Skip to contentSkip to conversation

ptp: netc: explicitly clear TMR_OFF during initialization

PATCH RESEND Applied [email protected] 4 participants 4 messages Jul 20 – Jul 23

Subsystems (5)
lore
Share & tools
lore
#0 lore raw

PATCH RESEND 1/1 ptp: netc: explicitly clear TMR_OFF during initialization

From: Clark Wang <[email protected]>

The NETC timer does not support function level reset, so TMR_OFF_L/H registers are not cleared by pcie_flr(). If TMR_OFF was set to a non-zero value in a previous binding, it will persist across driver rebind and cause inaccurate PTP time.

There is also a hardware issue: after a warm reset or soft reset, TMR_OFF_L/H registers appear to be cleared to zero, but the timer clock domain internally retains the stale value. When the timer is re-enabled, TMR_CUR_TIME continues to track the old offset until TMR_OFF is written explicitly. This can cause incorrect PTP timestamps and even PTP clock synchronization failures.

Per the recommendation from the IP team, explicitly write 0 to TMR_OFF in netc_timer_init() to flush the internally cached value and ensure TMR_CUR_TIME follows the freshly initialized counter.

Fixes 87a201d59963 ("ptp: netc: add NETC V4 Timer PTP driver support")

---
No changes, just collect Reviewed-by tag
v1 link: https://lore.kernel.org/imx/[email protected]/

---
 drivers/ptp/ptp_netc.c | 1 +
 1 file changed, 1 insertion(+)
drivers/ptp/ptp_netc.c +1 -0
index 94e952ee6990..5e381c354d74 100644
@@ -779,6 +779,7 @@ static void netc_timer_init(struct netc_timer *priv)
 	netc_timer_wr(priv, NETC_TMR_FIPER_CTRL, fiper_ctrl);
 	netc_timer_wr(priv, NETC_TMR_ECTRL, NETC_TMR_DEFAULT_ETTF_THR);
 
+	netc_timer_offset_write(priv, 0);
 	ktime_get_real_ts64(&now);
 	ns = timespec64_to_ns(&now);
 	netc_timer_cnt_write(priv, ns);
Signature · 1 line
2.34.1
Reviewed-by on v1 1/1
#1 lore raw
Quoted 14 lines · On Mon, Jul 20, 2026 at 09:25:08AM +0800, [email protected] wrote:

From: Clark Wang <[email protected]>

The NETC timer does not support function level reset, so TMR_OFF_L/H registers are not cleared by pcie_flr(). If TMR_OFF was set to a non-zero value in a previous binding, it will persist across driver rebind and cause inaccurate PTP time.

There is also a hardware issue: after a warm reset or soft reset, TMR_OFF_L/H registers appear to be cleared to zero, but the timer clock domain internally retains the stale value. When the timer is re-enabled, TMR_CUR_TIME continues to track the old offset until TMR_OFF is written explicitly. This can cause incorrect PTP timestamps and even PTP clock synchronization failures.

Quoted 8 lines

Per the recommendation from the IP team, explicitly write 0 to TMR_OFF in netc_timer_init() to flush the internally cached value and ensure TMR_CUR_TIME follows the freshly initialized counter.

Fixes 87a201d59963 ("ptp: netc: add NETC V4 Timer PTP driver support")
#2 lore raw
In reply to [email protected]
Thank you for your contribution! Sashiko AI review found 2 potential issue(s) to consider:

Thank you for your contribution! Sashiko AI review found 2 potential issue(s) to consider:

Pre-existing issues:
- [High] The driver requests the MSI-X interrupt during probe before sanitizing hardware interrupt state. Combined with a logic error in the ISR, this causes an uninterruptible interrupt storm (CPU lockup).
- [High] Race condition in netc_timer_remove allows user-space ioctls to leave hardware interrupts active after driver unbind.
--

--- Patch [1]: [PATCH RESEND v1 net] ptp: netc: explicitly clear TMR_OFF during initialization --- commit f3b81ecce462c4ab69c0789be51860d8eb8b3d21
Author: Clark Wang <[email protected]>

ptp: netc: explicitly clear TMR_OFF during initialization

The NETC timer does not support function level reset, so TMR_OFF_L/H registers are not cleared by pcie_flr(). If TMR_OFF was set to a non-zero value in a previous binding, it will persist across driver rebind and cause inaccurate PTP time.

drivers/ptp/ptp_netc.c static void netc_timer_init(struct netc_timer *priv) · line 782

Quoted 9 lines
drivers/ptp/ptp_netc.c +1 -0
index 94e952ee69902..5e381c354d746 100644
@@ -779,6 +779,7 @@ static void netc_timer_init(struct netc_timer *priv)
 	netc_timer_wr(priv, NETC_TMR_FIPER_CTRL, fiper_ctrl);
 	netc_timer_wr(priv, NETC_TMR_ECTRL, NETC_TMR_DEFAULT_ETTF_THR);
 
+	netc_timer_offset_write(priv, 0);

[Severity: High]
This is a pre-existing issue, but since netc_timer_init() is called during netc_timer_probe(), does the driver risk an interrupt storm?

In netc_timer_probe(), the driver calls request_irq() via netc_timer_init_msix_irq() before calling netc_timer_init():

drivers/ptp/ptp_netc.c:netc_timer_probe() {
    ...
	err = netc_timer_init_msix_irq(priv);
	if (err)
		goto timer_pci_remove;

	netc_timer_init(priv);
    ...
}

If hardware has a pending interrupt from a previous driver session, it could fire immediately. Furthermore, the ISR appears to mask out unhandled events before writing back to clear them, and unconditionally returns IRQ_HANDLED:

drivers/ptp/ptp_netc.c:netc_timer_isr() {
    ...
	tmr_event = netc_timer_rd(priv, NETC_TMR_TEVENT);
	tmr_event &= priv->tmr_emask;
	/* Clear interrupts status */
	netc_timer_wr(priv, NETC_TMR_TEVENT, tmr_event);
    ...
	return IRQ_HANDLED;
}

Since priv->tmr_emask is 0 during probe, tmr_event becomes 0, so the hardware interrupt is never cleared. Because IRQ_HANDLED is returned unconditionally, will this lead to an uninterruptible infinite interrupt storm?

[Severity: High]
This isn't a bug introduced by this patch, but is there a race condition during driver unbind that could leave hardware interrupts active?

In netc_timer_remove(), the driver writes to hardware registers to disable interrupts before unregistering the PTP clock:

drivers/ptp/ptp_netc.c:netc_timer_remove() {
    ...
	netc_timer_wr(priv, NETC_TMR_TEMASK, 0);
	netc_timer_wr(priv, NETC_TMR_CTRL, 0);
	ptp_clock_unregister(priv->clock);
    ...
}

Since ptp_clock_unregister() is the barrier that waits for ongoing ioctls and blocks new ones, can a concurrent ioctl execute after the mask is cleared but before unregistration finishes?

For example, a concurrent PTP_CLK_REQ_EXTTS ioctl calling netc_timer_enable_extts():

drivers/ptp/ptp_netc.c:netc_timer_enable_extts() {
    ...
	netc_timer_wr(priv, NETC_TMR_TEMASK, priv->tmr_emask);

	spin_unlock_irqrestore(&priv->lock, flags);
    ...
}

Could this re-enable interrupts while the driver unbinds, leading to a use-after-free or system instability on the next probe?

Quoted 3 lines
drivers/ptp/ptp_netc.c +0 -0
 	ktime_get_real_ts64(&now);
 	ns = timespec64_to_ns(&now);
 	netc_timer_cnt_write(priv, ns);
Signature · 1 line
Sashiko AI review · https://sashiko.dev/#/patchset/[email protected]?part=1
#3 lore raw
In reply to [email protected]
This patch was applied to netdev/net.git (main) by Jakub Kicinski <[email protected]>:

Hello:

This patch was applied to netdev/net.git (main)
by Jakub Kicinski <[email protected]>:

Quoted 8 lines · On Mon, 20 Jul 2026 09:25:08 +0800 you wrote:

From: Clark Wang <[email protected]>

The NETC timer does not support function level reset, so TMR_OFF_L/H registers are not cleared by pcie_flr(). If TMR_OFF was set to a non-zero value in a previous binding, it will persist across driver rebind and cause inaccurate PTP time.

[...]

Here is the summary with links:
- [RESEND,v1,net] ptp: netc: explicitly clear TMR_OFF during initialization

    https://git.kernel.org/netdev/net/c/c3f2fc231a39

You are awesome, thank you!

Signature · 2 lines
Deet-doot-dot, I am a bot.
https://korg.docs.kernel.org/patchwork/pwbot.html