Linux-ARM-Kernel Archive on lore.kernel.org
 help / color / mirror / Atom feed
From: Sudeep Holla <sudeep.holla@kernel.org>
To: arm-scmi@vger.kernel.org, linux-arm-kernel@lists.infradead.org
Cc: Sudeep Holla <sudeep.holla@kernel.org>,
	Cristian Marussi <cristian.marussi@arm.com>,
	Sashiko <sashiko-bot@kernel.org>
Subject: [PATCH v4 02/16][UPDATED] firmware: arm_scmi: Quiesce notifications before teardown
Date: Wed,  8 Jul 2026 21:55:22 +0100	[thread overview]
Message-ID: <20260708205522.139187-1-sudeep.holla@kernel.org> (raw)
In-Reply-To: <20260708-scmi_core_fixes-v4-2-53142cd60b63@kernel.org>

scmi_notification_exit() clears and releases the notification instance, but
transport callbacks can still deliver incoming notifications until the
TX/RX channels are freed. During remove, an RX interrupt in that window can
enter scmi_notify() while notification state is being torn down and then
dereference freed memory. The same ordering exists on the probe error path
after notification initialization.

The notification late-init worker has a separate lifetime issue: protocol
event registration queues ni->init_work on the system workqueue, so
destroying ni->notify_wq does not drain that work. If the devres group is
released while init_work is still pending or running, the late-init worker
can dereference the freed notification instance.

Unregister the requested device notifier first, quiesce the notification
core so no late-init work can run while channels are being torn down, then
clean up TX/RX channels before releasing the notification core resources.
The final notification exit path also cancels any pending late-init work
before destroying the notification workqueue and releasing the notification
devres group.

Fixes: 1e7cbfaa66d3 ("firmware: arm_scmi: Free mailbox channels if probe fails")
Reported-by: Sashiko <sashiko-bot@kernel.org>
Signed-off-by: Sudeep Holla <sudeep.holla@kernel.org>
---
 drivers/firmware/arm_scmi/driver.c | 13 ++++++------
 drivers/firmware/arm_scmi/notify.c | 32 +++++++++++++++++++++++++++++-
 drivers/firmware/arm_scmi/notify.h |  1 +
 3 files changed, 39 insertions(+), 7 deletions(-)

diff --git a/drivers/firmware/arm_scmi/driver.c b/drivers/firmware/arm_scmi/driver.c
index 1d1f5d25d773..1354de59ea84 100644
--- a/drivers/firmware/arm_scmi/driver.c
+++ b/drivers/firmware/arm_scmi/driver.c
@@ -3325,7 +3325,7 @@ static int scmi_probe(struct platform_device *pdev)
 			dev_err(dev, "%s", err_str);
 			return 0;
 		}
-		goto notification_exit;
+		goto raw_mode_cleanup;
 	}
 
 	mutex_lock(&scmi_list_mutex);
@@ -3367,17 +3367,18 @@ static int scmi_probe(struct platform_device *pdev)
 
 	return 0;
 
-notification_exit:
+raw_mode_cleanup:
 	if (IS_ENABLED(CONFIG_ARM_SCMI_RAW_MODE_SUPPORT))
 		scmi_raw_mode_cleanup(info->raw);
-	scmi_notification_exit(&info->handle);
 clear_dev_req_notifier:
 	blocking_notifier_chain_unregister(&scmi_requested_devices_nh,
 					   &info->dev_req_nb);
 clear_bus_notifier:
 	bus_unregister_notifier(&scmi_bus_type, &info->bus_nb);
 clear_txrx_setup:
+	scmi_notification_quiesce(&info->handle);
 	scmi_cleanup_txrx_channels(info);
+	scmi_notification_exit(&info->handle);
 clear_ida:
 	ida_free(&scmi_id, info->id);
 
@@ -3401,6 +3402,9 @@ static void scmi_remove(struct platform_device *pdev)
 	list_del(&info->node);
 	mutex_unlock(&scmi_list_mutex);
 
+	/* Stop transport callbacks before tearing down notifications. */
+	scmi_notification_quiesce(&info->handle);
+	scmi_cleanup_txrx_channels(info);
 	scmi_notification_exit(&info->handle);
 
 	mutex_lock(&info->protocols_mtx);
@@ -3415,9 +3419,6 @@ static void scmi_remove(struct platform_device *pdev)
 					   &info->dev_req_nb);
 	bus_unregister_notifier(&scmi_bus_type, &info->bus_nb);
 
-	/* Safe to free channels since no more users */
-	scmi_cleanup_txrx_channels(info);
-
 	ida_free(&scmi_id, info->id);
 }
 
diff --git a/drivers/firmware/arm_scmi/notify.c b/drivers/firmware/arm_scmi/notify.c
index 40ec184eedae..2491c848d4f3 100644
--- a/drivers/firmware/arm_scmi/notify.c
+++ b/drivers/firmware/arm_scmi/notify.c
@@ -209,6 +209,7 @@ struct scmi_registered_events_desc;
  * @init_work: A work item to perform final initializations of pending handlers
  * @notify_wq: A reference to the allocated Kernel cmwq
  * @pending_mtx: A mutex to protect @pending_events_handlers
+ * @shutting_down: Flag to block late-init work scheduling during teardown
  * @registered_protocols: A statically allocated array containing pointers to
  *			  all the registered protocol-level specific information
  *			  related to events' handling
@@ -225,6 +226,7 @@ struct scmi_notify_instance {
 	struct workqueue_struct	*notify_wq;
 	/* lock to protect pending_events_handlers */
 	struct mutex		pending_mtx;
+	bool			shutting_down;
 	struct scmi_registered_events_desc	**registered_protocols;
 	DECLARE_HASHTABLE(pending_events_handlers, SCMI_PENDING_HASH_SZ);
 };
@@ -845,7 +847,10 @@ int scmi_register_protocol_events(const struct scmi_handle *handle, u8 proto_id,
 	 * Finalize any pending events' handler which could have been waiting
 	 * for this protocol's events registration.
 	 */
-	schedule_work(&ni->init_work);
+	mutex_lock(&ni->pending_mtx);
+	if (!ni->shutting_down)
+		schedule_work(&ni->init_work);
+	mutex_unlock(&ni->pending_mtx);
 
 	return 0;
 }
@@ -1706,6 +1711,29 @@ int scmi_notification_init(struct scmi_handle *handle)
 	return -ENOMEM;
 }
 
+/**
+ * scmi_notification_quiesce()  - Stop notification late initialization
+ * @handle: The handle identifying the platform instance to quiesce
+ *
+ * Prevent new late-init work from being queued and wait for any already queued
+ * or running late-init work to complete before transport channels are torn
+ * down.
+ */
+void scmi_notification_quiesce(struct scmi_handle *handle)
+{
+	struct scmi_notify_instance *ni;
+
+	ni = scmi_notification_instance_data_get(handle);
+	if (!ni)
+		return;
+
+	mutex_lock(&ni->pending_mtx);
+	ni->shutting_down = true;
+	mutex_unlock(&ni->pending_mtx);
+
+	cancel_work_sync(&ni->init_work);
+}
+
 /**
  * scmi_notification_exit()  - Shutdown and clean Notification core
  * @handle: The handle identifying the platform instance to shutdown
@@ -1717,6 +1745,8 @@ void scmi_notification_exit(struct scmi_handle *handle)
 	ni = scmi_notification_instance_data_get(handle);
 	if (!ni)
 		return;
+
+	scmi_notification_quiesce(handle);
 	scmi_notification_instance_data_set(handle, NULL);
 
 	/* Destroy while letting pending work complete */
diff --git a/drivers/firmware/arm_scmi/notify.h b/drivers/firmware/arm_scmi/notify.h
index 76758a736cf4..f18f98c5ab3b 100644
--- a/drivers/firmware/arm_scmi/notify.h
+++ b/drivers/firmware/arm_scmi/notify.h
@@ -82,6 +82,7 @@ struct scmi_protocol_events {
 };
 
 int scmi_notification_init(struct scmi_handle *handle);
+void scmi_notification_quiesce(struct scmi_handle *handle);
 void scmi_notification_exit(struct scmi_handle *handle);
 int scmi_register_protocol_events(const struct scmi_handle *handle, u8 proto_id,
 				  const struct scmi_protocol_handle *ph,
-- 
2.43.0



  reply	other threads:[~2026-07-08 20:55 UTC|newest]

Thread overview: 18+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2026-07-08  8:59 [PATCH v4 00/16] firmware: arm_scmi: Fix SCMI core cleanup paths Sudeep Holla
2026-07-08  8:59 ` [PATCH v4 01/16] firmware: arm_scmi: Publish channel state before callbacks Sudeep Holla
2026-07-08  8:59 ` [PATCH v4 02/16] firmware: arm_scmi: Quiesce notifications before teardown Sudeep Holla
2026-07-08 20:55   ` Sudeep Holla [this message]
2026-07-08  8:59 ` [PATCH v4 03/16] firmware: arm_scmi: Clean up channels on setup failure Sudeep Holla
2026-07-08  9:00 ` [PATCH v4 04/16] firmware: arm_scmi: Free transport channel on IDR failure Sudeep Holla
2026-07-08  9:00 ` [PATCH v4 05/16] firmware: arm_scmi: Avoid IDR updates while cleaning channels Sudeep Holla
2026-07-08  9:00 ` [PATCH v4 06/16] firmware: arm_scmi: Reject out of range DT protocol IDs Sudeep Holla
2026-07-08  9:00 ` [PATCH v4 07/16] firmware: arm_scmi: Use channel ID for transport teardown Sudeep Holla
2026-07-08  9:00 ` [PATCH v4 08/16] firmware: arm_scmi: Unregister device notifier before IDR teardown Sudeep Holla
2026-07-08  9:00 ` [PATCH v4 09/16] firmware: arm_scmi: Protect device request lookup with RCU Sudeep Holla
2026-07-08  9:00 ` [PATCH v4 10/16] firmware: arm_scmi: Drop handle on protocol bind failures Sudeep Holla
2026-07-08  9:00 ` [PATCH v4 11/16] firmware: arm_scmi: Clear SystemPower flag on create failure Sudeep Holla
2026-07-08  9:00 ` [PATCH v4 12/16] firmware: arm_scmi: Fix OF node reference handling Sudeep Holla
2026-07-08  9:00 ` [PATCH v4 13/16] firmware: arm_scmi: Unwind TX receiver mailbox setup failure Sudeep Holla
2026-07-08  9:00 ` [PATCH v4 14/16] firmware: arm_scmi: Unwind P2A " Sudeep Holla
2026-07-08  9:00 ` [PATCH v4 15/16] firmware: arm_scmi: Fix SCMI device destroy lifetimes Sudeep Holla
2026-07-08  9:00 ` [PATCH v4 16/16] firmware: arm_scmi: Fix transport device teardown lookup Sudeep Holla

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=20260708205522.139187-1-sudeep.holla@kernel.org \
    --to=sudeep.holla@kernel.org \
    --cc=arm-scmi@vger.kernel.org \
    --cc=cristian.marussi@arm.com \
    --cc=linux-arm-kernel@lists.infradead.org \
    --cc=sashiko-bot@kernel.org \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox