Skip to contentSkip to conversation

net/mlx5: Fix OOB access and stack information leak in PTP event handling

PATCH In review Prathamesh Deshpande 3 participants 4 messages Apr 12 – Apr 15

Subsystems (4)
lore
Share & tools
lore
#0 lore raw

PATCH v4 1/1 net/mlx5: Fix OOB access and stack information leak in PTP event handling

In mlx5_pps_event(), several critical issues were identified:

1. The 'pin' index from the hardware event was used without bounds checking to index 'pin_config' and 'pps_info->start'. Check against MAX_PIN_NUM to prevent out-of-bounds access.
2. 'ptp_event' was not zero-initialized, potentially leaking stack memory through the union.
3. A NULL 'pin_config' could be dereferenced if initialization failed.
4. 'clock->ptp' could be NULL if ptp_clock_register() failed.

Fixes 7c39afb394c7 ("net/mlx5: PTP code migration to driver core section")

---
v4:
- Validate pin index against MAX_PIN_NUM instead of n_pins [Carolina]. v3:
- Fix union corruption by using a local timestamp variable [Sashiko].
- Validate pin index against n_pins with WARN_ON_ONCE [Carolina].
- Remove redundant pin < 0 check and cleanup TODO comment.
v2:
- Zero-initialize ptp_event to prevent stack information leak [Sashiko].
- Add bounds check for hardware pin index to prevent OOB access [Sashiko].
- Add NULL guard for pin_config to handle initialization failures [Sashiko].
- Add NULL check for clock->ptp as originally intended.

 .../net/ethernet/mellanox/mlx5/core/lib/clock.c | 17 ++++++++++++-----
 1 file changed, 12 insertions(+), 5 deletions(-)
drivers/net/ethernet/mellanox/mlx5/core/lib/clock.c +12 -5
index bd4e042077af..ff03dfa12a67 100644
@@ -1164,16 +1164,22 @@ static int mlx5_pps_event(struct notifier_block *nb,
 							       pps_nb);
 	struct mlx5_core_dev *mdev = clock_state->mdev;
 	struct mlx5_clock *clock = mdev->clock;
-	struct ptp_clock_event ptp_event;
+	struct ptp_clock_event ptp_event = {};
 	struct mlx5_eqe *eqe = data;
 	int pin = eqe->data.pps.pin;
 	unsigned long flags;
 	u64 ns;
 
+	if (!clock->ptp_info.pin_config)
+		return NOTIFY_OK;
+
+	if (WARN_ON_ONCE(pin >= MAX_PIN_NUM))
+		return NOTIFY_OK;
+
 	switch (clock->ptp_info.pin_config[pin].func) {
 	case PTP_PF_EXTTS:
 		ptp_event.index = pin;
-		ptp_event.timestamp = mlx5_real_time_mode(mdev) ?
+		ns = mlx5_real_time_mode(mdev) ?
 			mlx5_real_time_cyc2time(clock,
 						be64_to_cpu(eqe->data.pps.time_stamp)) :
 			mlx5_timecounter_cyc2time(clock,
@@ -1181,12 +1187,13 @@ static int mlx5_pps_event(struct notifier_block *nb,
 		if (clock->pps_info.enabled) {
 			ptp_event.type = PTP_CLOCK_PPSUSR;
 			ptp_event.pps_times.ts_real =
-					ns_to_timespec64(ptp_event.timestamp);
+					ns_to_timespec64(ns);
 		} else {
 			ptp_event.type = PTP_CLOCK_EXTTS;
+			ptp_event.timestamp = ns;
 		}
-		/* TODOL clock->ptp can be NULL if ptp_clock_register fails */
-		ptp_clock_event(clock->ptp, &ptp_event);
+		if (clock->ptp)
+			ptp_clock_event(clock->ptp, &ptp_event);
 		break;
 	case PTP_PF_PEROUT:
 		if (clock->shared) {
Signature · 1 line
2.43.0
Reviewed-by on v4 1/1
#1 lore raw
Quoted 10 lines · Prathamesh Deshpande

In mlx5_pps_event(), several critical issues were identified:

1. The 'pin' index from the hardware event was used without bounds checking to index 'pin_config' and 'pps_info->start'. Check against MAX_PIN_NUM to prevent out-of-bounds access.
2. 'ptp_event' was not zero-initialized, potentially leaking stack memory through the union.
3. A NULL 'pin_config' could be dereferenced if initialization failed.
4. 'clock->ptp' could be NULL if ptp_clock_register() failed.

Quoted 3 lines
Fixes 7c39afb394c7 ("net/mlx5: PTP code migration to driver core section")
#2 lore raw
In reply to Prathamesh Deshpande
Quoted 5 lines · Prathamesh Deshpande

In mlx5_pps_event(), several critical issues were identified:

1. The 'pin' index from the hardware event was used without bounds checking to index 'pin_config' and 'pps_info->start'. Check against MAX_PIN_NUM to prevent out-of-bounds access.

You were told more than once that this is impossible.

<...>

drivers/net/ethernet/mellanox/mlx5/core/lib/clock.c static int mlx5_pps_event(struct notifier_block *nb, · line 1177

Quoted 2 lines
drivers/net/ethernet/mellanox/mlx5/core/lib/clock.c +2 -0
+	if (WARN_ON_ONCE(pin >= MAX_PIN_NUM))
+		return NOTIFY_OK;

Let's not add useless checks in fast path.

Thanks

#3 lore raw
Quoted 7 lines · Leon Romanovsky
Quoted 5 lines · Prathamesh Deshpande

In mlx5_pps_event(), several critical issues were identified:

1. The 'pin' index from the hardware event was used without bounds checking to index 'pin_config' and 'pps_info->start'. Check against MAX_PIN_NUM to prevent out-of-bounds access.

Quoted 8 lines

You were told more than once that this is impossible.

<...>

drivers/net/ethernet/mellanox/mlx5/core/lib/clock.c static int mlx5_pps_event(struct notifier_block *nb, · line 1177

Quoted 2 lines
drivers/net/ethernet/mellanox/mlx5/core/lib/clock.c +2 -0
+	if (WARN_ON_ONCE(pin >= MAX_PIN_NUM))
+		return NOTIFY_OK;

Let's not add useless checks in fast path.

Hi Leon,

Thanks for the feedback. I've addressed this in v5 by dropping the redundant pin bounds and pin_config checks to keep the fast path clean, focusing strictly on the stack leak and NULL clock guard fixes.

Thanks,
Prathamesh