netdev - Re: [PATCH net-next v3 3/4] net: use skb->priority for overloading skb->dropcount and skb->reserved

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]

Date:	Tue, 24 Feb 2015 16:07:07 +0200
From:	Eyal Birger <eyal.birger@...il.com>
To:	Eric Dumazet <eric.dumazet@...il.com>
Cc:	David Miller <davem@...emloft.net>,
	Shmulik Ladkani <shmulik.ladkani@...il.com>,
	Eric Dumazet <edumazet@...gle.com>, netdev@...r.kernel.org
Subject: Re: [PATCH net-next v3 3/4] net: use skb->priority for overloading
 skb->dropcount and skb->reserved_tailroom instead of skb->mark

Hi,

On Tue, Feb 24, 2015 at 3:18 PM, Eric Dumazet <eric.dumazet@...il.com> wrote:
> On Tue, 2015-02-24 at 12:10 +0200, Eyal Birger wrote:
>
>>
>> Well, gave it a shot... it looks like several protocol families
>> (packet, rxrpc, bluetooth) do not
>> have room in skb->cb[] for the dropcount - at least on my 64 bit machine.
>
> No idea how you took a look ?
>
> sizeof(struct packet_skb_cb) == 24  : We have plenty of room ?
>

Well, I took a look, then I tried :)

It breaks in packet_rcv() in a BUILD_BUG_ON() assertion. The asserted size is:

sizeof(*PACKET_SKB_CB(skb)) + MAX_ADDR_LEN - 8

This was introduced in ffbc61117d32dc4e768 ("[PACKET]: Fix skb->cb
clobbering between aux and sockaddr")
It requires the ability to store the maximal possible address length (32).

>
> bluetooth : Whole struct rxrpc_skb_priv is not used when packet is
> stored in receive queue.
>
> We only need bt_cb(skb)->psm & bt_cb(skb)->bdaddr according to
> l2cap_skb_msg_name()
>
> An union will be possible.
>

I can look into that.

> rxpc is buggy right now anyway, as it reads skb->mark _and_ uses
> sock_recv_ts_and_drops(), so skb->mark value is pretty much void.
>

Yes, but it does not seem to use sock_queue_rcv_skb() which sets skb->dropcount.
BTW, aliasing priority instead of mark would allow someone to refactor rxrpc to
use sock_queue_rcv_skb() if desired.

> Note that resend_at field could probably converted into u32. tcp stack
> uses same (u32)jiffies trick (tcp_time_stamp)
>
>
> I never said it was going to be easy ;)
>

I don't mind :) though I wonder whether the added complexity in
partitioning skb->cb[]
between the protocol families and the socket layer is worth it.
--
To unsubscribe from this list: send the line "unsubscribe netdev" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html