On Tue, Jun 2, 2015 at 8:03 AM, Ben Pfaff <[email protected]> wrote:
> On Tue, Jun 02, 2015 at 07:58:48AM -0700, Ben Pfaff wrote:
>> On Mon, Jun 01, 2015 at 03:35:23PM -0700, Jesse Gross wrote:
>> > On Mon, Jun 1, 2015 at 2:55 PM, Ben Pfaff <[email protected]> wrote:
>> > > On Mon, Jun 01, 2015 at 01:49:30PM -0700, Jesse Gross wrote:
>> > >> We have a special flow_metadata structure to represent the parts
>> > >> of a packet that aren't carried in the payload itself. This is
>> > >> used in the case where we need to send the packet as a Packet In
>> > >> to an OpenFlow controller. This is a subset of the more general
>> > >> struct flow.
>> > >>
>> > >> In practice, almost all operations we do on this structure involve
>> > >> converting it to or from a match or have code that is the same as
>> > >> a match. Serialization to NXM and back is done as a match. There
>> > >> is special flow_metadata formatting code that is almost identical
>> > >> to match formatting.
>> > >>
>> > >> The uses for struct flow_metadata aren't performance critical
>> > >> when it comes to memory, so we can save quite a bit of code by
>> > >> just using a match structure directly instead. In addition, as
>> > >> metadata increases and becomes more complex (Geneve options require
>> > >> some special handling beyond just additional fields), using the
>> > >> match structure means we only have to do this work in one place.
>> > >>
>> > >> Signed-off-by: Jesse Gross <[email protected]>
>> > >
>> > > Most of this rationale makes sense to me, except that I'm surprised
>> > > that struct flow_metadata (which doesn't have masks) gets converted to
>> > > struct match (which does).  Maybe I should just dig into the code, but
>> > > is there some big reason why this is the case?
>> >
>> > The main reason is just that struct match is the more general superset
>> > and the encode/decoding/formatting code operates on that to be able to
>> > handle everything so it's cleaner to just use a match even if it is
>> > slightly less memory efficient.
>>
>> It just seems conceptually wrong to me to have masks around in cases
>> where there's no concept of one, where the masks just take up space and
>> get ignored.  I'm trying to figure out how much that bothers me.
>
> So, I guess, the summary of my thoughts about this right now is that for
> me it's a no-brainer to switch from a specialized structure to struct
> flow, but struct match is a little harder for me conceptually.

I guess I should have been more clear: we actually are using the
wildcards to indicate which fields are significant. Effectively the
purpose of the conversion code between struct flow_metadata and struct
match is to synthesize these wildcards. There would be little benefit
to switching to struct flow instead because we would still need to do
the same operation.
_______________________________________________
dev mailing list
[email protected]
http://openvswitch.org/mailman/listinfo/dev

Reply via email to