> From: Alan Stern <stern@xxxxxxxxxxxxxxxxxxx> > Sent: Monday, January 22, 2024 10:31 AM > To: yuan linyu <yuanlinyu@xxxxxxxxxxx> > Cc: Thinh Nguyen <Thinh.Nguyen@xxxxxxxxxxxx>; Greg Kroah-Hartman > <gregkh@xxxxxxxxxxxxxxxxxxx>; linux-usb@xxxxxxxxxxxxxxx > Subject: Re: [RFC PATCH] usb: udc: run disconnect callback before pull up zero > > > > > What is host error ? > > If the host is trying to communicate with the gadget at the time of the > mode switch, it will see that something is wrong because the gadget will > still be connected (the pullup will still be on) but it won't reply to > any USB packets. > Host will lost device soon. > > > > It is already change to dev_dbg() by Wesley Cheng for dwc3. > > But it also can enable by change log level. Only delete it will not show again. > > I think it's good to have those messages show up when the log level is > set to debug. They let developers know what's happening when they test > changes to the gadget drivers. If the messages really bother somebody, > all they have to do is reduce the log level to info. > > If you're still concerned about the behavior of the mass-storage > function, you could change it. Make it disable itself whenever it gets > a -ESHUTDOWN error, either while submitting a request or as a completion > status. This should reduce the number of error messages, although it > won't eliminate them. Thanks, will send a patch for mass storage only. > > (Of course, this still leaves the possibility of floods of debugging > messages from all the other function drivers...) > > > this thread just want to discuss if disable eps first then pull up zero acceptable > or > > good (reduce mode switch time ???) for all UDCs ? > > I think it's not a good thing to do. And I don't think it will reduce > the mode switch time, because both operations still have to occur so > they will require the same total amount of time. > > Alan Stern
