Hi Thomas/Ingo,
This is in response to the suggestions Ingo gave [1] on the shortcomings of clockevents core's state machine.
This first separates out the RESUME functionality from other states as its a special case. Then it defines a new enum to map possible states of a clockevent device.
Ideally it should only be available for the core, but as bL switcher is using it today, it is exposed in clockchips.h. That dependency will go away after applying the Thomas's work (Sent out be Peter) and then we can move this enum to somewhere in kernel/time/.
The last patch moves the legacy check to the legacy code.
Please see if this meets your expectation or if you have some suggestions on it.
Rebased of tip/master as there were some dependencies: 575daeea39a3 Merge branch 'tools/kvm'
This along with migration of few clockevents drivers is pushed here: git://git.kernel.org/pub/scm/linux/kernel/git/vireshk/linux.git clkevt/manage-state
-- Viresh
[1] https://lkml.org/lkml/2015/2/20/107
Viresh Kumar (3): clockevents: Handle tick device's resume separately clockevents: Manage device's state separately for the core clockevents: Don't validate dev->mode against CLOCK_EVT_MODE_UNUSED for new interface
arch/arm/common/bL_switcher.c | 8 +-- include/linux/clockchips.h | 48 +++++++++++------ kernel/time/clockevents.c | 118 ++++++++++++++++++++++++------------------ kernel/time/tick-broadcast.c | 22 ++++---- kernel/time/tick-common.c | 9 ++-- kernel/time/tick-internal.h | 1 + kernel/time/tick-oneshot.c | 6 +-- kernel/time/timer_list.c | 16 +++--- 8 files changed, 134 insertions(+), 94 deletions(-)