[flow-tools] Re: persistent bug in flow-stat.c for format23

[Mark Fullmer]

Thanks for the fix.  The bug got introduced recently while working towards
flow-tools internals being less Cisco NetFlow centric.

The flow-capture memory leak should be fixed in 0.58.

mark

On Tue, Jun 18, 2002 at 09:32:50PM +0400, Y.Lachin@transtk.ru wrote:
> Hi, Mark
> 
> First, thanks for the great and flexible flow-tools :-). Its nice to see
> the "Right Thing"  after looking at buggy commercial monsters.
> 
> I beleive there is a small bug  I found in recent versions of flow-tools
> (at least, starting from v5.6, didn't  look at older versions).
> I expected it to be fixed in v.5.8, but it's still there.
> 
> In function:
> 
> int format23(struct fmtargs *args)
> 
> lines
> 
>     ftch_recif2.input = *((u_char*)(rec+fo.input));
>     ftch_recif2.output = *((u_char*)(rec+fo.output));
> 
> should obviously be changed to
> 
>     ftch_recif2.input = *((u_int16*)(rec+fo.input));
>     ftch_recif2.output = *((u_int16*)(rec+fo.output));
> 
> otherwise it doesn't work correctly. There might be similar pointer type
> casting errors in other places, too, but I didn't check.
> 
> I noticed and tried this on Sparc Solaris
> (gcc-2.95.3, SunOS kestrel 5.8 Generic_108528-11 sun4u sparc
> SUNW,UltraAX-i2).
> 
> Strange enough, that nobody reported it before. Looks like it doesn't show
> up under some circumstances (small number of interfaces?).
> Can't check it now, but as far as I remember,  I haven't seen it
> (half-year ago) when running flow-stat  on x86 Linux machine?!  (it was
> version 5.5 or 5.6).
> 
> 
> Regards,
>  Yuri
> 
> PS: I think that there is a possible  memory leak in flow-capture code:
> after running flow-capture  non-stop for several days I saw its size in
> memory
> grown 10-20 times. Don't have time now to find it.
> 
> --
> Yuri Lachin   Company TransTeleCom <y.lachin@transtk.ru>
> +7(095)784-6670 ext.7280
> 
>