irq: Remove IRQ_EXIT_OFFSET workaround
authorFrederic Weisbecker <fweisbec@gmail.com>
Thu, 21 Feb 2013 23:05:07 +0000 (00:05 +0100)
committerFrederic Weisbecker <fweisbec@gmail.com>
Thu, 21 Feb 2013 23:05:07 +0000 (00:05 +0100)
The IRQ_EXIT_OFFSET trick was used to make sure the irq
doesn't get preempted after we substract the HARDIRQ_OFFSET
until we are entirely done with any code in irq_exit().

This workaround was necessary because some archs may call
irq_exit() with irqs enabled and there is still some code
in the end of this function that is not covered by the
HARDIRQ_OFFSET but want to stay non-preemptible.

Now that irq are always disabled in irq_exit(), the whole code
is guaranteed not to be preempted. We can thus remove this hack.

Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Ingo Molnar <mingo@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
include/linux/hardirq.h
kernel/softirq.c

index 29eb805..c1d6555 100644 (file)
 
 #ifdef CONFIG_PREEMPT_COUNT
 # define preemptible() (preempt_count() == 0 && !irqs_disabled())
-# define IRQ_EXIT_OFFSET (HARDIRQ_OFFSET-1)
 #else
 # define preemptible() 0
-# define IRQ_EXIT_OFFSET HARDIRQ_OFFSET
 #endif
 
 #if defined(CONFIG_SMP) || defined(CONFIG_GENERIC_HARDIRQS)
index 24a921b..f42ff97 100644 (file)
@@ -343,7 +343,7 @@ void irq_exit(void)
 
        account_irq_exit_time(current);
        trace_hardirq_exit();
-       sub_preempt_count(IRQ_EXIT_OFFSET);
+       sub_preempt_count(HARDIRQ_OFFSET);
        if (!in_interrupt() && local_softirq_pending())
                invoke_softirq();
 
@@ -353,7 +353,6 @@ void irq_exit(void)
                tick_nohz_irq_exit();
 #endif
        rcu_irq_exit();
-       sched_preempt_enable_no_resched();
 #ifndef __ARCH_IRQ_EXIT_IRQS_DISABLED
        local_irq_restore(flags);
 #endif