[PATCH net v2] ptp: vclock: Switch from RCU to SRCU

[PATCH net v2] ptp: vclock: Switch from RCU to SRCU

[Kurt Kanzenbach]

The usage of PTP vClocks leads immediately to the following issues with
ptp4l with LOCKDEP and DEBUG_ATOMIC_SLEEP enabled: "BUG: sleeping function
called from invalid context".

ptp_convert_timestamp() acquires a mutex_t within a RCU read section.  This
is illegal, because acquiring a mutex_t can result in voluntary scheduling
request which is not allowed within a RCU read section.

Replace the RCU usage with SRCU where sleeping is allowed.

Reported-by: Florian Zeitz <florian.zeitz@schettke.com>
Closes: https://lore.kernel.org/all/00a8cce8-410e-4038-98af-49be6d93d7bd@schettke.com/
Fixes: 67d93ffc0f3c ("ptp: vclock: use mutex to fix "sleep on atomic" bug")
Signed-off-by: Kurt Kanzenbach <kurt@linutronix.de>
---
Changes in v2:
- Remove RFC. No comments.
- Link to v1: https://patch.msgid.link/20260521-vclock_rcu-v1-1-9c345c18bd72@linutronix.de
---
 drivers/ptp/ptp_vclock.c | 14 +++++++++-----
 1 file changed, 9 insertions(+), 5 deletions(-)

diff --git a/drivers/ptp/ptp_vclock.c b/drivers/ptp/ptp_vclock.c
index 915a4f6defc9..84cb527f59cc 100644
--- a/drivers/ptp/ptp_vclock.c
+++ b/drivers/ptp/ptp_vclock.c
@@ -19,6 +19,8 @@ static DEFINE_SPINLOCK(vclock_hash_lock);
 
 static DEFINE_READ_MOSTLY_HASHTABLE(vclock_hash, 8);
 
+DEFINE_STATIC_SRCU(vclock_srcu);
+
 static void ptp_vclock_hash_add(struct ptp_vclock *vclock)
 {
 	spin_lock(&vclock_hash_lock);
@@ -37,7 +39,7 @@ static void ptp_vclock_hash_del(struct ptp_vclock *vclock)
 
 	spin_unlock(&vclock_hash_lock);
 
-	synchronize_rcu();
+	synchronize_srcu(&vclock_srcu);
 }
 
 static int ptp_vclock_adjfine(struct ptp_clock_info *ptp, long scaled_ppm)
@@ -276,14 +278,16 @@ ktime_t ptp_convert_timestamp(const ktime_t *hwtstamp, int vclock_index)
 {
 	unsigned int hash = vclock_index % HASH_SIZE(vclock_hash);
 	struct ptp_vclock *vclock;
-	u64 ns;
 	u64 vclock_ns = 0;
+	int srcu_idx;
+	u64 ns;
 
 	ns = ktime_to_ns(*hwtstamp);
 
-	rcu_read_lock();
+	srcu_idx = srcu_read_lock(&vclock_srcu);
 
-	hlist_for_each_entry_rcu(vclock, &vclock_hash[hash], vclock_hash_node) {
+	hlist_for_each_entry_srcu(vclock, &vclock_hash[hash], vclock_hash_node,
+				  srcu_read_lock_held(&vclock_srcu)) {
 		if (vclock->clock->index != vclock_index)
 			continue;
 
@@ -294,7 +298,7 @@ ktime_t ptp_convert_timestamp(const ktime_t *hwtstamp, int vclock_index)
 		break;
 	}
 
-	rcu_read_unlock();
+	srcu_read_unlock(&vclock_srcu, srcu_idx);
 
 	return ns_to_ktime(vclock_ns);
 }

---
base-commit: 422b5233b607476ac7176bfa2a101b9a103d7653
change-id: 20260521-vclock_rcu-e29c9c107c3b

Best regards,
--  
Kurt Kanzenbach <kurt@linutronix.de>

Re: [PATCH net v2] ptp: vclock: Switch from RCU to SRCU

[Sebastian Andrzej Siewior]

On 2026-05-29 19:11:47 [+0200], Kurt Kanzenbach wrote:
> The usage of PTP vClocks leads immediately to the following issues with
> ptp4l with LOCKDEP and DEBUG_ATOMIC_SLEEP enabled: "BUG: sleeping function
> called from invalid context".
> 
> ptp_convert_timestamp() acquires a mutex_t within a RCU read section.  This
> is illegal, because acquiring a mutex_t can result in voluntary scheduling
> request which is not allowed within a RCU read section.
> 
> Replace the RCU usage with SRCU where sleeping is allowed.
> 
> Reported-by: Florian Zeitz <florian.zeitz@schettke.com>
> Closes: https://lore.kernel.org/all/00a8cce8-410e-4038-98af-49be6d93d7bd@schettke.com/
> Fixes: 67d93ffc0f3c ("ptp: vclock: use mutex to fix "sleep on atomic" bug")
> Signed-off-by: Kurt Kanzenbach <kurt@linutronix.de>

Reviewed-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de>

Sebastian