[PATCH] usb: musb: Fix runtime PM race in musb_queue_resume_work

[PATCH] usb: musb: Fix runtime PM race in musb_queue_resume_work

[Paul Cercueil]

musb_queue_resume_work() would call the provided callback if the runtime
PM status was 'active'. Otherwise, it would enqueue the request if the
hardware was still suspended (musb->is_runtime_suspended is true).

This causes a race with the runtime PM handlers, as it is possible to be
in the case where the runtime PM status is not yet 'active', but the
hardware has been awaken (PM resume function has been called).

When hitting the race, the resume work was not enqueued, which probably
triggered other bugs further down the stack. For instance, a telnet
connection on Ingenic SoCs would result in a 50/50 chance of a
segmentation fault somewhere in the musb code.

Rework the code so that either we call the callback directly if
(musb->is_runtime_suspended == 0), or enqueue the query otherwise.

Fixes: ea2f35c01d5e ("usb: musb: Fix sleeping function called from invalid context for hdrc glue")
Cc: stable@vger.kernel.org # v4.9
Signed-off-by: Paul Cercueil <paul@crapouillou.net>
---
 drivers/usb/musb/musb_core.c | 31 +++++++++++++++++--------------
 1 file changed, 17 insertions(+), 14 deletions(-)

diff --git a/drivers/usb/musb/musb_core.c b/drivers/usb/musb/musb_core.c
index 384a8039a7fd..462c10d7455a 100644
--- a/drivers/usb/musb/musb_core.c
+++ b/drivers/usb/musb/musb_core.c
@@ -2241,32 +2241,35 @@ int musb_queue_resume_work(struct musb *musb,
 {
 	struct musb_pending_work *w;
 	unsigned long flags;
+	bool is_suspended;
 	int error;
 
 	if (WARN_ON(!callback))
 		return -EINVAL;
 
-	if (pm_runtime_active(musb->controller))
-		return callback(musb, data);
+	spin_lock_irqsave(&musb->list_lock, flags);
+	is_suspended = musb->is_runtime_suspended;
+
+	if (is_suspended) {
+		w = devm_kzalloc(musb->controller, sizeof(*w), GFP_ATOMIC);
+		if (!w) {
+			error = -ENOMEM;
+			goto out_unlock;
+		}
 
-	w = devm_kzalloc(musb->controller, sizeof(*w), GFP_ATOMIC);
-	if (!w)
-		return -ENOMEM;
+		w->callback = callback;
+		w->data = data;
 
-	w->callback = callback;
-	w->data = data;
-	spin_lock_irqsave(&musb->list_lock, flags);
-	if (musb->is_runtime_suspended) {
 		list_add_tail(&w->node, &musb->pending_list);
 		error = 0;
-	} else {
-		dev_err(musb->controller, "could not add resume work %p\n",
-			callback);
-		devm_kfree(musb->controller, w);
-		error = -EINPROGRESS;
 	}
+
+out_unlock:
 	spin_unlock_irqrestore(&musb->list_lock, flags);
 
+	if (!is_suspended)
+		error = callback(musb, data);
+
 	return error;
 }
 EXPORT_SYMBOL_GPL(musb_queue_resume_work);
-- 
2.28.0

Re: [PATCH] usb: musb: Fix runtime PM race in musb_queue_resume_work

[Tony Lindgren]

* Paul Cercueil <paul@crapouillou.net> [200809 12:54]:
> musb_queue_resume_work() would call the provided callback if the runtime
> PM status was 'active'. Otherwise, it would enqueue the request if the
> hardware was still suspended (musb->is_runtime_suspended is true).
> 
> This causes a race with the runtime PM handlers, as it is possible to be
> in the case where the runtime PM status is not yet 'active', but the
> hardware has been awaken (PM resume function has been called).
> 
> When hitting the race, the resume work was not enqueued, which probably
> triggered other bugs further down the stack. For instance, a telnet
> connection on Ingenic SoCs would result in a 50/50 chance of a
> segmentation fault somewhere in the musb code.
> 
> Rework the code so that either we call the callback directly if
> (musb->is_runtime_suspended == 0), or enqueue the query otherwise.

Yes we should use is_runtime_suspended, thanks for fixing it.
Things still work for me so:

Reviewed-by: Tony Lindgren <tony@atomide.com>
Tested-by: Tony Lindgren <tony@atomide.com>

Re: [PATCH] usb: musb: Fix runtime PM race in musb_queue_resume_work

[Paul Cercueil]

Hi,

Le lun. 17 août 2020 à 13:59, Tony Lindgren <tony@atomide.com> a 
écrit :
> * Paul Cercueil <paul@crapouillou.net> [200809 12:54]:
>>  musb_queue_resume_work() would call the provided callback if the 
>> runtime
>>  PM status was 'active'. Otherwise, it would enqueue the request if 
>> the
>>  hardware was still suspended (musb->is_runtime_suspended is true).
>> 
>>  This causes a race with the runtime PM handlers, as it is possible 
>> to be
>>  in the case where the runtime PM status is not yet 'active', but the
>>  hardware has been awaken (PM resume function has been called).
>> 
>>  When hitting the race, the resume work was not enqueued, which 
>> probably
>>  triggered other bugs further down the stack. For instance, a telnet
>>  connection on Ingenic SoCs would result in a 50/50 chance of a
>>  segmentation fault somewhere in the musb code.
>> 
>>  Rework the code so that either we call the callback directly if
>>  (musb->is_runtime_suspended == 0), or enqueue the query otherwise.
> 
> Yes we should use is_runtime_suspended, thanks for fixing it.
> Things still work for me so:
> 
> Reviewed-by: Tony Lindgren <tony@atomide.com>
> Tested-by: Tony Lindgren <tony@atomide.com>

Bin, can you take this patch?

Thanks,
-Paul