On Fri, Dec 07, 2012 at 05:07:24PM +0100, walter harms wrote:

Am 07.12.2012 12:10, schrieb Dan Carpenter:

quoted

We pass IFLA_BRPORT_MAX to nla_parse_nested() so we need
IFLA_BRPORT_MAX + 1 elements.  Also Smatch complains that we read past
the end of the array when in br_set_port_flag() when it's called with
IFLA_BRPORT_FAST_LEAVE.

I have no clue why nla_parse_nested() need IFLA_BRPORT_MAX elements.
but the majory of loop look like
for(i=0;i<max;++)
most programmers will think this way.
So it seems the place to fix is nla_parse_nested().
doing not so is asking for trouble (in the long run).
At least this function needs a big warning label that (max-1)
is actually needed.

Yeah, nla_parse_nested() is actually documented already.

regards,
dan carpenter