or not.
Regards,
Krishna,
2. For this particular case where we are trying to get info on number of
ports present (dwc->num_usb2_ports), we can add compatible data for
sc8280-mp and provide input to driver telling num ports is 4.
That may also work as a way to avoid parsing the xhci registers, but I'm
still not sure why simply counting the PHYs in DT would not work.
Johan
