6.9-stable review patch. If anyone has any objections, please let me know.
------------------
From: Radu Rendec <rrendec@xxxxxxxxxx>
[ Upstream commit 9a0c28efeec6383ef22e97437616b920e7320b67 ]
The use-after-free is actually in rswitch_tx_free(), which is inlined in
rswitch_poll(). Since `skb` and `gq->skbs[gq->dirty]` are in fact the
same pointer, the skb is first freed using dev_kfree_skb_any(), then the
value in skb->len is used to update the interface statistics.
Let's move around the instructions to use skb->len before the skb is
freed.
This bug is trivial to reproduce using KFENCE. It will trigger a splat
every few packets. A simple ARP request or ICMP echo request is enough.
Fixes: 271e015b9153 ("net: rswitch: Add unmap_addrs instead of dma address in each desc")
Signed-off-by: Radu Rendec <rrendec@xxxxxxxxxx>
Reviewed-by: Yoshihiro Shimoda <yoshihiro.shimoda.uh@xxxxxxxxxxx>
Reviewed-by: Niklas Söderlund <niklas.soderlund+renesas@xxxxxxxxxxxx>
Link: https://patch.msgid.link/20240702210838.2703228-1-rrendec@xxxxxxxxxx
Signed-off-by: Jakub Kicinski <kuba@xxxxxxxxxx>
Signed-off-by: Sasha Levin <sashal@xxxxxxxxxx>
---
drivers/net/ethernet/renesas/rswitch.c | 4 ++--
1 file changed, 2 insertions(+), 2 deletions(-)
diff --git a/drivers/net/ethernet/renesas/rswitch.c b/drivers/net/ethernet/renesas/rswitch.c
index dcab638c57fe8..24c90d8f5a442 100644
--- a/drivers/net/ethernet/renesas/rswitch.c
+++ b/drivers/net/ethernet/renesas/rswitch.c
@@ -871,13 +871,13 @@ static void rswitch_tx_free(struct net_device *ndev)
dma_rmb();
skb = gq->skbs[gq->dirty];
if (skb) {
+ rdev->ndev->stats.tx_packets++;
+ rdev->ndev->stats.tx_bytes += skb->len;
dma_unmap_single(ndev->dev.parent,
gq->unmap_addrs[gq->dirty],
skb->len, DMA_TO_DEVICE);
dev_kfree_skb_any(gq->skbs[gq->dirty]);
gq->skbs[gq->dirty] = NULL;
- rdev->ndev->stats.tx_packets++;
- rdev->ndev->stats.tx_bytes += skb->len;
}
desc->desc.die_dt = DT_EEMPTY;
}
--
2.43.0
