On Thu, Jul 16, 2009 at 07:00:12PM -0700, Elina Pasheva wrote: > > On Friday, July 09 2009 Greg wrote: > >Do the autosuspend on a per-device basis, not allowing it for devices > that do not support it. > >As you know which ones do and do not handle this properly, just add > them to the tables in the > >driver. > > > The sierra.c driver is sufficient to interact over USB interfaces for all of > our modem products, so one driver fits all. However many of our modems today > do not support Selective Suspend – they may crash or behave unpredictably when > suspended - but they may eventually support it. Creating two separate drivers, > one for SS modems and another for non-SS modems is not an option either because > of the configuration management nightmare that would create. No, don't create 2 drivers! > We also do not > want to make any assumptions about the way a particular Linux kernel is > configured for Power Management; our experience with Ubuntu is that power/level > is ON so far, but other distros may be different. We like to retain some > control over our own destiny by using the module parameter as a kind of > master-enable. The default setting is to disable autosuspend so that the > majority of users of this driver don’t see problems if Autosuspend is > somehow enabled on their system. Those who want Selective Suspend and who > know what they are doing can enable the parameter in the driver. > So far, of all our customers, only one needs this feature and that’s why we’ve > been working so hard on it lately. > > As far as whether a driver-wide or a device-wide parameter is the way to go, > I think you could argue it either way. But since implementing it on a > per-device basis can be configured to work the same as the driver-wide case, > we are looking at ways to do this. One approach is to add a device attribute > with read/write capabilities that will have specific value based on the device. > We’d be interested in getting your opinion as to whether this approach may > meet with further opposition. Yes, per-device attribute would work, but please follow the standard api for this (I think there is one for USB devices, can't recall it at the moment, sorry.) thanks, greg k-h -- To unsubscribe from this list: send the line "unsubscribe linux-usb" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
