netdev - Re: [PATCH] atm: Preserve value of skb->truesize when accounting to vcc

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: Windows password security audit tool. GUI, reports in PDF.

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

Message-ID: <10912b7594637a95567a2ffe3ef5dbc1.squirrel@twosheds.infradead.org>
Date:   Sat, 16 Jun 2018 15:19:48 -0000
From:   "David Woodhouse" <dwmw2@...radead.org>
To:     "Eric Dumazet" <eric.dumazet@...il.com>
Cc:     "David Woodhouse" <dwmw2@...radead.org>, netdev@...r.kernel.org,
        ldir@...byshire-bryant.me.uk
Subject: Re: [PATCH] atm: Preserve value of skb->truesize when accounting to
 vcc


>> Commit 14afee4b609 ("net: convert sock.sk_wmem_alloc from atomic_t to
>> refcount_t") did exactly what it was intended to do, and turned this
>> mostly-theoretical problem into a real one, causing PPPoATM to fail
>> immediately as sk_wmem_alloc underflows and atm_may_send() *immediately*
>> starts refusing to allow new packets.

 ...

>> Fixes: 14afee4b ("net: convert sock.sk_wmem_alloc from atomic_t to
>> refcount_t")
>
> This Fixes tag shoots the messenger really.

A little bit, yes. The text hopefully made that clear.

> I suggest to instead use :
>
> Fixes: 158f323b9868 ("net: adjust skb->truesize in pskb_expand_head()")
>
> Because even without the conversion to refcount_t, we could have a LOCKDEP
> splat in :
>
> filter = rcu_dereference_check(sk->sk_filter,
>                                atomic_read(&sk->sk_wmem_alloc) == 0);
>
> Note that some places make a further check even when LOCKDEP is not used.
>
> net/ipv4/af_inet.c:154:	WARN_ON(refcount_read(&sk->sk_wmem_alloc));
> net/iucv/af_iucv.c:405:	WARN_ON(refcount_read(&sk->sk_wmem_alloc));
> net/key/af_key.c:112:	WARN_ON(refcount_read(&sk->sk_wmem_alloc));
> net/netlink/af_netlink.c:410:	WARN_ON(refcount_read(&sk->sk_wmem_alloc));
> net/packet/af_packet.c:1286:	WARN_ON(refcount_read(&sk->sk_wmem_alloc));
> net/rxrpc/af_rxrpc.c:852:	WARN_ON(refcount_read(&sk->sk_wmem_alloc));
> net/unix/af_unix.c:490:	WARN_ON(refcount_read(&sk->sk_wmem_alloc));
>
>
> We might factorize these checks into __sk_destruct()
>


How many of those were likely to trigger in practice on an ATM VCC though?
If we are taking the Fixes: tag as a hint about which stable kernels we
might want to backport to, rather than a moral assignment of blame, then
14afee4b is probably not the worst place to point it. But I don't mind
much...

-- 
dwmw2