[ovs-discuss] how nw_src_mask and nw_dst_mask are generated in ovs?

Tahir Rauf tahir.rauf1 at gmail.com
Sat Aug 20 14:19:29 PDT 2011


Thanks Ben for your prompt answer. Your answer eliminates the confusion.

On Sun, Aug 21, 2011 at 1:57 AM, Ben Pfaff <blp at nicira.com> wrote:

> On Sat, Aug 20, 2011 at 11:12:10PM +0500, Tahir Rauf wrote:
> > I got a little question regarding wildcards of nw_src address. I have
> > written a controller application which sets the wildcards as following
> >
> > 1. ofm->match.wildcards = htonl((((OFPFW_ALL & ~ OFPFW_DL_SRC) & ~
> > OFPFW_DL_DST)  & ~ OFPFW_DL_TYPE) & ~ OFPFW_NW_SRC_ALL);
> > The above statement yields wildcards    0000 0000 0011 1111 11*01 1111*
> 1110
> > 0011.
> >
> > 2. The bits in 'bold' represents the nw_src_mask. The 0 at extreme left
> > tells that not every bit of nw_src is wildcarded. (if that bit is '1', it
> > will say to wildcard each and every bit of nw_src. right?)
>
> Yes.
>
> > 3. The value of rest of bits represents that how many LSBs should be
> > wildcarded. In our example, it tells that least 31 bits should be
> > wildcarded. O.K?
>
> Not "at least" but exactly 31 bits.
>
> > An IP netmask is generated with a 1 in each bit that must match and a
> > 0 in each bit that is wildcarded.  I am expecting that OVS should
> > generate the value of nw_src_mask = 1111 1111 1111 1111 1111 1111 1111
> > 1111.
>
> I don't see how that wildcards any bits at all, let alone 31 bits.
>
> > (On the other hand, the value of nw_dst_mask = 0000 0000 0000 0000
> > 0000 0000 0000 0000).  For nw_dst_mask, the results are according to
> > expectations. But this is not the case for nw_src_mask. OVS is
> > generating the nw_src_mask = 0x80000000.
>
> 0x80000000 specifies that the highest bit must match and that the other
> 31 bits are wildcarded.  Isn't that what you are asking for?
>



-- 
Tahir Rauf
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://openvswitch.org/pipermail/discuss/attachments/20110821/5ad44acf/attachment-0001.htm>


More information about the discuss mailing list