On Mon, 5 Nov 2007, Rafael J. Wysocki wrote: > On Monday, 5 of November 2007, Alan Stern wrote: > > Rafael: > > > > This patch adds the icebox, for use by kernel threads that want to > > freeze themselves without using the freezer -- which is likely to crop > > up when the freezer gets eliminated. > > > > It's straightforward enough. It could be used as-is, for example for > > freezing workqueue threads. Using it for other sorts of kernel threads > > seems to be rather difficult, requiring a fair amount of extra code > > that's hard to centralize since each thread will have to be woken up > > its own way. > > > > Do you have any particular suggestions? > > I like the idea, but I think that start_icebox() should be called after > freeze_processes(). Similarly, IMO it's better to call stop_icebox() before > thaw_processes(). The same thought occurred to me. But consider: If at some point we do remove the freezer, then freezable kernel threads will rely on the PM_HIBERNATION_PREPARE notification to know when they should go into the icebox. Hence the icebox has to be started before the notifications are sent, and that means before freeze_processes. (In fact, once the freezer is gone it won't even make sense to ask whether start_icebox comes before or after freeze_processes!) The experience of working on this has made we wonder whether we really should get rid of the freezer entirely. Even if it is kept merely for freezing the handful of kernel threads (not user tasks!) that want to be frozen, it's still likely to be more efficient that having a bunch of independent notifier routines, each sending a particular thread to the icebox. If we do keep the freezer then maybe it's okay to start the icebox after the notications are sent. Alan Stern _______________________________________________ linux-pm mailing list linux-pm@xxxxxxxxxxxxxxxxxxxxxxxxxx https://lists.linux-foundation.org/mailman/listinfo/linux-pm
