Skip to contentSkip to conversation

sfc: check for NULL efx->ptp_data in efx_ptp_event

PATCH Superseded Edward Cree 2 participants 2 messages Feb 24, 2014

Subsystems (2)

v1 of 2 Superseded Latest is v2 →

lore
Share & tools
lore
#0 lore raw

PATCH 1/1 sfc: check for NULL efx->ptp_data in efx_ptp_event

If we receive a PTP event from the NIC when we haven't set up PTP state in the driver, we attempt to read through a NULL pointer efx->ptp_data, triggering a panic.

---
While this shouldn't happen with existing released firmware, future sfn7322f firmware is planned to enable PPS irrespective of whether timestamping (and thus PTP) is enabled. We are considering workarounds, but this defensive patch seemed like a good idea anyway in case the firmware gets in a muddle.
Also, in principle this shouldn't happen on 3.14+ kernels as the driver should always set up PTP state at probe time (though there might still be a race against the firmware sending its first PPS), however 3.12 and 3.13 both supported EF10 but not PTP-on-EF10 and thus they never set up the PTP state; hence this fix should also be backported to those series.

--- linux-3.13/drivers/net/ethernet/sfc/ptp.c.orig 2014-01-20
02:40:07.000000000 +0000
+++ linux-3.13/drivers/net/ethernet/sfc/ptp.c 2014-02-24
17:46:31.753497000 +0000
@@ -1360,6 +1360,13 @@ void efx_ptp_event(struct efx_nic *efx,
struct efx_ptp_data *ptp = efx->ptp_data; int code = EFX_QWORD_FIELD(*ev, MCDI_EVENT_CODE);

+ if (!ptp) {
+ if (net_ratelimit())
+ netif_warn(efx, drv, efx->net_dev,
+ "Received PTP event but PTP not set up\n");
+ return;
+ }
+
if (!ptp->enabled)
return;

#1 lore raw
Quoted 5 lines · Edward Cree

If we receive a PTP event from the NIC when we haven't set up PTP state in the driver, we attempt to read through a NULL pointer efx->ptp_data, triggering a panic.

Quoted 11 lines

---
While this shouldn't happen with existing released firmware, future sfn7322f firmware is planned to enable PPS irrespective of whether timestamping (and thus PTP) is enabled. We are considering workarounds, but this defensive patch seemed like a good idea anyway in case the firmware gets in a muddle.
Also, in principle this shouldn't happen on 3.14+ kernels as the driver should always set up PTP state at probe time (though there might still be a race against the firmware sending its first PPS), however 3.12 and 3.13 both supported EF10 but not PTP-on-EF10 and thus they never set up the PTP state; hence this fix should also be backported to those series.

Good thinking.

Quoted 4 lines

--- linux-3.13/drivers/net/ethernet/sfc/ptp.c.orig 2014-01-20
02:40:07.000000000 +0000
+++ linux-3.13/drivers/net/ethernet/sfc/ptp.c 2014-02-24
17:46:31.753497000 +0000

This patch has been word-wrapped and can't be applied; see Documentation/email-clients.txt

Ben.

Quoted 14 lines
(quoted diff) +7 -0
@@ -1360,6 +1360,13 @@ void efx_ptp_event(struct efx_nic *efx,
 	struct efx_ptp_data *ptp = efx->ptp_data;
 	int code = EFX_QWORD_FIELD(*ev, MCDI_EVENT_CODE);
 
+	if (!ptp) {
+		if (net_ratelimit())
+			netif_warn(efx, drv, efx->net_dev,
+				   "Received PTP event but PTP not set up\n");
+		return;
+	}
+
 	if (!ptp->enabled)
 		return;
 
Signature · 3 lines
Ben Hutchings
Beware of bugs in the above code;
I have only proved it correct, not tried it. - Donald Knuth