On Thu, 14 Feb 2002, Greg KH wrote: > > * usb_hub_port_debounce() provides the required debounce delay including > > the "restart delay when spurious disconnect detected" behaviour > > Looks good. But why aren't you restarting the timeout when you detect a > disconnect? > Right now it looks if you see a disconnect for 30ms, then > you error out.
Every time the port status is seen disconnected, the delay_time counter _is_ reset to 0 - hence the timeout restarts as required. And delay_time is only advanced when portstatus was connected. The bogus_discon counter is there to detect and treat three (30ms) consecutive status==disconnect samples as real disconnect. It's not in the specs but something like this is required or you would end up looping and resetting the timeout forever, should there be a real disconnect during the debounce delay. Easy to trigger with the dev. board. However, I think there is something which could be done better: currently the portstatus is sampled every 10 ms and the connect _status_ is what we look at. Hence we might miss short transient disconnect/reconnect changes between the sample. I'm going to change this to use the port connect _change_ information. > > Patch created and tested with 2.4.18-pre9. > > Again, I'd like to try this out in 2.5.x for a while before adding it to > 2.4. Absolutely. It's just my target platform is 2.4 and 2.5.x has too many quirks for me at the moment to use it in parallel. Hence stuff gets created and tested for 2.4. Martin _______________________________________________ [EMAIL PROTECTED] To unsubscribe, use the last form field at: https://lists.sourceforge.net/lists/listinfo/linux-usb-devel
