netdev - Re: [PATCH] usbnet: ipheth: fix potential recvmsg bug and recvmsg bug 2

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] [day] [month] [year] [list]

Message-Id: <20181127.145950.2178672310020332163.davem@davemloft.net>
Date:   Tue, 27 Nov 2018 14:59:50 -0800 (PST)
From:   David Miller <davem@...emloft.net>
To:     3erndeckstein@...il.com
Cc:     linux-usb@...r.kernel.org, netdev@...r.kernel.org,
        linux-kernel@...r.kernel.org, 3ernd.Eckstein@...il.com,
        Oliver.Zweigle@...o.com
Subject: Re: [PATCH] usbnet: ipheth: fix potential recvmsg bug and recvmsg
 bug 2

From: Bernd Eckstein <3erndeckstein@...il.com>
Date: Fri, 23 Nov 2018 13:51:26 +0100

> The bug is not easily reproducable, as it may occur very infrequently
> (we had machines with 20minutes heavy downloading before it occurred)
> However, on a virual machine (VMWare on Windows 10 host) it occurred
> pretty frequently (1-2 seconds after a speedtest was started)
> 
> dev->tx_skb mab be freed via dev_kfree_skb_irq on a callback
> before it is set.
> 
> This causes the following problems:
> - double free of the skb or potential memory leak
> - in dmesg: 'recvmsg bug' and 'recvmsg bug 2' and eventually
>   general protection fault
> 
> Example dmesg output:
 ...
> The proposed patch eliminates a potential racing condition.
> Before, usb_submit_urb was called and _after_ that, the skb was attached
> (dev->tx_skb). So, on a callback it was possible, however unlikely that the
> skb was freed before it was set. That way (because dev->tx_skb was not set
> to NULL after it was freed), it could happen that a skb from a earlier
> transmission was freed a second time (and the skb we should have freed did
> not get freed at all)
> 
> Now we free the skb directly in ipheth_tx(). It is not passed to the
> callback anymore, eliminating the posibility of a double free of the same
> skb. Depending on the retval of usb_submit_urb() we use dev_kfree_skb_any()
> respectively dev_consume_skb_any() to free the skb.
> 
> Signed-off-by: Oliver Zweigle <Oliver.Zweigle@...o.com>
> Signed-off-by: Bernd Eckstein <3ernd.Eckstein@...il.com>

Applied and queued up for -stable, thanks.