> I don't agree with changes #1 and #2. > > If you are going to go to a model where every single configuration operation > is recorded in software and performed at resume time, then really do it and > fix it in the whole driver. As currently coded you are leaving lots of known > bugs in the driver. > > #2 is even a worse situation. If you are going to handle things this way you > must sync the counters when the suspend happens, so that the statistics get > call receives up to date values.
Thanks for your advice. I will send #3 patch first. Then modify #1 and #2 patch and resend these two patches. ------Please consider the environment before printing this e-mail.
