netdev - Re: [PATCH net-next v4] Add Common Applications Kept Enhanced (cake) qdisc

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite for Android: free password hash cracker in your pocket

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-ID: <87in8c4vn7.fsf@toke.dk>
Date:   Fri, 27 Apr 2018 15:38:20 +0200
From:   Toke Høiland-Jørgensen <toke@...e.dk>
To:     Eric Dumazet <eric.dumazet@...il.com>, netdev@...r.kernel.org
Cc:     cake@...ts.bufferbloat.net, Dave Taht <dave.taht@...il.com>
Subject: Re: [PATCH net-next v4] Add Common Applications Kept Enhanced (cake) qdisc

Eric Dumazet <eric.dumazet@...il.com> writes:

> On 04/27/2018 05:17 AM, Toke Høiland-Jørgensen wrote:
>
> ...
>
>> +
>> +static struct sk_buff *cake_ack_filter(struct cake_sched_data *q,
>> +				       struct cake_flow *flow)
>> +{
>> +	int seglen;
>> +	struct sk_buff *skb = flow->tail, *skb_check, *skb_check_prev;
>> +	struct iphdr *iph, *iph_check;
>> +	struct ipv6hdr *ipv6h, *ipv6h_check;
>> +	struct tcphdr *tcph, *tcph_check;
>> +	bool otherconn_ack_seen = false;
>> +	struct sk_buff *otherconn_checked_to = NULL;
>> +	bool thisconn_redundant_seen = false, thisconn_seen_last = false;
>> +	struct sk_buff *thisconn_checked_to = NULL, *thisconn_ack = NULL;
>> +	bool aggressive = q->ack_filter == CAKE_ACK_AGGRESSIVE;
>> +
>> +	/* no other possible ACKs to filter */
>> +	if (flow->head == skb)
>> +		return NULL;
>> +
>> +	iph = skb->encapsulation ? inner_ip_hdr(skb) : ip_hdr(skb);
>> +	ipv6h = skb->encapsulation ? inner_ipv6_hdr(skb) : ipv6_hdr(skb);
>> +
>> +	/* check that the innermost network header is v4/v6, and contains TCP */
>> +	if (pskb_may_pull(skb, ((unsigned char *)iph - skb->head) + sizeof(struct iphdr)) &&
>> +	    iph->version == 4) {
>> +		if (iph->protocol != IPPROTO_TCP)
>> +			return NULL;
>> +		seglen = ntohs(iph->tot_len) - (4 * iph->ihl);
>> +		tcph = (struct tcphdr *)((void *)iph + (4 * iph->ihl));
>> +		if (!pskb_may_pull(skb, ((unsigned char *)tcph - skb->head) + sizeof(struct tcphdr)))
>> +			return NULL;
>> +	} else if (pskb_may_pull(skb, ((unsigned char *)ipv6h - skb->head) + sizeof(struct ipv6hdr) + sizeof(struct tcphdr)) &&
>> +	           ipv6h->version == 6) {
>> +		if (ipv6h->nexthdr != IPPROTO_TCP)
>> +			return NULL;
>> +		seglen = ntohs(ipv6h->payload_len);
>> +		tcph = (struct tcphdr *)((void *)ipv6h +
>> +					 sizeof(struct ipv6hdr));
>> +	} else {
>> +		return NULL;
>> +	}
>> +
>
>
> This is still broken.
>
> After pskb_may_pull(), skb->head might have been reallocated.
>
> You need to recompute iph , ipv6h, tcph, otherwise you are reading
> freed memory and crash kernels with sufficient debugging (KASAN and
> other CONFIG_DEBUG_PAGEALLOC / CONFIG_DEBUG_SLAB like options)

Ah, right. Will fix.

Is it safe to dereference the iph pointer before calling
pskb_may_pull()?

-Toke