Re: "PM / wakeup: Show wakeup sources stats in sysfs" causes boot warnings

[Stephen Boyd <swboyd@chromium.org>]

Quoting Qian Cai (2019-08-13 14:32:56)
> The linux-next commit "PM / wakeup: Show wakeup sources stats in sysfs" [1]
> introduced some baddies during boot on several x86 servers. Reverted the commit
> fixed the issue.
> 
> [1] https://lore.kernel.org/lkml/20190807014846.143949-4-trong@android.com/
> 
> [   39.195053][    T1] serio: i8042 KBD port at 0x60,0x64 irq 1
> [   39.197347][    T1] kobject_add_internal failed for wakeup (error: -2 parent:
> serio0)
> [   39.199845][    T1] INFO: trying to register non-static key.
> [   39.201582][    T1] the code is fine but needs lockdep annotation.
> [   39.203477][    T1] turning off the locking correctness validator.
> [   39.205399][    T1] CPU: 12 PID: 1 Comm: swapper/0 Not tainted 5.3.0-rc4-
> next-20190813 #3
> [   39.207938][    T1] Hardware name: HP ProLiant XL420 Gen9/ProLiant XL420
> Gen9, BIOS U19 12/27/2015
> [   39.210606][    T1] Call Trace:
> [   39.210606][    T1]  dump_stack+0x62/0x9a
> [   39.210606][    T1]  register_lock_class+0x95a/0x960
> [   39.210606][    T1]  ? __platform_driver_probe+0xcd/0x230
> [   39.210606][    T1]  ? __platform_create_bundle+0xc0/0xe0
> [   39.210606][    T1]  ? i8042_init+0x4ec/0x578
> [   39.210606][    T1]  ? do_one_initcall+0xfe/0x45a
> [   39.219571][    T1]  ? kernel_init_freeable+0x614/0x6a7
> [   39.219571][    T1]  ? kernel_init+0x11/0x138
> [   39.219571][    T1]  ? ret_from_fork+0x35/0x40
> [   39.219571][    T1]  ? is_dynamic_key+0xf0/0xf0
> [   39.219571][    T1]  ? rwlock_bug.part.0+0x60/0x60
> [   39.219571][    T1]  ? __debug_check_no_obj_freed+0x8e/0x250
> [   39.219571][    T1]  __lock_acquire.isra.13+0x5f/0x830
> [   39.229491][    T1]  ? __debug_check_no_obj_freed+0x152/0x250
> [   39.229491][    T1]  lock_acquire+0x107/0x220
> [   39.229491][    T1]  ? __pm_relax.part.2+0x21/0xa0
> [   39.229491][    T1]  _raw_spin_lock_irqsave+0x35/0x50
> [   39.229491][    T1]  ? __pm_relax.part.2+0x21/0xa0
> [   39.229491][    T1]  __pm_relax.part.2+0x21/0xa0
> [   39.239588][    T1]  wakeup_source_destroy.part.3+0x18/0x190
> [   39.239588][    T1]  wakeup_source_register+0x43/0x50

We shouldn't call wakeup_source_destroy() from the error path in
wakeup_source_register() because that calls __pm_relax() and that takes
a lock that isn't initialized until wakeup_source_add() is called. Can
you try this patch?

----8<----
diff --git a/drivers/base/power/wakeup.c b/drivers/base/power/wakeup.c
index 3a7f5803aa81..f7925820b5ca 100644
--- a/drivers/base/power/wakeup.c
+++ b/drivers/base/power/wakeup.c
@@ -137,6 +137,13 @@ static void wakeup_source_record(struct wakeup_source *ws)
 	spin_unlock_irqrestore(&deleted_ws.lock, flags);
 }
 
+static void wakeup_source_free(struct wakeup_source *ws)
+{
+	ida_free(&wakeup_ida, ws->id);
+	kfree_const(ws->name);
+	kfree(ws);
+}
+
 /**
  * wakeup_source_destroy - Destroy a struct wakeup_source object.
  * @ws: Wakeup source to destroy.
@@ -150,9 +157,7 @@ void wakeup_source_destroy(struct wakeup_source *ws)
 
 	__pm_relax(ws);
 	wakeup_source_record(ws);
-	ida_free(&wakeup_ida, ws->id);
-	kfree_const(ws->name);
-	kfree(ws);
+	wakeup_source_free(ws);
 }
 EXPORT_SYMBOL_GPL(wakeup_source_destroy);
 
@@ -217,7 +222,7 @@ struct wakeup_source *wakeup_source_register(struct device *dev,
 	if (ws) {
 		ret = wakeup_source_sysfs_add(dev, ws);
 		if (ret) {
-			wakeup_source_destroy(ws);
+			wakeup_source_free(ws);
 			return NULL;
 		}
 		wakeup_source_add(ws);