►
From YouTube: DASH Behavioral Model WG Feb 2 2023
Description
WIP snapshot of something that compiles with the P4 DPDK compiler
https://github.com/sonic-net/DASH/pull/324
Thank you Andy!
Add Private Link mapping by marian-pritsak ยท https://github.com/sonic-net/DASH/pull/327
Discuss: Add/Transition from bmv2 to P4 DPDK target
AI: Andy will try to run maybe next week or so to see if it executes (check for gaps)
Should not be a mess of IF/DEFs
Chris created a docker file w/DPDK backend for Andy, can compile this version of code w/every version of CI
Marian: execute on real DPDK BE software
Add-on-miss available in PNA, implemented in DPDK, there is an active development community there. We want to make sure there is no degradation if we switch to different BE (do not break out party hats yet folks)
https://github.com/sonic-net/DASH/pull/331
Correct syntax errors in PNA_CONNTRACK version of code - Andy
These had IF/DEF PNA conntrack around them
Q: range port list and list port list (in bmv2) public?
A: Marian: simplified version s/be on par
unroll the range by looping through it
Saad: open contrail issue in P4; need proven xn tracking implementation
paste code into new work item
Kristina opened: https://github.com/orgs/Azure/projects/129/views/6?pane=issue&itemId=19631774
A
Are
coming
to
dash
behavioral
model
meeting
for
February,
2nd
I'll,
go
ahead
and
turn
off
my
camera
and
let
everyone
look
through
the
notes
from
last
time.
So
just
briefly
went
over
what
we
talked
about
last
time
and
I
was
wondering
today.
If
we
are
going
to
talk
about
the
the
item
we've
been
wanting
to
talk
about
for
a
couple
of
weeks,
the
bmv2
for
service
tunnel
private
link
from
maybe
either
Prince
or
Marion.
B
I
guess
bmv2
will
still
need
another
week
because
we
don't
get
to
reviewing
it
together
with
Prince
and
since
we're
doing
this
together,
both
of
us
on
different
sides
of
that
feature.
The
pull
request
itself
is
available
for
preview.
If
anyone
wants
to
look
at
it,
but
I
want
to
make
sure
that
we
figure
out
everything
in
between
us
before
presenting
it
in
the
community.
B
C
News
so
I
created
a
couple
of
pull
requests
since
last
week.
Not
one
of
them
is
not
in
a
mergeable
form.
At
the
moment
it
was
sort
of
just
put
out
there
as
a
demonstration
of
what's
possible.
Okay,
that
is,
if
you
want
to
know
the
number.
It's
sorry
looking
really
quickly
here-
number
324
on
the
repo,
so
that
was
just
a
sort
of
a
work
in
progress
snapshot
of
something
that
compiles
with
the
dpdk
compiler.
C
There
are
a
couple
of
small
gaps
and
it's
P4
Implement
p,
a
implementation
that
don't
support
a
couple
minor
features,
but
except
for
that,
it
looks
like
everything
compiles
now
note:
I
haven't
tried
it
on
the
dpdk
backend
running
packets
through
it,
so
there
could
be
future
additional
gaps
that
I
haven't
identified.
Yet
that
would
be
revealed
by
doing
so.
Okay.
E
F
I
think
there
is
one
other
thing
I
I
also
saw
reshma
is
that
you
know,
for
the
benefit
of
others.
Is
that
you
know
we
have
moved
from
V1
model
architecture
to
PNA
right
right
as
part
of
this
one
as
well.
So
I
think
there
are
two
things
that
we
should
probably
bring
the
community
up
to
speed
on.
That
is
that
the
transition
from
V1
model
to
PNA
and
a
translation
from
BMV
to
to
P
Force
P4
DPT
Target
right
right.
So
this
is
yeah.
Okay,
yeah
go
ahead.
Please
thanks.
C
So
Chris
Summers
took
a
look
at
it.
Several
other
people
commented
on
it.
I
am
as
as
I
didn't
quite
get
it
done
before
this
meeting,
but
there
was
a
couple
people
that
commented.
Should
these
be,
you
know,
should
we
merge
these
into
the
existing
files
using
IF
deaths?
Should
we
put
these
in
separate
files?
There
seem
to
be
maybe
a
slightly
more
people.
C
F
F
D
C
So
I
haven't
run
it
yet
I
will
find
out
and
give
instructions
to
the
group
to
run
it
using
open
source
when
I
have
them
myself,
but
deeply
there.
There
is
open
source
implementation
of
the
dpd
backend
for
the
P4
compiler
that
exists.
That
I
just
haven't
tried
myself
yet
for
running
packets
through
it.
D
G
So
I
think
there's
a
few
things
here.
You
know
we
can.
One
debate
is
whether
we
merge
the
code
or
not
and
I
lean
towards
merging
it,
but
it
depends
on
what
it
really
looks
like
you
know,
if
it's
it
doesn't
sound
like
it'll,
be
a
god-awful
mess
of
if
deaths
no
worse
than
what's
already
there,
and
it's
not
a
not
a
criticism.
It's
just.
The
stuff
happens
like
that.
So
there's
the
compilation
of
it
and
I
I
am
I,
came
up
with
a
Docker
file
and
made
it
available
to
Andy.
G
G
D
G
G
It'll
just
be
a
bona
fide
build
Target
within
the
dash
repo,
and
we
can
compile
the
code,
make
sure
it
passes
every
time
and
and
so
there's
work
on
Andy's
part
to
try
that
out,
right
and
kind
of
take
the
next
step,
but
I
can
help
formalize
it
and
then
the
next
part
which
Marion
just
alluded
to
is
executing
it
on
a
real
dpdk
back
end,
which
would
be
a
software.
So
that
would
be
a
follow-on
project
and
you
know
if
people
can
kind
of
pave
the
way.
G
I
can
also
help
put
that
into
the
dash
workflow
so
that
we're
running
cicd
tests
on
the
dpdk
target,
assuming
we
we
can
get
there
and
and
then
we'll
have
a
new
build
Target
and
a
new
execution
engine.
You
know,
subject
to
whatever
limitations
it
has
so
I
think
that's.
This
was
really
encouraging,
but
we'll
see
if
it
executes
first
right,
I
don't
want
to
have
a
party
yet
yeah.
B
D
F
F
F
C
F
C
G
And
we
got
the
code
on
the
gurney
right
now.
This
is
getting
weird
Andy.
Can
you
talk
about
PR
331
a
little
bit
because
I
made
another
comment
on
that
just
a
while
ago
is
relevant.
I
think
that
is
you're
correcting
impact
errors,
which
is
good,
I,
think
it
brings
up
two
things.
One
is:
how
did
you
expose
these
errors,
and
should
we
be-
and
assuming
you
know,
your
fix
is
good.
How
do
we
also
make
a
let's
say
a
CI
screen
test
this
with
every
build.
G
So
will
we
inherently
if
we
start
enabling
this
dpdk
activist
build
Target
as
a
CI
action,
we'll
also
be
checking
this
every
time
as
well
right,
correct,
it'll
be
intrinsic,
so
we
don't
need
yet
another
CI
pipeline
to
address
this.
It's
it's
going
to
be
part
and
parcel
of
the
other
PR
correct.
Okay,.
G
G
C
A
C
A
F
B
Yeah,
so
we
did
with
Microsoft
core
integration.
We
did
the
simplified
version
kind
of
like
what
Andy
mentioned
before
it's
already
available
inside
the
implant,
or
should
be
I
actually
didn't
check,
because
it's
not
me
who
was
doing
that.
But
what
and
the
mentioned
just
having
this
implemented
in
the
site
Library
instead
of
having
the
having
reloaded
lists
in
in
the
back
end,
essentially.
B
A
E
Yes,
hello,
okay,
I
wanted
to
ask
a
very
quick
question.
If
you
don't
mind
it's
about
connection
tracking
in
the
data
path,
so
obviously
the
data
path
needs
to
keep
track
of
the
connection
States.
Now.
My
question
is:
that
is
the
dpu
expected
to
implement
all
the
states?
All
the,
for
example,
TCP
State
machine,
the
entire
TCP
State
machine,
or
only
a
subset
of
tcpd
State
states
are
required.
E
H
A
H
H
A
D
A
A
A
H
Right
because,
like
you,
might
want
to
have
a
different
timer
like
based
on
like
this,
like
the
opening
of
a
connection,
then
like
a
different
timer
when
you're
in
the
established
state
and
then
like
a
different
timer
when
you're
closing
the
connection,
and
so
like
you
know
this,
like
the
states
that
this
maintains
like
would
allow
for
such
you
know,
ability
to
like
yeah.
This
is
this
is
the
this?
Is
the
code
from
open
control
if
you
scroll
down
a
little
bit
more
like.
A
A
C
H
I
think
the
idea
was
and
I
think
Mario
is
not
on
the
call,
but
he's
on
this
email
I
think
the
idea
was
that
this
logic
would
be
implemented
in
P4
as
part
of
the
behavioral
model,
and
then
once
it
was
implemented
in
P4
as
part
of
the
behavioral
model.
That
would
be
the
reference
for
connection
tracking
and
so
but
I
think,
like
it,
I
think
Mario
never
I
actually
got
to
an
implementation
in
P4.
F
A
A
C
A
C
A
H
H
And
I
was
a
little
more
than
that,
but
it's
it's
pretty
understandable
and
I
think
the
reason
why,
like
we
proposed
to
like
this
as
being
the
logic,
is
because
this
is
like
a
proven
connection.
Tracking
implementation
I
mean
Linux
has
has
contract
I
mean
there
are
other
connection
tracking
implementations
out
there,
but
this
one
looked
pretty
straightforward.
A
Okay,
all
right,
okay,
cool,
so
it's
11..
Do
you
guys
want
to
stop
for
the
day
and
then
next
week?
Maybe
we
can
everyone
take
a
look
at
the
service
tunnel.
Private
link,
PR's
I'll,
send
notes
with
the
links
to
them
and
then,
if
we
get
time
to
go
ahead
and
try
Andy's
experiment
great,
you
know
if
we
don't
we'll
keep
working
on
it.
Does
that
sound
good.