On Thu, Aug 06, 2020 at 05:43:54AM +0000, Asano, Yasushi (ADITJ/SWG) wrote: > Dear Alan > Dear Greg > > Thank you for your feedback. > I really appreciate your concrete proposal. > > > So let's change the code to do 3 tries with each scheme. > I understood. I will try to modify it so that the number of > attempts will decrease. It is 6 attempts in total both old and > new schemes, but msleep is executed at various places in > hub_port_connect and hub_port_init. apart from a timeout. > > For example, msleep(100) is executed every time in the loop of > GET_DESCRIPTOR_TRIES[8] of new scheme. and In the old scheme, > msleep(200) is executed in the loop of SET_ADDRESS_TRIES[10]. > From my measurement, it does not subside within 30 seconds, > but it is around 32 seconds. > > From these things, I would like you to reconsider the number of attempts. > Is it OK to set the new scheme to 3 times and the old scheme to > 2 times(no change as it is)? In other words > > [plan 1] > 3 * new scheme, then 2 * old scheme, or else > 2 * old scheme, then 3 * new_scheme, > depending on the old_scheme_first parameter. Yes, that's all right. Although you might want to make the second case be: 3 * old scheme, then 2 * new scheme. > Also, although it is a "better plan", the original processing is in the following. > > 6 * new scheme, then 6 * new scheme, > then 2 * old scheme, then 2 * old scheme > > if it will be modified from above to below, It seems that the structure > of the loop has to be greatly revised. I think. > > 2 * new scheme, then 2 * old scheme, > then 1 * new scheme, then 1 * old scheme If you want to use only five attempts, you'll have to get rid of the last one. > The fix is likely to be large, so Can I proceed with a patch in plan 1? Okay. Alan Stern > I will post the patch after confirming the behavior of the patch with > the customer board with the PET tool. please give me a little time. > > Best Regards > Yasushi Asano
