mt7615_queue_rx_skb() walks the TXS entries up to skb->len, but that length only describes the whole packet. For RX skbs that are not fully linear, such as the ones built from the fragmented USB RX path, skb->data only covers skb_headlen(skb) bytes, so the loop advances past the skb head buffer and mt7615_mac_add_txs() reads outside of the allocated slab object: BUG: KASAN: slab-out-of-bounds in mt7615_mac_add_txs+0x5c8/0x650 Read of size 4 at addr ffff888005097b40 Bound the iteration by skb_headlen(skb) instead. Fully linear skbs keep the previous behaviour, since both lengths are equal there. Fixes: eb99cc95c3b6 ("mt76: mt7615: introduce mt7663u support") Cc: stable@vger.kernel.org Co-developed-by: Bin Yu Signed-off-by: Bin Yu Co-developed-by: Cheng Wen Signed-off-by: Cheng Wen Signed-off-by: YuanZhe Liu <25031212351@stu.xidian.edu.cn> --- drivers/net/wireless/mediatek/mt76/mt7615/mac.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/drivers/net/wireless/mediatek/mt76/mt7615/mac.c b/drivers/net/wireless/mediatek/mt76/mt7615/mac.c index aad232c5a6fa..86c6b443ef71 100644 --- a/drivers/net/wireless/mediatek/mt76/mt7615/mac.c +++ b/drivers/net/wireless/mediatek/mt76/mt7615/mac.c @@ -1620,7 +1620,7 @@ void mt7615_queue_rx_skb(struct mt76_dev *mdev, enum mt76_rxq_id q, { struct mt7615_dev *dev = container_of(mdev, struct mt7615_dev, mt76); __le32 *rxd = (__le32 *)skb->data; - __le32 *end = (__le32 *)&skb->data[skb->len]; + __le32 *end = (__le32 *)&skb->data[skb_headlen(skb)]; enum rx_pkt_type type; u16 flag; -- 2.34.1