drivers/net/phy/mscc/mscc_ptp.c | 18 ++++++++---------- 1 file changed, 8 insertions(+), 10 deletions(-)
When transmitting a PTP frame which is timestamp using 2 step, the
following warning appears if CONFIG_PROVE_LOCKING is enabled:
=============================
[ BUG: Invalid wait context ]
6.17.0-rc1-00326-ge6160462704e #427 Not tainted
-----------------------------
ptp4l/119 is trying to lock:
c2a44ed4 (&vsc8531->ts_lock){+.+.}-{3:3}, at: vsc85xx_txtstamp+0x50/0xac
other info that might help us debug this:
context-{4:4}
4 locks held by ptp4l/119:
#0: c145f068 (rcu_read_lock_bh){....}-{1:2}, at: __dev_queue_xmit+0x58/0x1440
#1: c29df974 (dev->qdisc_tx_busylock ?: &qdisc_tx_busylock){+...}-{2:2}, at: __dev_queue_xmit+0x5c4/0x1440
#2: c2aaaad0 (_xmit_ETHER#2){+.-.}-{2:2}, at: sch_direct_xmit+0x108/0x350
#3: c2aac170 (&lan966x->tx_lock){+.-.}-{2:2}, at: lan966x_port_xmit+0xd0/0x350
stack backtrace:
CPU: 0 UID: 0 PID: 119 Comm: ptp4l Not tainted 6.17.0-rc1-00326-ge6160462704e #427 NONE
Hardware name: Generic DT based system
Call trace:
unwind_backtrace from show_stack+0x10/0x14
show_stack from dump_stack_lvl+0x7c/0xac
dump_stack_lvl from __lock_acquire+0x8e8/0x29dc
__lock_acquire from lock_acquire+0x108/0x38c
lock_acquire from __mutex_lock+0xb0/0xe78
__mutex_lock from mutex_lock_nested+0x1c/0x24
mutex_lock_nested from vsc85xx_txtstamp+0x50/0xac
vsc85xx_txtstamp from lan966x_fdma_xmit+0xd8/0x3a8
lan966x_fdma_xmit from lan966x_port_xmit+0x1bc/0x350
lan966x_port_xmit from dev_hard_start_xmit+0xc8/0x2c0
dev_hard_start_xmit from sch_direct_xmit+0x8c/0x350
sch_direct_xmit from __dev_queue_xmit+0x680/0x1440
__dev_queue_xmit from packet_sendmsg+0xfa4/0x1568
packet_sendmsg from __sys_sendto+0x110/0x19c
__sys_sendto from sys_send+0x18/0x20
sys_send from ret_fast_syscall+0x0/0x1c
Exception stack(0xf0b05fa8 to 0xf0b05ff0)
5fa0: 00000001 0000000e 0000000e 0004b47a 0000003a 00000000
5fc0: 00000001 0000000e 00000000 00000121 0004af58 00044874 00000000 00000000
5fe0: 00000001 bee9d420 00025a10 b6e75c7c
So, instead of using the ts_lock for tx_queue, use the spinlock that
skb_buff_head has.
Reviewed-by: Vadim Fedorenko <vadim.fedorenko@linux.dev>
Fixes: 7d272e63e0979d ("net: phy: mscc: timestamping and PHC support")
Signed-off-by: Horatiu Vultur <horatiu.vultur@microchip.com>
---
v2->v3:
- replace skb_queue_len with skb_queue_len_lockless
v1->v2:
- initialize tx_queue in ptp_probe
- purge the tx_queue when the driver is removed or when TX timestamping
is OFF
---
drivers/net/phy/mscc/mscc_ptp.c | 18 ++++++++----------
1 file changed, 8 insertions(+), 10 deletions(-)
diff --git a/drivers/net/phy/mscc/mscc_ptp.c b/drivers/net/phy/mscc/mscc_ptp.c
index 72847320cb652..d692df7d975c7 100644
--- a/drivers/net/phy/mscc/mscc_ptp.c
+++ b/drivers/net/phy/mscc/mscc_ptp.c
@@ -456,12 +456,12 @@ static void vsc85xx_dequeue_skb(struct vsc85xx_ptp *ptp)
*p++ = (reg >> 24) & 0xff;
}
- len = skb_queue_len(&ptp->tx_queue);
+ len = skb_queue_len_lockless(&ptp->tx_queue);
if (len < 1)
return;
while (len--) {
- skb = __skb_dequeue(&ptp->tx_queue);
+ skb = skb_dequeue(&ptp->tx_queue);
if (!skb)
return;
@@ -486,7 +486,7 @@ static void vsc85xx_dequeue_skb(struct vsc85xx_ptp *ptp)
* packet in the FIFO right now, reschedule it for later
* packets.
*/
- __skb_queue_tail(&ptp->tx_queue, skb);
+ skb_queue_tail(&ptp->tx_queue, skb);
}
}
@@ -1068,6 +1068,7 @@ static int vsc85xx_hwtstamp(struct mii_timestamper *mii_ts,
case HWTSTAMP_TX_ON:
break;
case HWTSTAMP_TX_OFF:
+ skb_queue_purge(&vsc8531->ptp->tx_queue);
break;
default:
return -ERANGE;
@@ -1092,9 +1093,6 @@ static int vsc85xx_hwtstamp(struct mii_timestamper *mii_ts,
mutex_lock(&vsc8531->ts_lock);
- __skb_queue_purge(&vsc8531->ptp->tx_queue);
- __skb_queue_head_init(&vsc8531->ptp->tx_queue);
-
/* Disable predictor while configuring the 1588 block */
val = vsc85xx_ts_read_csr(phydev, PROCESSOR,
MSCC_PHY_PTP_INGR_PREDICTOR);
@@ -1180,9 +1178,7 @@ static void vsc85xx_txtstamp(struct mii_timestamper *mii_ts,
skb_shinfo(skb)->tx_flags |= SKBTX_IN_PROGRESS;
- mutex_lock(&vsc8531->ts_lock);
- __skb_queue_tail(&vsc8531->ptp->tx_queue, skb);
- mutex_unlock(&vsc8531->ts_lock);
+ skb_queue_tail(&vsc8531->ptp->tx_queue, skb);
return;
out:
@@ -1548,6 +1544,7 @@ void vsc8584_ptp_deinit(struct phy_device *phydev)
if (vsc8531->ptp->ptp_clock) {
ptp_clock_unregister(vsc8531->ptp->ptp_clock);
skb_queue_purge(&vsc8531->rx_skbs_list);
+ skb_queue_purge(&vsc8531->ptp->tx_queue);
}
}
@@ -1571,7 +1568,7 @@ irqreturn_t vsc8584_handle_ts_interrupt(struct phy_device *phydev)
if (rc & VSC85XX_1588_INT_FIFO_ADD) {
vsc85xx_get_tx_ts(priv->ptp);
} else if (rc & VSC85XX_1588_INT_FIFO_OVERFLOW) {
- __skb_queue_purge(&priv->ptp->tx_queue);
+ skb_queue_purge(&priv->ptp->tx_queue);
vsc85xx_ts_reset_fifo(phydev);
}
@@ -1591,6 +1588,7 @@ int vsc8584_ptp_probe(struct phy_device *phydev)
mutex_init(&vsc8531->phc_lock);
mutex_init(&vsc8531->ts_lock);
skb_queue_head_init(&vsc8531->rx_skbs_list);
+ skb_queue_head_init(&vsc8531->ptp->tx_queue);
/* Retrieve the shared load/save GPIO. Request it as non exclusive as
* the same GPIO can be requested by all the PHYs of the same package.
--
2.34.1On Tue, 2 Sep 2025 14:12:59 +0200 Horatiu Vultur wrote:
> - len = skb_queue_len(&ptp->tx_queue);
> + len = skb_queue_len_lockless(&ptp->tx_queue);
> if (len < 1)
> return;
>
> while (len--) {
> - skb = __skb_dequeue(&ptp->tx_queue);
> + skb = skb_dequeue(&ptp->tx_queue);
> if (!skb)
> return;
Isn't checking len completely unnecessary? skb_dequeue() will return
null if the list is empty, which you are already handling.
--
pw-bot: cr
The 09/03/2025 17:05, Jakub Kicinski wrote:
Hi Jakub,
>
> On Tue, 2 Sep 2025 14:12:59 +0200 Horatiu Vultur wrote:
> > - len = skb_queue_len(&ptp->tx_queue);
> > + len = skb_queue_len_lockless(&ptp->tx_queue);
> > if (len < 1)
> > return;
> >
> > while (len--) {
> > - skb = __skb_dequeue(&ptp->tx_queue);
> > + skb = skb_dequeue(&ptp->tx_queue);
> > if (!skb)
> > return;
>
> Isn't checking len completely unnecessary? skb_dequeue() will return
> null if the list is empty, which you are already handling.
I don't think so, because if the skb doesn't match the sig then the skb
is added back to the tx_queue.
/* Valid signature but does not match the one of the
* packet in the FIFO right now, reschedule it for later
* packets.
*/
skb_queue_tail(&ptp->tx_queue, skb);
Then if we change to check until skb_dequeue returns NULL we might be in
an infinite loop.
> --
> pw-bot: cr
--
/Horatiu
© 2016 - 2025 Red Hat, Inc.