On Thu, Feb 12, 2015 at 09:09:30AM +0800, Viresh Kumar wrote:
On 11 February 2015 at 18:38, Peter Zijlstra peterz@infradead.org wrote:
I think its better to be strict; esp. with new interfaces. It avoids confusion.
Suppose a driver writer sees these new methods and thinks to use one while still having the set_mode() one -- ie. he didn't actually read the comment. We'd better make sure he fails and goes back to read it.
Okay, does below diff looks good to you ?
diff --git a/kernel/time/clockevents.c b/kernel/time/clockevents.c index e646fbe44b05..75d221c7e9cc 100644 --- a/kernel/time/clockevents.c +++ b/kernel/time/clockevents.c @@ -444,8 +444,12 @@ EXPORT_SYMBOL_GPL(clockevents_unbind); static int clockevents_sanity_check(struct clock_event_device *dev) { /* Legacy set_mode() callback */
if (dev->set_mode)
if (dev->set_mode) {/* We shouldn't be supporting new modes now */WARN_ON(dev->set_mode_periodic || dev->set_mode_oneshot ||dev->set_mode_shutdown || dev->set_mode_resume); return 0;} if (dev->features & CLOCK_EVT_FEAT_DUMMY) return 0;
Yep, that looks entirely reasonable.