[ECOS] Problems with sockets after 1132 requests

[ECOS] Problems with sockets after 1132 requests

[Michael Sig Birkmose]

Hi List,

I have some problems handling many requests to a tcp server in eCos 2.0.

I am using the code found in net/common/v2_0/tests/server_test.c (I don't
modify it at all)

I have no problems connecting to the server on that port. However, when I
continuosly send tcp requests to the eCos tcp server, reads the response,
and then closes the socket, the server always stops responding after 1132
requests has been handled.

Furthermore I tried running gdb remote and debugging the server. Then it
stop responding after 15 requests, and I was not able to figure out why.

Does anyone have an idea of what could causing this?

The hardware platform is i386, and I am using redboot.

-- 
Michael Birkmose

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

Re: [ECOS] Problems with sockets after 1132 requests

[Andrew Lunn]

On Tue, Dec 02, 2003 at 01:18:02PM +0100, Michael Sig Birkmose wrote:
> Hi List,
> 
> I have some problems handling many requests to a tcp server in eCos 2.0.
> 
> I am using the code found in net/common/v2_0/tests/server_test.c (I don't
> modify it at all)
> 
> I have no problems connecting to the server on that port. However, when I
> continuosly send tcp requests to the eCos tcp server, reads the response,
> and then closes the socket, the server always stops responding after 1132
> requests has been handled.

What are you using for a client?

I've got a test running at the moment using the synth target and a
little script:

while true ; do cat /dev/zero | netcat 192.168.10.10 7734  ; done

Its run for other 20000 connections without a problem. 

    Andrew



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

Re: [ECOS] Problems with sockets after 1132 requests

[sandeep]

Hi Michael,
Your mail vaguely reminds of something similar we had observed here (perhaps
more than an year back) on a different platform and in slightly different
context. It could be possible mbuff leakage problem (perhaps mbuffs not getting
freed). Here is what you can try (I am not a networks expert)

- enabling "Network MBUF diagnostics"
- play around with the "Memory designated for networking buffers" to see if
numbers change from 1132 / 15
- increasing/decreasing the value of "Max number of open sockets" and see if
that bears any relations to the figures you observe

peace
sandeep

> I have some problems handling many requests to a tcp server in eCos 2.0.
>
> I am using the code found in net/common/v2_0/tests/server_test.c (I don't
> modify it at all)
>
> I have no problems connecting to the server on that port. However, when I
> continuosly send tcp requests to the eCos tcp server, reads the response,
> and then closes the socket, the server always stops responding after 1132
> requests has been handled.
>
> Furthermore I tried running gdb remote and debugging the server. Then it
> stop responding after 15 requests, and I was not able to figure out why.



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

Re: [ECOS] Problems with sockets after 1132 requests

[Andrew Lunn]

On Tue, Dec 02, 2003 at 06:36:34PM +0530, sandeep wrote:
> Hi Michael,
> Your mail vaguely reminds of something similar we had observed here (perhaps
> more than an year back) on a different platform and in slightly different
> context. It could be possible mbuff leakage problem (perhaps mbuffs not getting
> freed). Here is what you can try (I am not a networks expert)
> 
> - enabling "Network MBUF diagnostics"
> - play around with the "Memory designated for networking buffers" to see if
> numbers change from 1132 / 15
> - increasing/decreasing the value of "Max number of open sockets" and see if
> that bears any relations to the figures you observe

A simpler test would be to ping the board while running the server
test. If it continues to ping after the connections stop, its not an
MBUF problem.

     Andrew

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