linux/arch
Dimitri Sivanich 71416bea5a [IA64] disable irq's and check need_resched before safe_halt
While sending interrupts to a cpu to repeatedly wake a thread, on occasion
that thread will take a full timer tick cycle (4002 usec in my case)
to wakeup.

The problem concerns a race condition in the code around the safe_halt()
call in the default_idle() routine.  Setting 'nohalt' on the kernel
command line causes the long wakeups to disappear.

void
default_idle (void)
{
        local_irq_enable();
        while (!need_resched()) {
-->             if (can_do_pal_halt)
-->                     safe_halt();
                else

A timer tick could arrive between the check for !need_resched and the
actual call to safe_halt() (which does a pal call to PAL_HALT_LIGHT).
By the time the timer tick completes, a thread that might now need to run
could get held up for as long as a timer tick waiting for the halted cpu.

I'm proposing that we disable irq's and check need_resched again before
calling safe_halt().  Does anyone see any problem with this approach?

Signed-off-by: Dimitri Sivanich <sivanich@sgi.com>
Signed-off-by: Tony Luck <tony.luck@intel.com>
2007-08-13 10:17:23 -07:00
..
alpha alpha: -Werror fixes for sys_titan.c 2007-08-11 15:47:42 -07:00
arm [ARM] rpc: update defconfig 2007-08-06 16:48:49 +01:00
avr32 Merge branch 'for-linus' of git://git.o-hand.com/linux-rpurdie-leds 2007-07-22 11:22:01 -07:00
blackfin Blackfin arch: after removing fs.h from mm.h, fix the broken on Blackfin arch 2007-08-11 15:47:39 -07:00
cris cris: drivers/cdrom/Kconfig no longer exists 2007-08-11 15:47:41 -07:00
frv FRV: connect up fallocate 2007-08-11 15:47:40 -07:00
h8300 PTRACE_POKEDATA consolidation 2007-07-17 10:23:03 -07:00
i386 genirq: mark io_apic level interrupts to avoid resend 2007-08-12 11:05:45 -07:00
ia64 [IA64] disable irq's and check need_resched before safe_halt 2007-08-13 10:17:23 -07:00
m32r m32r: build fix for removing fs.h from mm.h 2007-07-30 10:52:10 -07:00
m68k Remove fs.h from mm.h 2007-07-29 17:09:29 -07:00
m68knommu m68knommu: get rid of duplicate include 2007-07-31 15:39:40 -07:00
mips [MIPS] Add smp_call_function_single() 2007-07-31 21:35:37 +01:00
parisc Remove fs.h from mm.h 2007-07-29 17:09:29 -07:00
powerpc [POWERPC] Fix size check for hugetlbfs 2007-08-10 21:04:42 +10:00
ppc Merge git://git.kernel.org/pub/scm/linux/kernel/git/wim/linux-2.6-watchdog 2007-07-31 20:43:52 -07:00
s390 [S390] remove DEFAULT_MIGRATION_COST 2007-08-10 14:32:37 +02:00
sh sh: fix defconfigs for sh7751r boards 2007-08-03 10:17:47 +09:00
sh64 sh64: Kill off virt_to_bus()/bus_to_virt(). 2007-07-31 13:11:25 +09:00
sparc [SPARC]: Centralize find_in_proplist() instead of duplicating N times. 2007-08-07 18:46:36 -07:00
sparc64 [SPARC64]: Fix memory leak when cpu hotplugging. 2007-08-08 17:33:52 -07:00
um UML: console should handle spurious IRQS 2007-07-31 15:39:37 -07:00
v850 PTRACE_POKEDATA consolidation 2007-07-17 10:23:03 -07:00
x86_64 genirq: mark io_apic level interrupts to avoid resend 2007-08-12 11:05:45 -07:00
xtensa Merge git://git.kernel.org/pub/scm/linux/kernel/git/sam/kbuild 2007-07-19 14:28:19 -07:00