On 2017年02月27日 12:09, Hailiang Zhang wrote:
On 2017/2/27 11:40, Jason Wang wrote:
On 2017年02月27日 11:11, Hailiang Zhang wrote:
On 2017/2/23 12:16, Jason Wang wrote:
On 2017年02月22日 16:51, Hailiang Zhang wrote:
On 2017/2/22 16:45, Hailiang Zhang wrote:
On 2017/2/22 16:07, Jason Wang wrote:
On 2017年02月22日 11:46, zhanghailiang wrote:
After a net connection is closed, we didn't clear its releated
resources
in connection_track_table, which will lead to memory leak.
Not a real leak but would lead reset of hash table if too many
closed
connections.
Yes, you are right, there will be lots of stale connection data in
hash table
if we don't remove it while it is been closed. Which
Ok, so let's come up with a better title of the patch.
OK.
Let't track the state of net connection, if it is closed, its
related
resources will be cleared up.
The issue is the state were tracked partially, do we need a full
state
machine here?
Not, IMHO, we only care about the last state of it, because, we will
do nothing
even if we track the intermedial states.
Well, you care at least syn state too. Without a complete state
machine,
it's very hard to track even partial state I believe. And you will
fail
to track some state transition for sure which makes the code fragile.
Agree, but here things are a little different. There are some extreme
cases
that we may can't track the complete process of closing connection.
For example (I have explained that in the bellow, it seems that you
didn't
got it ;) ).
If VM is running before we want to make it goes into COLO FT state,
there maybe some connections exist already, in extreme case, VM is
going into
COLO state while some connections are in half closing state, we can
only track
the bellow half closing state in filter-rewriter and colo compare
object.
[...]
Er, here we track the last two states 'FIN=1, ACK=1' and 'ACK=1' (
which asks
the 'FIN=1,ACK=1' packet, We will remove the connection while got the
'ACK=1'
packet, so is it enough ?
But the connection is not closed in fact, no? It's legal for remote to
continue sending tons of packet to us even after this.
Er, I'm a little confused, Here, for server side,
i think after got the 'ACK=1,seq=u+1,ack=w+1', it is closed,
so i remove it from hash table, wrong ?
Client: Server:
ESTABLISHED| |
| -> FIN=1,seq=u -> |
This is case A and ACK should be set in this segment too.
FIN_WAIT_1 | |
| <- ACK=1,seq=v,ack=u+1 <- |
FINA_WAIT_2| |CLOSE_WAIT
| <- FIN=1,ACK=1,seq=w,ack=u+1<-|
| |LAST+ACK
This is case B.
| -> ACK=1,seq=u+1,ack=w+1 |
TIME_WAIT | |CLOSED
CLOSED | |
I think the issue is that your code can not differ A from B.