On Mon, 2019-04-29 at 10:53 +0200, Benjamin Tissoires wrote: > On Mon, Apr 29, 2019 at 10:38 AM Jarkko Nikula > <jarkko.nikula@xxxxxxxxxxxxxxx> wrote: > > I got another thought about this. I noticed these input drivers need > > SMBus Host Notify, maybe that explain the PM dependency? If that's the > > only dependency then we could prevent the controller suspend if there is > > a client needing host notify mechanism. IMHO that's less hack than the > > patch to rmi_smbus.c. > > So currently, AFAIK, only Synaptics (rmi4) and Elantech are using > SMBus Host Notify. > So this patch would prevent the same bugs for those 2 vendors, which is good. > > It took me some time to understand why this would be less than a hack. > And indeed, given that Host Notify relies on the I2C connection to be > ready for the IRQ, we can not put the controller in suspend like we do > for others where the IRQ controller is still ready. > > So yes, that could work from me. Not sure what Wolfram and Jean would > say though. I would say OK with me, this looks like the cleanest solution to me, so if testing is positive, let's go with it. -- Jean Delvare SUSE L3 Support
