[ECOS] __tcp_close() can't close connection in Redboot

[ECOS] __tcp_close() can't close connection in Redboot

[Huang Sun I]

[-- Warning: decoded text below may be mangled, UTF-8 assumed --]
[-- Attachment #1: Type: text/plain; charset=big5, Size: 1570 bytes --]

Hi,
  I am trying to build up httpd service in RedBoot.
And I met problem that I can't close connection by
calling __tcp_close() after replying data to
client,and the httpd sock state is always TIME_WAIT.
The client http can't get response from server after
recevive replying.I have to open another client http
to browser,but only once,problem is the same with
above
I described.

The following are rough code structure..
1> into listening state by calling __tcp_listen().

2>Checking TCP incoming connection
httpd_accept()
{
    if (!is_idle) return;
    if (!have_net) return;

    __tcp_poll();
    if (httpd_state != httpd_sock.state) {
        // Something has changed
        if (httpd_sock.state == _ESTABLISHED) {
            // A new connection has arrived
            //check Rx data and response
            // close connection by calling
__tcp_close() .
        }
        if (httpd_sock.state == _CLOSED) {
            // Get ready for another connection
            //into listening state by calling
__tcp_listen().
        }
    }
    httpd_state = httpd_sock.state;      
}
RedBoot_idle(httpd_accept, RedBoot_IDLE_HTTPD/* 6000
*/);


Do I miss any procedure or other way that could fix
it?


thank you!
Huang Sun I


___________________________________________________  ×îаæ Yahoo!ÆæĦ¼´•rͨӍ 7.0 beta£¬ÃâÙM¾W·ëŠÔ’ÈÎÄã´ò£¡  http://messenger.yahoo.com.tw/beta.html

-- 
Before posting, please read the FAQ: http://ecos.sourceware.org/fom/ecos
and search the list archive: http://ecos.sourceware.org/ml/ecos-discuss

[ECOS] Re: __tcp_close() can't close connection in Redboot

[Grant Edwards]

In gmane.os.ecos.general, you wrote:

>   I am trying to build up httpd service in RedBoot.
> And I met problem that I can't close connection by
> calling __tcp_close() after replying data to
> client,and the httpd sock state is always TIME_WAIT.

Right. 

After you close a tcp connection (you send a FIN packet and
receive a FIN back) the state is _supposed_ to be TIME_WAIT.

http://tangentsoft.net/wskfaq/articles/debugging-tcp.html

-- 
Grant Edwards                   grante             Yow!  Mr and Mrs PED, can
                                  at               I borrow 26.7
                               visi.com            

-- 
Before posting, please read the FAQ: http://ecos.sourceware.org/fom/ecos
and search the list archive: http://ecos.sourceware.org/ml/ecos-discuss

Re: [ECOS] Re: __tcp_close() can't close connection in Redboot

[Andrew Lunn]

On Wed, Oct 12, 2005 at 07:55:11AM -0500, Grant Edwards wrote:
> In gmane.os.ecos.general, you wrote:
> 
> >   I am trying to build up httpd service in RedBoot.
> > And I met problem that I can't close connection by
> > calling __tcp_close() after replying data to
> > client,and the httpd sock state is always TIME_WAIT.
> 
> Right. 
> 
> After you close a tcp connection (you send a FIN packet and
> receive a FIN back) the state is _supposed_ to be TIME_WAIT.
> 
> http://tangentsoft.net/wskfaq/articles/debugging-tcp.html

The purpose of this is to catch old packets still floating around the
network including retransmitted FIN packets.

Normally this is not a problem. The server goes back to its listen
socket and accepts the next connection.  After 2x maximum segment
lifetime the old socket in TIME_WAIT is destroyed. 

I don't know the redboot tcp stack. Does it implement the same
concept? ie simply accept the next connection waiting on the listen
socket?

        Andrew

-- 
Before posting, please read the FAQ: http://ecos.sourceware.org/fom/ecos
and search the list archive: http://ecos.sourceware.org/ml/ecos-discuss

Re: [ECOS] Re: __tcp_close() can't close connection in Redboot

[Gary Thomas]

On Wed, 2005-10-12 at 15:16 +0200, Andrew Lunn wrote:
> On Wed, Oct 12, 2005 at 07:55:11AM -0500, Grant Edwards wrote:
> > In gmane.os.ecos.general, you wrote:
> > 
> > >   I am trying to build up httpd service in RedBoot.
> > > And I met problem that I can't close connection by
> > > calling __tcp_close() after replying data to
> > > client,and the httpd sock state is always TIME_WAIT.
> > 
> > Right. 
> > 
> > After you close a tcp connection (you send a FIN packet and
> > receive a FIN back) the state is _supposed_ to be TIME_WAIT.
> > 
> > http://tangentsoft.net/wskfaq/articles/debugging-tcp.html
> 
> The purpose of this is to catch old packets still floating around the
> network including retransmitted FIN packets.
> 
> Normally this is not a problem. The server goes back to its listen
> socket and accepts the next connection.  After 2x maximum segment
> lifetime the old socket in TIME_WAIT is destroyed. 
> 
> I don't know the redboot tcp stack. Does it implement the same
> concept? ie simply accept the next connection waiting on the listen
> socket?

RedBoot hangs once it sees the CLOSE_WAIT condition.  You need to
continually call __tcp_poll() which will eventually flush the connection
which then goes to CLOSED.  At that time, you can create a new 
connection.

Note: this code is untested for more than one active TCP connection,
so beware.

-- 
------------------------------------------------------------
Gary Thomas                 |  Consulting for the
MLB Associates              |    Embedded world
------------------------------------------------------------


-- 
Before posting, please read the FAQ: http://ecos.sourceware.org/fom/ecos
and search the list archive: http://ecos.sourceware.org/ml/ecos-discuss

[ECOS] Re: __tcp_close() can't close connection in Redboot

[Grant Edwards]

In gmane.os.ecos.general, you wrote:
> On Wed, Oct 12, 2005 at 07:55:11AM -0500, Grant Edwards wrote:
>> In gmane.os.ecos.general, you wrote:
>> 
>> >   I am trying to build up httpd service in RedBoot.
>> > And I met problem that I can't close connection by
>> > calling __tcp_close() after replying data to
>> > client,and the httpd sock state is always TIME_WAIT.
>> 
>> Right. 
>> 
>> After you close a tcp connection (you send a FIN packet and
>> receive a FIN back) the state is _supposed_ to be TIME_WAIT.
>> 
>> http://tangentsoft.net/wskfaq/articles/debugging-tcp.html
>
> The purpose of this is to catch old packets still floating around the
> network including retransmitted FIN packets.
>
> Normally this is not a problem. The server goes back to its listen
> socket and accepts the next connection.  After 2x maximum segment
> lifetime the old socket in TIME_WAIT is destroyed. 

Right.  You can reduct that period from 120 seconds to 1 second
by setting the reuse option on the socket.

> I don't know the redboot tcp stack. Does it implement the same
> concept? ie simply accept the next connection waiting on the listen
> socket?

After the 2MSL TIME_WAIT is done, you need to call
__tcp_listen() on the socket to make it listen for a new
connection (accept a SYN packet).  There is no such thing in
RedBoot as a "listen socket" on which pending connections wait.

IOW, you use a single socket structure to both listen for the
connection and communicate via that connection.

-- 
Grant Edwards                   grante             Yow!  FUN is never having
                                  at               to say you're SUSHI!!
                               visi.com            

-- 
Before posting, please read the FAQ: http://ecos.sourceware.org/fom/ecos
and search the list archive: http://ecos.sourceware.org/ml/ecos-discuss

[ECOS] Re: __tcp_close() can't close connection in Redboot

[Grant Edwards]

In gmane.os.ecos.general, you wrote:

> RedBoot hangs once it sees the CLOSE_WAIT condition.

I've never seen that happen.

> You need to continually call __tcp_poll() which will
> eventually flush the connection which then goes to CLOSED.

You have to continually call __tcp_poll() no matter what state
the socket is, since RedBoot is single-threaded.

> At that time, you can create a new connection.
>
> Note: this code is untested for more than one active TCP connection,
> so beware.

It worked OK for me. :)

-- 
Grant Edwards                   grante             Yow!  My forehead feels
                                  at               like a PACKAGE of moist
                               visi.com            CRANBERRIES in a remote
                                                   FRENCH OUTPOST!!

-- 
Before posting, please read the FAQ: http://ecos.sourceware.org/fom/ecos
and search the list archive: http://ecos.sourceware.org/ml/ecos-discuss

Re: [ECOS] Re: __tcp_close() can't close connection in Redboot

[Gary Thomas]

On Wed, 2005-10-12 at 08:34 -0500, Grant Edwards wrote:
> In gmane.os.ecos.general, you wrote:
> 
> > RedBoot hangs once it sees the CLOSE_WAIT condition.
> 
> I've never seen that happen.

You won't - it just won't look for another 'telnet' session
when the current one closes until it's completely closed.

> 
> > You need to continually call __tcp_poll() which will
> > eventually flush the connection which then goes to CLOSED.
> 
> You have to continually call __tcp_poll() no matter what state
> the socket is, since RedBoot is single-threaded.
> 
> > At that time, you can create a new connection.
> >
> > Note: this code is untested for more than one active TCP connection,
> > so beware.
> 
> It worked OK for me. :)

Nice to know.

-- 
------------------------------------------------------------
Gary Thomas                 |  Consulting for the
MLB Associates              |    Embedded world
------------------------------------------------------------


-- 
Before posting, please read the FAQ: http://ecos.sourceware.org/fom/ecos
and search the list archive: http://ecos.sourceware.org/ml/ecos-discuss

Re: [ECOS] Re: __tcp_close() can't close connection in Redboot

[Huang Sun I]

[-- Warning: decoded text below may be mangled, UTF-8 assumed --]
[-- Attachment #1: Type: text/plain; charset=big5, Size: 1931 bytes --]

Hi,
   thanks for replying.
After __tcp_close(),I keep calling __tcp_poll() until
state is _CLOSED. But it seems endless,even setup
reuse option to 1sec in sock,the state is still
TIME_WAIT..,

Have you met this situation before?
Could force to destory the state to _CLOSED?
If so,is there any potential problem?

thank you
Sam

 
--- Gary Thomas <gary@mlbassoc.com> »¡¡G

> On Wed, 2005-10-12 at 08:34 -0500, Grant Edwards
> wrote:
> > In gmane.os.ecos.general, you wrote:
> > 
> > > RedBoot hangs once it sees the CLOSE_WAIT
> condition.
> > 
> > I've never seen that happen.
> 
> You won't - it just won't look for another 'telnet'
> session
> when the current one closes until it's completely
> closed.
> 
> > 
> > > You need to continually call __tcp_poll() which
> will
> > > eventually flush the connection which then goes
> to CLOSED.
> > 
> > You have to continually call __tcp_poll() no
> matter what state
> > the socket is, since RedBoot is single-threaded.
> > 
> > > At that time, you can create a new connection.
> > >
> > > Note: this code is untested for more than one
> active TCP connection,
> > > so beware.
> > 
> > It worked OK for me. :)
> 
> Nice to know.
> 
> -- 
>
------------------------------------------------------------
> Gary Thomas                 |  Consulting for the
> MLB Associates              |    Embedded world
>
------------------------------------------------------------
> 
> 
> -- 
> Before posting, please read the FAQ:
> http://ecos.sourceware.org/fom/ecos
> and search the list archive:
> http://ecos.sourceware.org/ml/ecos-discuss
> 
> 


___________________________________________________  ×îаæ Yahoo!ÆæĦ¼´•rͨӍ 7.0 beta£¬ÃâÙM¾W·ëŠÔ’ÈÎÄã´ò£¡  http://messenger.yahoo.com.tw/beta.html

-- 
Before posting, please read the FAQ: http://ecos.sourceware.org/fom/ecos
and search the list archive: http://ecos.sourceware.org/ml/ecos-discuss