Skip to contentSkip to conversation

hippi: fix possible buffer overflow caused by bad DMA value in rr_start_xmit()

PATCH Superseded Huai-Yuan Liu 3 participants 3 messages Jun 20, 2024 – Jun 21, 2024

Subsystems (2)

v2 of 4 Superseded Latest is v4 →

lore
Share & tools
lore
#0 lore raw

PATCH v2 1/1 hippi: fix possible buffer overflow caused by bad DMA value in rr_start_xmit()

The value rrpriv->info->tx_ctrl is stored in DMA memory, and it is assigned to txctrl, so txctrl->pi can be modified at any time by malicious hardware. Becausetxctrl->pi is assigned to index, buffer overflow may occur when the code "rrpriv->tx_skbuff[index]" is executed.

To address this issue, the index should be checked.

Fixes f33a7251c825 ("hippi: switch from 'pci_' to 'dma_' API")

---
V2:
* In patch V2, we remove the first condition in if statement and use netdev_err() instead of printk().
Thanks Paolo Abeni for helpful advice.

---
 drivers/net/hippi/rrunner.c | 5 +++++
 1 file changed, 5 insertions(+)
drivers/net/hippi/rrunner.c +5 -0
index aa8f828a0ae7..6b0056ced922 100644
@@ -1440,6 +1440,11 @@ static netdev_tx_t rr_start_xmit(struct sk_buff *skb,
 	txctrl = &rrpriv->info->tx_ctrl;
 
 	index = txctrl->pi;
+	if (index >= TX_RING_ENTRIES) {
+		netdev_err(dev, "invalid index value %02x\n", index);
+		spin_unlock_irqrestore(&rrpriv->lock, flags);
+		return NETDEV_TX_BUSY;
+	}
 
 	rrpriv->tx_skbuff[index] = skb;
 	set_rraddr(&rrpriv->tx_ring[index].addr,
Signature · 1 line
2.34.1
#1 lore raw
Quoted 23 lines

The value rrpriv->info->tx_ctrl is stored in DMA memory, and it is assigned to txctrl, so txctrl->pi can be modified at any time by malicious hardware. Becausetxctrl->pi is assigned to index, buffer overflow may occur when the code "rrpriv-

Quoted 1 line

tx_skbuff[index]" is executed.

To address this issue, the index should be checked.

Fixes f33a7251c825 ("hippi: switch from 'pci_' to 'dma_' API")

---
V2:
* In patch V2, we remove the first condition in if statement and use netdev_err() instead of printk().
Thanks Paolo Abeni for helpful advice.
---
drivers/net/hippi/rrunner.c | 5 +++++

1 file changed, 5 insertions(+)
drivers/net/hippi/rrunner.c index renamed from drivers/net/hippi/rrunner.c +0 -0

aa8f828a0ae7..6b0056ced922 100644

drivers/net/hippi/rrunner.c static netdev_tx_t rr_start_xmit(struct sk_buff *skb, · line 1444

Quoted 5 lines
drivers/net/hippi/rrunner.c +2 -0
@@ -1440,6 +1440,11 @@ static netdev_tx_t rr_start_xmit(struct sk_buff *skb,
 	txctrl = &rrpriv->info->tx_ctrl;
 
 	index = txctrl->pi;
+	if (index >= TX_RING_ENTRIES) {
+		netdev_err(dev, "invalid index value %02x\n", index);

Much better would be to use netif_msg_tx_err which can be enabled/disabled instead of dumping on console, which would be annoying if there are many errors.

Thanks,
Sunil.

maintainer
#2 lore raw
Quoted 6 lines · On Fri, 21 Jun 2024 05:11:41 +0000 Sunil Kovvuri Goutham wrote:

drivers/net/hippi/rrunner.c static netdev_tx_t rr_start_xmit(struct sk_buff *skb, · line 1444

Quoted 2 lines
drivers/net/hippi/rrunner.c +2 -0
+	if (index >= TX_RING_ENTRIES) {
+		netdev_err(dev, "invalid index value %02x\n", index);  

Much better would be to use netif_msg_tx_err which can be enabled/disabled instead of dumping on console, which would be annoying if there are many errors.

Doesn't it require ethtool to be changed tho? This driver doesn't have any ethtool ops AFAICT. I'd go for netdev_err_once().

But more importantly, I think you should stop the queue before returning BUSY.