On Wed, 27 Sep 2000, Eric Haszlakiewicz wrote:

> On Wed, Sep 27, 2000 at 05:31:28PM -0700, Andrei Petrov wrote:
> > Thank you for help, yep kernel assumes and use longword alignment,
> > and arp doesn't. Seems that ROUNDUP is popular in net*.
> > I submitted patch for this into problem report.
> 	just about what I have.  In addition to using ROUNDUP I am also
> going to use the length field instead of sizeof since one of those addresses
> is actually only 8 bytes long.  This will also allow better bounds checking
> in the kernel.
> 	The kernel will now return EINVAL unless it finds exactly the number of
> addresses you say you're giving it and the amount of data you pass is exactly
> the same as the sum of the sa_len fields.
> 	Could you please update to arp.c:1.25 and rtsock.c:1.42 and make sure
> I didn't break anything?
> 
What worked before works as before. The test is around publishing
entry. New arp works. 
Old arp rejected with EINVAL, that's good; my modified is rejected and I
known why.

Thank you,
	Andrey