2022-08-15 15:00:48

by Aleksander Jan Bajkowski

Subject: [PATCH net v2 1/3] net: lantiq_xrx200: confirm skb is allocated before using

xrx200_hw_receive() assumes build_skb() always works and goes straight
to skb_reserve(). However, build_skb() can fail under memory pressure.

Add a check in case build_skb() failed to allocate and return NULL.

Fixes: e015593573b3 ("net: lantiq_xrx200: convert to build_skb")
Reported-by: Eric Dumazet <[email protected]>
Signed-off-by: Aleksander Jan Bajkowski <[email protected]>
---
drivers/net/ethernet/lantiq_xrx200.c | 7 +++++++
1 file changed, 7 insertions(+)

diff --git a/drivers/net/ethernet/lantiq_xrx200.c b/drivers/net/ethernet/lantiq_xrx200.c
index 5edb68a8aab1..6a83a6c19484 100644
--- a/drivers/net/ethernet/lantiq_xrx200.c
+++ b/drivers/net/ethernet/lantiq_xrx200.c
@@ -239,6 +239,13 @@ static int xrx200_hw_receive(struct xrx200_chan *ch)
}

skb = build_skb(buf, priv->rx_skb_size);
+ if (!skb) {
+ skb_free_frag(buf);
+ net_dev->stats.rx_dropped++;
+ netdev_err(net_dev, "failed to build skb\n");
+ return -ENOMEM;
+ }
+
skb_reserve(skb, NET_SKB_PAD);
skb_put(skb, len);

--
2.30.2

2022-08-17 03:48:07

by Jakub Kicinski

Subject: Re: [PATCH net v2 1/3] net: lantiq_xrx200: confirm skb is allocated before using

On Mon, 15 Aug 2022 16:57:38 +0200 Aleksander Jan Bajkowski wrote:
> + netdev_err(net_dev, "failed to build skb\n");

I don't thinkg this is needed, is build_skb() using GFP_NOWARN?
Otherwise there will be an OOM splat anyway. Do check, I'm not sure.

If there won't be an OOM splat this line should be rate limited.