[PATCH v2] mm/kfence: add reboot notifier to disable KFENCE on shutdown

[PATCH v2] mm/kfence: add reboot notifier to disable KFENCE on shutdown

[Breno Leitao]

During system shutdown, KFENCE can cause IPI synchronization issues if
it remains active through the reboot process. To prevent this, register
a reboot notifier that disables KFENCE and cancels any pending timer
work early in the shutdown sequence.

This is only necessary when CONFIG_KFENCE_STATIC_KEYS is enabled, as
this configuration sends IPIs that can interfere with shutdown. Without
static keys, no IPIs are generated and KFENCE can safely remain active.

The notifier uses maximum priority (INT_MAX) to ensure KFENCE shuts
down before other subsystems that might still depend on stable memory
allocation behavior.

This fixes a late kexec CSD lockup[1] when kfence is trying to IPI a CPU
that is busy in a IRQ-disabled context printing characters to the
console.

Link: https://lore.kernel.org/all/sqwajvt7utnt463tzxgwu2yctyn5m6bjwrslsnupfexeml6hkd@v6sqmpbu3vvu/ [1]

Cc: stable@vger.kernel.org
Signed-off-by: Breno Leitao <leitao@debian.org>
Reviewed-by: Marco Elver <elver@google.com>
Fixes: 0ce20dd84089 ("mm: add Kernel Electric-Fence infrastructure")
---
Changes in v2:
- Adding Fixes: tag and CCing stable (akpm)
- Link to v1: https://patch.msgid.link/20251126-kfence-v1-1-5a6e1d7c681c@debian.org
---
 mm/kfence/core.c | 24 ++++++++++++++++++++++++
 1 file changed, 24 insertions(+)

diff --git a/mm/kfence/core.c b/mm/kfence/core.c
index 727c20c94ac5..162a026871ab 100644
--- a/mm/kfence/core.c
+++ b/mm/kfence/core.c
@@ -26,6 +26,7 @@
 #include <linux/panic_notifier.h>
 #include <linux/random.h>
 #include <linux/rcupdate.h>
+#include <linux/reboot.h>
 #include <linux/sched/clock.h>
 #include <linux/seq_file.h>
 #include <linux/slab.h>
@@ -820,6 +821,25 @@ static struct notifier_block kfence_check_canary_notifier = {
 static struct delayed_work kfence_timer;
 
 #ifdef CONFIG_KFENCE_STATIC_KEYS
+static int kfence_reboot_callback(struct notifier_block *nb,
+				  unsigned long action, void *data)
+{
+	/*
+	 * Disable kfence to avoid static keys IPI synchronization during
+	 * late shutdown/kexec
+	 */
+	WRITE_ONCE(kfence_enabled, false);
+	/* Cancel any pending timer work */
+	cancel_delayed_work_sync(&kfence_timer);
+
+	return NOTIFY_OK;
+}
+
+static struct notifier_block kfence_reboot_notifier = {
+	.notifier_call = kfence_reboot_callback,
+	.priority = INT_MAX, /* Run early to stop timers ASAP */
+};
+
 /* Wait queue to wake up allocation-gate timer task. */
 static DECLARE_WAIT_QUEUE_HEAD(allocation_wait);
 
@@ -901,6 +921,10 @@ static void kfence_init_enable(void)
 	if (kfence_check_on_panic)
 		atomic_notifier_chain_register(&panic_notifier_list, &kfence_check_canary_notifier);
 
+#ifdef CONFIG_KFENCE_STATIC_KEYS
+	register_reboot_notifier(&kfence_reboot_notifier);
+#endif
+
 	WRITE_ONCE(kfence_enabled, true);
 	queue_delayed_work(system_unbound_wq, &kfence_timer, 0);
 

---
base-commit: ab084f0b8d6d2ee4b1c6a28f39a2a7430bdfa7f0
change-id: 20251126-kfence-42c93f9b3979

Best regards,
--  
Breno Leitao <leitao@debian.org>

Re: [PATCH v2] mm/kfence: add reboot notifier to disable KFENCE on shutdown

[Chris Mason]

On Thu, 27 Nov 2025 06:51:54 -0800 Breno Leitao <leitao@debian.org> wrote:

> During system shutdown, KFENCE can cause IPI synchronization issues if
> it remains active through the reboot process. To prevent this, register
> a reboot notifier that disables KFENCE and cancels any pending timer
> work early in the shutdown sequence.
>

Hi Breno,

I'm running v6.19 MM commits through the patch review automation and this
one was flagged.  Is this deadlock possible?

Automated review output:

> @@ -820,6 +821,25 @@ static struct notifier_block kfence_check_canary_notifier = {
>  static struct delayed_work kfence_timer;
>
>  #ifdef CONFIG_KFENCE_STATIC_KEYS
> +static int kfence_reboot_callback(struct notifier_block *nb,
> +				  unsigned long action, void *data)
> +{
> +	/*
> +	 * Disable kfence to avoid static keys IPI synchronization during
> +	 * late shutdown/kexec
> +	 */
> +	WRITE_ONCE(kfence_enabled, false);
> +	/* Cancel any pending timer work */
> +	cancel_delayed_work_sync(&kfence_timer);
                   ^^^^^^^^^^^^^^^

Can cancel_delayed_work_sync() deadlock here?

If toggle_allocation_gate() is currently executing and blocked inside
wait_event_idle() (waiting for kfence_allocation_gate > 0), then
cancel_delayed_work_sync() will block forever waiting for the work to
complete.

The wait_event_idle() condition depends only on allocations occurring
to increment kfence_allocation_gate - setting kfence_enabled to false
does not wake up this wait. During shutdown when allocations may have
stopped, the work item could remain blocked indefinitely, causing the
reboot notifier to hang.

The call chain is:
  kfence_reboot_callback()
    -> cancel_delayed_work_sync(&kfence_timer)
       -> __flush_work()
          -> wait_for_completion(&barr.done)
             // waits forever because...

  toggle_allocation_gate() [currently running]
    -> wait_event_idle(allocation_wait, kfence_allocation_gate > 0)
       // never wakes up if no allocations happen

Would it be safer to use cancel_delayed_work() (non-sync) here, or add
a mechanism to wake up the wait_event_idle() when kfence_enabled becomes
false?