Re: 6.16-rc2+ lockdep circular locking between iwlwifi and thermal_zone code

[Benjamin Berg <benjamin@xxxxxxxxxxxxxxxx>]

Hi Hans,


On Fri, 2025-06-20 at 17:56 +0200, Hans de Goede wrote:
> While testing 6.16-rc2+ on a Dell XPS 9640 I got the following lockdep report:
> 
> (Note this was a build without debuginfo, so I did not run decode_stacktrace.sh)
> 
> [   19.690210] ======================================================
> [   19.690212] WARNING: possible circular locking dependency detected
> [   19.690214] 6.16.0-rc2+ #3 Tainted: G            E      
> [   19.690217] ------------------------------------------------------
> [   19.690218] modprobe/906 is trying to acquire lock:
> [   19.690220] ffff89da8e948768 (&rdev->wiphy.mtx){+.+.}-{4:4}, at: iwl_mld_tzone_get_temp+0x2f/0x1d0 [iwlmld]
> [   19.690269] 
>                but task is already holding lock:
> [   19.690270] ffff89da41ac2708 (&tz->lock){+.+.}-{4:4}, at: thermal_zone_device_set_mode+0x20/0xa0
> [   19.690284] 
>                which lock already depends on the new lock.
> [SNIP]

I am thinking that this is a false-positive and the

    lock(cpuhp_state-up);
    lock(&tz->lock);

chain is happening on a different type of thermal device. A solution
might be to create a separate lock class for every thermal_zone_device.
Something like the below, I suspect.

Benjamin

diff --git a/drivers/thermal/thermal_core.c
b/drivers/thermal/thermal_core.c
index 17ca5c082643..fcb060898733 100644
--- a/drivers/thermal/thermal_core.c
+++ b/drivers/thermal/thermal_core.c
@@ -1556,6 +1556,7 @@ thermal_zone_device_register_with_trips(const
char *type,
 	INIT_LIST_HEAD(&tz->trips_invalid);
 	ida_init(&tz->ida);
 	mutex_init(&tz->lock);
+	lockdep_set_class(&tz->lock, &tz->lock_class);
 	init_completion(&tz->removal);
 	init_completion(&tz->resume);
 	id = ida_alloc(&thermal_tz_ida, GFP_KERNEL);
diff --git a/drivers/thermal/thermal_core.h
b/drivers/thermal/thermal_core.h
index bdadd141aa24..9ad5c37620f3 100644
--- a/drivers/thermal/thermal_core.h
+++ b/drivers/thermal/thermal_core.h
@@ -141,6 +141,7 @@ struct thermal_zone_device {
 	void *governor_data;
 	struct ida ida;
 	struct mutex lock;
+	struct lock_class_key lock_class;
 	struct list_head node;
 	struct delayed_work poll_queue;
 	enum thermal_notify_event notify_event;