[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <ZVcuqzgmLqzVBNdP@gauss3.secunet.de>
Date: Fri, 17 Nov 2023 10:13:15 +0100
From: Steffen Klassert <steffen.klassert@...unet.com>
To: Antony Antony <antony.antony@...unet.com>
CC: Herbert Xu <herbert@...dor.apana.org.au>, "David S. Miller"
<davem@...emloft.net>, <devel@...ux-ipsec.org>, Jakub Kicinski
<kuba@...nel.org>, <netdev@...r.kernel.org>
Subject: Re: [PATCH v2 ipsec-next 1/2] xfrm: introduce forwarding of ICMP
Error messages
On Fri, Oct 27, 2023 at 10:16:29AM +0200, Antony Antony wrote:
> +
> +static struct sk_buff *xfrm_icmp_flow_decode(struct sk_buff *skb,
> + unsigned short family,
> + struct flowi *fl,
> + struct flowi *fl1)
> +{
> + struct net *net = dev_net(skb->dev);
> + struct sk_buff *newskb = skb_clone(skb, GFP_ATOMIC);
> + int hl = family == AF_INET ? (sizeof(struct iphdr) + sizeof(struct icmphdr)) :
> + (sizeof(struct ipv6hdr) + sizeof(struct icmp6hdr));
> + skb_reset_network_header(newskb);
This is not needed there.
> +
> + if (!pskb_pull(newskb, hl))
> + return NULL;
This leaks newskb.
> + skb_reset_network_header(newskb);
> +
> + if (xfrm_decode_session_reverse(net, newskb, fl1, family) < 0) {
> + kfree_skb(newskb);
The newskb is not dropped because of an error, maybe better use
consume_skb().
> + XFRM_INC_STATS(net, LINUX_MIB_XFRMINHDRERROR);
This might bump a second stat counter for this packet. Also
this is not really an error, because you just can't parse
the payload of the icmp packet.
> + return NULL;
> + }
> +
> + fl1->flowi_oif = fl->flowi_oif;
> + fl1->flowi_mark = fl->flowi_mark;
> + fl1->flowi_tos = fl->flowi_tos;
> + nf_nat_decode_session(newskb, fl1, family);
> +
> + return newskb;
Why do you return newskb here? It is not used in all
of the calling functions.
The rest looks good, thanks!
Powered by blists - more mailing lists