Re: Memory leak in Libnet

Hi Frederic,

"AL" is a structure of the following type:

struct libnet_ifaddr_list
{
    u_int32_t addr;
    int8_t *device;
};

So although "al" itself is static variable, its field "device"
points to dynamically allocated memory (in this case, allocated
by strdup()). Thus the old memory allocated for al->device need
to be freed before new memory is allocated for al->device.

In fact, parasoft's insure++ reports a memory leak at this strdup()
which is where I come from.

I did the following patch and it seems the "memory leak" is gone:

[libnet_if_addr.c]
     238
 +               if (al->device) free(al->device);
     239         if ((al->device = strdup(device)) == NULL)
     240         {
     241             snprintf(errbuf, LIBNET_ERRBUF_SIZE,
     242                     "%s(): strdup not enough memory\n", __func__);
     243             return(-1);
     244         }

Thanks

Kevin

--- Frédéric Raynal <[email protected]> wrote:
> Hello
> 
> On Mon, Apr 05, 2004 at 06:08:12PM -0700, [email protected] wrote:
> > Hi Mike,
> > 
> > It seems the following "strdup" is a memory leak
> > 
> > libnet-1.1.2:
> > 
> > [libnet_if_addr.c]
> >     238
> >     239         if ((al->device = strdup(device)) == NULL)
> >     240         {
> >     241             snprintf(errbuf, LIBNET_ERRBUF_SIZE,
> >     242                     "%s(): strdup not enough memory\n", __func__);
> >     243             return(-1);
> >     244         }
> > 
> > 
> > Should we do a "free" before each strdup() except when al->device 
> > is NULL ?
> > 
> 
> I dont think so : al is a pointer on a static array of structures.
> It can easily be checked with valgrind for instance ... but as I am on
> vacation and connected on a Mac OS X through a 56k line, I'll let that
> to the reader as an exercise ;-) 
> 
> 
> 	Fred Raynal
>