On Wed, Jun 12, 2013 at 04:06:47PM +0400, Kirill Tkhai wrote: > There are many constructions like: > > spin_unlock_irq(lock); > schedule(); > > In case of preemptible kernel we check if task needs reschedule > at the end of spin_unlock(). So if TIF_NEED_RESCHED is set > we call schedule() twice and we have a little overhead here. > Add primitives to avoid these situations. > > Signed-off-by: Kirill Tkhai <[email protected]> > CC: Steven Rostedt <[email protected]> > CC: Ingo Molnar <[email protected]> > CC: Peter Zijlstra <[email protected]> > --- > include/linux/spinlock.h | 27 +++++++++++++++++++++++++++ > include/linux/spinlock_api_smp.h | 37 +++++++++++++++++++++++++++++++++++++ > include/linux/spinlock_api_up.h | 13 +++++++++++++ > kernel/spinlock.c | 20 ++++++++++++++++++++ > 4 files changed, 97 insertions(+), 0 deletions(-) > diff --git a/include/linux/spinlock.h b/include/linux/spinlock.h > index 7d537ce..35caa32 100644 > --- a/include/linux/spinlock.h > +++ b/include/linux/spinlock.h > @@ -221,13 +221,24 @@ static inline void do_raw_spin_unlock(raw_spinlock_t > *lock) __releases(lock) > #define raw_spin_lock_irq(lock) _raw_spin_lock_irq(lock) > #define raw_spin_lock_bh(lock) _raw_spin_lock_bh(lock) > #define raw_spin_unlock(lock) _raw_spin_unlock(lock) > +#define raw_spin_unlock_no_resched(lock) \ > + _raw_spin_unlock_no_resched(lock) > + > #define raw_spin_unlock_irq(lock) _raw_spin_unlock_irq(lock) > +#define raw_spin_unlock_irq_no_resched(lock) \ > + _raw_spin_unlock_irq_no_resched(lock) > > #define raw_spin_unlock_irqrestore(lock, flags) \ > do { \ > typecheck(unsigned long, flags); \ > _raw_spin_unlock_irqrestore(lock, flags); \ > } while (0) > +#define raw_spin_unlock_irqrestore_no_resched(lock, flags) \ > + do { \ > + typecheck(unsigned long, flags); \ > + _raw_spin_unlock_irqrestore_no_resched(lock, flags); \ > + } while (0)
So I absolutely hate this API because people can (and invariably will) abuse it; much like they did/do preempt_enable_no_resched(). IIRC Thomas even maps preempt_enable_no_resched() to preempt_enable() in -rt to make sure we don't miss preemption points due to stupidity. He converted the 'few' sane sites to use schedule_preempt_disabled(). In that vein, does it make sense to introduce schedule_spin_locked()? Also, your patch 'fails' to make use of the new API. -- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to [email protected] More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/
