ACPI: don't cond_resched if irq is disabled
authorXiaotian Feng <dfeng@redhat.com>
Thu, 10 Dec 2009 11:56:45 +0000 (19:56 +0800)
committerLen Brown <len.brown@intel.com>
Sat, 16 Jan 2010 06:35:36 +0000 (01:35 -0500)
commit 8bd108d adds preemption point after each opcode parse, then
a sleeping function called from invalid context bug was founded
during suspend/resume stage. this was fixed in commit abe1dfa by
don't cond_resched when irq_disabled. But recent commit 138d156 changes
the behaviour to don't cond_resched when in_atomic. This makes the
sleeping function called from invalid context bug happen again, which
is reported in http://lkml.org/lkml/2009/12/1/371.

This patch also fixes http://bugzilla.kernel.org/show_bug.cgi?id=14483

Reported-and-bisected-by: Larry Finger <Larry.Finger@lwfinger.net>
Reported-and-bisected-by: Justin P. Mattock <justinmattock@gmail.com>
Signed-off-by: Xiaotian Feng <dfeng@redhat.com>
Acked-by: Alexey Starikovskiy <astarikovskiy@suse.de>
Signed-off-by: Len Brown <len.brown@intel.com>
include/acpi/platform/aclinux.h

index 9d7febd..0946997 100644 (file)
@@ -152,7 +152,7 @@ static inline void *acpi_os_acquire_object(acpi_cache_t * cache)
 #include <linux/hardirq.h>
 #define ACPI_PREEMPTION_POINT() \
        do { \
-               if (!in_atomic_preempt_off()) \
+               if (!in_atomic_preempt_off() && !irqs_disabled()) \
                        cond_resched(); \
        } while (0)