[PATCH v4] hwrng: core - Stop/start hwrng_fillfn() kthread before/after suspend-resume

[PATCH v4] hwrng: core - Stop/start hwrng_fillfn() kthread before/after suspend-resume

[Thomas Richard (TI)]

The hwrng_fillfn() kernel thread accesses the RNG device directly. During
suspend and resume sequences, hwrng_fillfn() may attempt to access the RNG
device while it is suspended. To address this, the hwrng_fillfn() kernel
thread is stopped before suspend, and restarted after resume. This is done
using the pm_notifier mechanism.

Issue was found while doing suspend-to-ram on J721S2 EVM board with
omap-rng driver.

echo mem > /sys/power/state
[   27.922259] PM: suspend entry (deep)
[   27.927191] Filesystems sync: 0.000 seconds
[   27.933858] Freezing user space processes
[   27.939119] Freezing user space processes completed (elapsed 0.001 seconds)
[   27.946090] OOM killer disabled.
[   27.949315] Freezing remaining freezable tasks
[   27.954887] Freezing remaining freezable tasks completed (elapsed 0.001 seconds)
[   27.963337] GFP mask restricted
[   27.967069] omap_rng 4e10000.rng: PM: calling platform_pm_suspend @ 195, parent: 4e00000.crypto
[   27.967072] mmcblk mmc1:9fb0: PM: calling mmc_bus_suspend @ 122, parent: mmc1
[   27.968636] mmcblk mmc1:9fb0: PM: mmc_bus_suspend returned 0 after 1546 usecs
[   27.975778] omap_rng 4e10000.rng: PM: platform_pm_suspend returned 0 after 3 usecs
...
[   33.510667] ti-sci 44083000.system-controller: PM: ti_sci_suspend_noirq returned 0 after 0 usecs
[   33.510671] SError Interrupt on CPU0, code 0x00000000bf000000 -- SError
[   33.510681] CPU: 0 UID: 0 PID: 132 Comm: hwrng Tainted: G   M    W           7.0.0-12695-g8923b7a6e11d #19 PREEMPT
[   33.510690] Tainted: [M]=MACHINE_CHECK, [W]=WARN
[   33.510693] Hardware name: Texas Instruments J721S2 EVM (DT)
[   33.510697] pstate: 60000005 (nZCv daif -PAN -UAO -TCO -DIT -SSBS BTYPE=--)
[   33.510701] pc : omap_rng_do_read+0x3c/0xe0
[   33.510709] lr : omap_rng_do_read+0x58/0xe0
[   33.510712] sp : ffff80008942be00
[   33.510713] x29: ffff80008942be00 x28: 0000000000000000 x27: 0000000000000000
[   33.510719] x26: 0000000000000010 x25: 0000000000000010 x24: ffff0008065644e8
[   33.510724] x23: ffff8000878b3370 x22: ffff00080148b2c0 x21: 0000000000000000
[   33.510728] x20: ffff000806564480 x19: 0000000000000064 x18: 0000000000000000
[   33.510732] x17: 6573752031207265 x16: 7466612030206465 x15: 6e72757465722071
[   33.510737] x14: ffff0008062c8080 x13: 000031702bc0da42 x12: 0000000000000001
[   33.510741] x11: 00000000000000c0 x10: 0000000000000b30 x9 : ffff80008942bc80
[   33.510745] x8 : ffff0008062c8b90 x7 : ffff000b7dfa34c0 x6 : 0000000805ca16c1
[   33.510749] x5 : 0000000000000000 x4 : ffff800080e17bfc x3 : ffff800087389c68
[   33.510753] x2 : 0000000000000000 x1 : 0000000000000010 x0 : 000000000000a7c6
[   33.510759] Kernel panic - not syncing: Asynchronous SError Interrupt
[   33.510762] CPU: 0 UID: 0 PID: 132 Comm: hwrng Tainted: G   M    W           7.0.0-12695-g8923b7a6e11d #19 PREEMPT
[   33.510767] Tainted: [M]=MACHINE_CHECK, [W]=WARN
[   33.510768] Hardware name: Texas Instruments J721S2 EVM (DT)
[   33.510770] Call trace:
[   33.510772]  show_stack+0x18/0x24 (C)
[   33.510780]  dump_stack_lvl+0x34/0x8c
[   33.510788]  dump_stack+0x18/0x24
[   33.510792]  vpanic+0x47c/0x4dc
[   33.510799]  do_panic_on_target_cpu+0x0/0x1c
[   33.510803]  add_taint+0x0/0xbc
[   33.510807]  arm64_serror_panic+0x70/0x80
[   33.510812]  do_serror+0x3c/0x70
[   33.510815]  el1h_64_error_handler+0x34/0x50
[   33.510823]  el1h_64_error+0x6c/0x70
[   33.510827]  omap_rng_do_read+0x3c/0xe0 (P)
[   33.510831]  hwrng_fillfn+0x98/0x330
[   33.510834]  kthread+0x130/0x13c
[   33.510845]  ret_from_fork+0x10/0x20
[   33.510850] SMP: stopping secondary CPUs
[   33.519442] Kernel Offset: disabled
[   33.519444] CPU features: 0x04000000,800a0008,00040001,0400421b
[   33.519448] Memory Limit: none
[   33.732904] ---[ end Kernel panic - not syncing: Asynchronous SError Interrupt ]---

Signed-off-by: Thomas Richard (TI) <thomas.richard@bootlin.com>
---
This is the fourth iteration, as requested by Herbert I replaced BUG_ON()
by lockdep_assert_held().
---
Changes in v4:
- use lockdep_assert_held() instead of BUG_ON()
- Link to v3: https://lore.kernel.org/r/20260703-hw-random-fix-hwrng-fillfn-crash-suspend-resume-v3-1-b7165cb9cf38@bootlin.com

Changes in v3:
- use pm_notifier to stop/start the hwrng_fillfn() kthread before/after
  suspend-resume.
- Link to v2: https://lore.kernel.org/r/20260601-hw-random-fix-hwrng-fillfn-crash-suspend-resume-v2-1-667ce5da32ee@bootlin.com

Changes in v2:
- Initialize rc variable.
- Link to v1: https://lore.kernel.org/r/20260513-hw-random-fix-hwrng-fillfn-crash-suspend-resume-v1-1-b12551c1c7dd@bootlin.com
---
 drivers/char/hw_random/core.c | 90 +++++++++++++++++++++++++++++++++++--------
 1 file changed, 75 insertions(+), 15 deletions(-)

diff --git a/drivers/char/hw_random/core.c b/drivers/char/hw_random/core.c
index 6931657ad2ca..82cc8f825494 100644
--- a/drivers/char/hw_random/core.c
+++ b/drivers/char/hw_random/core.c
@@ -17,14 +17,17 @@
 #include <linux/hw_random.h>
 #include <linux/kernel.h>
 #include <linux/kthread.h>
+#include <linux/lockdep.h>
 #include <linux/miscdevice.h>
 #include <linux/module.h>
+#include <linux/notifier.h>
 #include <linux/random.h>
 #include <linux/rcupdate.h>
 #include <linux/sched.h>
 #include <linux/sched/signal.h>
 #include <linux/slab.h>
 #include <linux/string.h>
+#include <linux/suspend.h>
 #include <linux/sysfs.h>
 #include <linux/uaccess.h>
 #include <linux/workqueue.h>
@@ -57,6 +60,8 @@ MODULE_PARM_DESC(default_quality,
 
 static int hwrng_init(struct hwrng *rng);
 static int hwrng_fillfn(void *unused);
+static void hwrng_start_hwrng_fillfn(void);
+static void hwrng_stop_hwrng_fillfn(void);
 
 static size_t rng_buffer_size(void)
 {
@@ -114,13 +119,7 @@ static int set_current_rng(struct hwrng *rng)
 	}
 
 	/* if necessary, start hwrng thread */
-	if (!hwrng_fill) {
-		hwrng_fill = kthread_run(hwrng_fillfn, NULL, "hwrng");
-		if (IS_ERR(hwrng_fill)) {
-			pr_err("hwrng_fill thread creation failed\n");
-			hwrng_fill = NULL;
-		}
-	}
+	hwrng_start_hwrng_fillfn();
 
 	return 0;
 }
@@ -137,10 +136,7 @@ static void drop_current_rng(void)
 	RCU_INIT_POINTER(current_rng, NULL);
 	synchronize_rcu();
 
-	if (hwrng_fill) {
-		kthread_stop(hwrng_fill);
-		hwrng_fill = NULL;
-	}
+	hwrng_stop_hwrng_fillfn();
 
 	/* decrease last reference for triggering the cleanup */
 	kref_put(&rng->ref, cleanup_rng);
@@ -506,6 +502,29 @@ static struct attribute *rng_dev_attrs[] = {
 
 ATTRIBUTE_GROUPS(rng_dev);
 
+static void hwrng_start_hwrng_fillfn(void)
+{
+	lockdep_assert_held(&rng_mutex);
+
+	if (!hwrng_fill) {
+		hwrng_fill = kthread_run(hwrng_fillfn, NULL, "hwrng");
+		if (IS_ERR(hwrng_fill)) {
+			pr_err("hwrng_fill thread creation failed\n");
+			hwrng_fill = NULL;
+		}
+	}
+}
+
+static void hwrng_stop_hwrng_fillfn(void)
+{
+	lockdep_assert_held(&rng_mutex);
+
+	if (hwrng_fill) {
+		kthread_stop(hwrng_fill);
+		hwrng_fill = NULL;
+	}
+}
+
 static int hwrng_fillfn(void *unused)
 {
 	size_t entropy, entropy_credit = 0; /* in 1/1024 of a bit */
@@ -559,6 +578,35 @@ static int hwrng_fillfn(void *unused)
 	return 0;
 }
 
+static int hwrng_pm_notifier(struct notifier_block *nb, unsigned long action,
+			     void *data)
+{
+	switch (action) {
+	case PM_SUSPEND_PREPARE:
+	case PM_HIBERNATION_PREPARE:
+	case PM_RESTORE_PREPARE:
+		mutex_lock(&rng_mutex);
+		hwrng_stop_hwrng_fillfn();
+		mutex_unlock(&rng_mutex);
+		break;
+
+	case PM_POST_SUSPEND:
+	case PM_POST_HIBERNATION:
+	case PM_POST_RESTORE:
+		mutex_lock(&rng_mutex);
+		if (rcu_access_pointer(current_rng))
+			hwrng_start_hwrng_fillfn();
+		mutex_unlock(&rng_mutex);
+		break;
+	}
+
+	return NOTIFY_DONE;
+}
+
+static struct notifier_block hwrng_pm_nb = {
+	.notifier_call = hwrng_pm_notifier,
+};
+
 int hwrng_register(struct hwrng *rng)
 {
 	int err = -EINVAL;
@@ -705,10 +753,20 @@ static int __init hwrng_modinit(void)
 	}
 
 	ret = misc_register(&rng_miscdev);
-	if (ret) {
-		kfree(rng_fillbuf);
-		kfree(rng_buffer);
-	}
+	if (ret)
+		goto misc_err;
+
+	ret = register_pm_notifier(&hwrng_pm_nb);
+	if (ret)
+		goto pm_err;
+
+	return 0;
+
+pm_err:
+	misc_deregister(&rng_miscdev);
+misc_err:
+	kfree(rng_fillbuf);
+	kfree(rng_buffer);
 
 	return ret;
 }
@@ -721,6 +779,8 @@ static void __exit hwrng_modexit(void)
 	kfree(rng_fillbuf);
 	mutex_unlock(&rng_mutex);
 
+	unregister_pm_notifier(&hwrng_pm_nb);
+
 	misc_deregister(&rng_miscdev);
 }
 

---
base-commit: f4b80d91193e2453f7569c0249e59d6f2805d248
change-id: 20260513-hw-random-fix-hwrng-fillfn-crash-suspend-resume-29fdb0638f59

Best regards,
-- 
Thomas Richard (TI) <thomas.richard@bootlin.com>

Re: [PATCH v4] hwrng: core - Stop/start hwrng_fillfn() kthread before/after suspend-resume

[Herbert Xu]

On Mon, Jul 13, 2026 at 01:25:37PM +0200, Thomas Richard (TI) wrote:
>
> This is the fourth iteration, as requested by Herbert I replaced BUG_ON()
> by lockdep_assert_held().
> ---
> Changes in v4:
> - use lockdep_assert_held() instead of BUG_ON()
> - Link to v3: https://lore.kernel.org/r/20260703-hw-random-fix-hwrng-fillfn-crash-suspend-resume-v3-1-b7165cb9cf38@bootlin.com

https://sashiko.dev/#/patchset/20260713-hw-random-fix-hwrng-fillfn-crash-suspend-resume-v4-1-b5fb4c520b05%40bootlin.com

Sashiko is asking about whether user-space can restart the hwrng
thread via sysfs after you stop it in the notifier.  I know nothing
about suspension so I can't say whether that's possible.  But if
it is possible then you'd need to close the race window with a flag
of some kind.

Thanks,
-- 
Email: Herbert Xu <herbert@gondor.apana.org.au>
Home Page: http://gondor.apana.org.au/~herbert/
PGP Key: http://gondor.apana.org.au/~herbert/pubkey.txt

Re: [PATCH v4] hwrng: core - Stop/start hwrng_fillfn() kthread before/after suspend-resume

[Thomas Richard]

On 7/27/26 1:55 AM, Herbert Xu wrote:
> On Mon, Jul 13, 2026 at 01:25:37PM +0200, Thomas Richard (TI) wrote:
>>
>> This is the fourth iteration, as requested by Herbert I replaced BUG_ON()
>> by lockdep_assert_held().
>> ---
>> Changes in v4:
>> - use lockdep_assert_held() instead of BUG_ON()
>> - Link to v3: https://lore.kernel.org/r/20260703-hw-random-fix-hwrng-fillfn-crash-suspend-resume-v3-1-b7165cb9cf38@bootlin.com
> 
> https://sashiko.dev/#/patchset/20260713-hw-random-fix-hwrng-fillfn-crash-suspend-resume-v4-1-b5fb4c520b05%40bootlin.com
> 
> Sashiko is asking about whether user-space can restart the hwrng
> thread via sysfs after you stop it in the notifier.  I know nothing
> about suspension so I can't say whether that's possible.  But if
> it is possible then you'd need to close the race window with a flag
> of some kind.

Oh yes, indeed Sashiko is right. PM notifiers are run before user space
processes freezing.

What do you think if I lock the rng_mutex at the very beginning of
suspend and release it at the end of resume?

static int hwrng_pm_notifier(struct notifier_block *nb,
			     unsigned long action,
			     void *data)
{
	switch (action) {
	case PM_SUSPEND_PREPARE:
	case PM_HIBERNATION_PREPARE:
	case PM_RESTORE_PREPARE:
		mutex_lock(&rng_mutex);
		hwrng_stop_hwrng_fillfn();
		break;

	case PM_POST_SUSPEND:
	case PM_POST_HIBERNATION:
	case PM_POST_RESTORE:
		if (rcu_access_pointer(current_rng))
			hwrng_start_hwrng_fillfn();
		mutex_unlock(&rng_mutex);
		break;
	}

	return NOTIFY_DONE;
}

IIUC once rng_mutex is locked nobody can start the hwrng_thread.

And even if suspend sequence fails somewhere, POST PM notifier will be
run, so mutex will be released in any cases.

Best Regards,
Thomas

Re: [PATCH v4] hwrng: core - Stop/start hwrng_fillfn() kthread before/after suspend-resume

[Herbert Xu]

On Mon, Jul 27, 2026 at 11:30:41AM +0200, Thomas Richard wrote:
>
> Oh yes, indeed Sashiko is right. PM notifiers are run before user space
> processes freezing.
> 
> What do you think if I lock the rng_mutex at the very beginning of
> suspend and release it at the end of resume?

Wouldn't this just cause a hang if said race occurred during
suspension?

Ideally, we want the sysfs store to either result in an error
due to the ongoing suspension, or at least fail silently without
affecting the suspension.

Thanks,
-- 
Email: Herbert Xu <herbert@gondor.apana.org.au>
Home Page: http://gondor.apana.org.au/~herbert/
PGP Key: http://gondor.apana.org.au/~herbert/pubkey.txt