optee: ffa: fix sleep in atomic context

The OP-TEE driver registers the function notif_callback() for FF-A
notifications. However, this function is called in an atomic context
leading to errors like this when processing asynchronous notifications:

 | BUG: sleeping function called from invalid context at kernel/locking/mutex.c:258
 | in_atomic(): 1, irqs_disabled(): 1, non_block: 0, pid: 9, name: kworker/0:0
 | preempt_count: 1, expected: 0
 | RCU nest depth: 0, expected: 0
 | CPU: 0 UID: 0 PID: 9 Comm: kworker/0:0 Not tainted 6.14.0-00019-g657536ebe0aa #13
 | Hardware name: linux,dummy-virt (DT)
 | Workqueue: ffa_pcpu_irq_notification notif_pcpu_irq_work_fn
 | Call trace:
 |  show_stack+0x18/0x24 (C)
 |  dump_stack_lvl+0x78/0x90
 |  dump_stack+0x18/0x24
 |  __might_resched+0x114/0x170
 |  __might_sleep+0x48/0x98
 |  mutex_lock+0x24/0x80
 |  optee_get_msg_arg+0x7c/0x21c
 |  simple_call_with_arg+0x50/0xc0
 |  optee_do_bottom_half+0x14/0x20
 |  notif_callback+0x3c/0x48
 |  handle_notif_callbacks+0x9c/0xe0
 |  notif_get_and_handle+0x40/0x88
 |  generic_exec_single+0x80/0xc0
 |  smp_call_function_single+0xfc/0x1a0
 |  notif_pcpu_irq_work_fn+0x2c/0x38
 |  process_one_work+0x14c/0x2b4
 |  worker_thread+0x2e4/0x3e0
 |  kthread+0x13c/0x210
 |  ret_from_fork+0x10/0x20

Fix this by adding work queue to process the notification in a
non-atomic context.

Fixes: d0476a59de06 ("optee: ffa_abi: add asynchronous notifications")
Cc: stable@vger.kernel.org
Reviewed-by: Sumit Garg <sumit.garg@oss.qualcomm.com>
Tested-by: Sudeep Holla <sudeep.holla@arm.com>
Link: https://lore.kernel.org/r/20250602120452.2507084-1-jens.wiklander@linaro.org
Signed-off-by: Jens Wiklander <jens.wiklander@linaro.org>

2 files changed, 34 insertions, 9 deletions

diff --git a/drivers/tee/optee/ffa_abi.c b/drivers/tee/optee/ffa_abi.c
index f3af5666bb11..f9ef7d94cebd 100644
--- a/drivers/tee/optee/ffa_abi.c
+++ b/drivers/tee/optee/ffa_abi.c
@@ -728,12 +728,21 @@ static bool optee_ffa_exchange_caps(struct ffa_device *ffa_dev,
 	return true;
 }
 
+static void notif_work_fn(struct work_struct *work)
+{
+	struct optee_ffa *optee_ffa = container_of(work, struct optee_ffa,
+						   notif_work);
+	struct optee *optee = container_of(optee_ffa, struct optee, ffa);
+
+	optee_do_bottom_half(optee->ctx);
+}
+
 static void notif_callback(int notify_id, void *cb_data)
 {
 	struct optee *optee = cb_data;
 
 	if (notify_id == optee->ffa.bottom_half_value)
-		optee_do_bottom_half(optee->ctx);
+		queue_work(optee->ffa.notif_wq, &optee->ffa.notif_work);
 	else
 		optee_notif_send(optee, notify_id);
 }
@@ -817,9 +826,11 @@ static void optee_ffa_remove(struct ffa_device *ffa_dev)
 	struct optee *optee = ffa_dev_get_drvdata(ffa_dev);
 	u32 bottom_half_id = optee->ffa.bottom_half_value;
 
-	if (bottom_half_id != U32_MAX)
+	if (bottom_half_id != U32_MAX) {
 		ffa_dev->ops->notifier_ops->notify_relinquish(ffa_dev,
 							      bottom_half_id);
+		destroy_workqueue(optee->ffa.notif_wq);
+	}
 	optee_remove_common(optee);
 
 	mutex_destroy(&optee->ffa.mutex);
@@ -835,6 +846,13 @@ static int optee_ffa_async_notif_init(struct ffa_device *ffa_dev,
 	u32 notif_id = 0;
 	int rc;
 
+	INIT_WORK(&optee->ffa.notif_work, notif_work_fn);
+	optee->ffa.notif_wq = create_workqueue("optee_notification");
+	if (!optee->ffa.notif_wq) {
+		rc = -EINVAL;
+		goto err;
+	}
+
 	while (true) {
 		rc = ffa_dev->ops->notifier_ops->notify_request(ffa_dev,
 								is_per_vcpu,
@@ -851,19 +869,24 @@ static int optee_ffa_async_notif_init(struct ffa_device *ffa_dev,
 		 * notifications in that case.
 		 */
 		if (rc != -EACCES)
-			return rc;
+			goto err_wq;
 		notif_id++;
 		if (notif_id >= OPTEE_FFA_MAX_ASYNC_NOTIF_VALUE)
-			return rc;
+			goto err_wq;
 	}
 	optee->ffa.bottom_half_value = notif_id;
 
 	rc = enable_async_notif(optee);
-	if (rc < 0) {
-		ffa_dev->ops->notifier_ops->notify_relinquish(ffa_dev,
-							      notif_id);
-		optee->ffa.bottom_half_value = U32_MAX;
-	}
+	if (rc < 0)
+		goto err_rel;
+
+	return 0;
+err_rel:
+	ffa_dev->ops->notifier_ops->notify_relinquish(ffa_dev, notif_id);
+err_wq:
+	destroy_workqueue(optee->ffa.notif_wq);
+err:
+	optee->ffa.bottom_half_value = U32_MAX;
 
 	return rc;
 }
diff --git a/drivers/tee/optee/optee_private.h b/drivers/tee/optee/optee_private.h
index dc0f355ef72a..9526087f0e68 100644
--- a/drivers/tee/optee/optee_private.h
+++ b/drivers/tee/optee/optee_private.h
@@ -165,6 +165,8 @@ struct optee_ffa {
 	/* Serializes access to @global_ids */
 	struct mutex mutex;
 	struct rhashtable global_ids;
+	struct workqueue_struct *notif_wq;
+	struct work_struct notif_work;
 };
 
 struct optee;