Re: econet iovec

[Dan Rosenberg <dan.j.rosenberg () gmail com>]

Yes, this size calculation can overflow, but there's no negative
effect, since it is only used to construct a UDP packet, and UDP is
not susceptible to overflow issues in its sendmsg() path.

On the other hand, the check on line 331 to put an upper bound on the
total size can overflow, causing an underallocation on line 344 and a
kernel panic on subsequent usage due to bad skbuff alignment.  This
only affects people using actual native Econet hardware.  This was
already fixed by recently added checks in iovec size calculations and
in the sendto() path for maximum packet size.

-Dan

On Sun, Nov 14, 2010 at 9:56 AM, Thomas Pollet <thomas.pollet () gmail com> wrote:
> Hi,
>
> the AF_ECONET sendmsg iovec code also appears to be vulnerable to an integer
> overflow that will be fixed by the verify_iovec changes in the 2.6.37
> kernel.
> on line 469: size += iov_len
>
> Regards,
> Thomas