if CONFIG_TRACE_IRQFLAGS opened, we got that warning:
[  214.078880] ------------[ cut here ]------------

[  214.083521] WARNING: CPU: 4 PID: 1 at kernel/locking/lockdep.c:3557
check_flags.part.40+0x1c0/0x1d0()

        [  214.092778] DEBUG_LOCKS_WARN_ON(current->hardirqs_enabled)

        [  214.098110] Modules linked in:

        [  214.101373] CPU: 4 PID: 1 Comm: systemd Tainted: G        W
        4.1.15+ #15

        [  214.108537] Hardware name: Hisilicon Hi1610 16core Pangea
        Board (DT)

        [  214.114916] Call trace:

        [  214.117373] [<ffffffc00008af10>] dump_backtrace+0x0/0x1a0

        [  214.122794] [<ffffffc00008b0d0>] show_stack+0x20/0x30

        [  214.127867] [<ffffffc000b53454>] dump_stack+0x8c/0xb8

        [  214.132939] [<ffffffc0000c2308>]
        warn_slowpath_common+0xa8/0xe0

        [  214.138882] [<ffffffc0000c23b4>] warn_slowpath_fmt+0x74/0xa0

        [  214.144563] [<ffffffc000121c20>]
        check_flags.part.40+0x1c0/0x1d0

        [  214.150594] [<ffffffc000122760>] lock_is_held+0x80/0xa0

        [  214.155839] [<ffffffc00015125c>]
        rcu_note_context_switch+0x3dc/0x410

        [  214.162219] [<ffffffc000b57dcc>] __schedule+0x5c/0xb00

        [  214.167377] [<ffffffc000b588b0>] schedule+0x40/0xa0

        [  214.172272] ---[ end trace 2d1dc460f8460cd7 ]---

        [  214.176906] possible reason: unannotated irqs-off.

It's because irqflags is not the same as current->hardirqs_enabled,
I think maybe use local_irq_XXX to instead raw_local_irq_XXX can fix
that warning

Signed-off-by: Wang Yufen <wangyu...@huawei.com>
Signed-off-by: Ma Jun <majun...@huawei.com>
---
 kernel/locking/lockdep.c |    4 ++--
 1 files changed, 2 insertions(+), 2 deletions(-)

diff --git a/kernel/locking/lockdep.c b/kernel/locking/lockdep.c
index 716547f..1ed8b35 100644
--- a/kernel/locking/lockdep.c
+++ b/kernel/locking/lockdep.c
@@ -3620,13 +3620,13 @@ int lock_is_held(struct lockdep_map *lock)
        if (unlikely(current->lockdep_recursion))
                return 1; /* avoid false negative lockdep_assert_held() */
 
-       raw_local_irq_save(flags);
+       local_irq_save(flags);
        check_flags(flags);
 
        current->lockdep_recursion = 1;
        ret = __lock_is_held(lock);
        current->lockdep_recursion = 0;
-       raw_local_irq_restore(flags);
+       local_irq_restore(flags);
 
        return ret;
 }
-- 
1.7.1


Reply via email to