decreasing MAXSEG (MSS)

Hello:

I have a question on setting MSS (MAXSEG option):

--[client-program....code-snip]
---------------------------------------------------------------------------------

sockfd = socket(AF_INET, SOCK_STREAM, 0);
maxseg = 0;
maxseg_len = sizeof (maxseg);
getsockopt(sockfd, IPPROTO_TCP, TCP_MAXSEG, &maxseg, &maxseg_len);

maxseg = 1024;
setsockopt(sockfd, IPPROTO_TCP, TCP_MAXSEG, &maxseg, maxseg_len)

connect (.....)
.....
--------------------------------------------------------------------------------------------------------

Using TCPDUMP, I notice that the MSS value sent by my client m/c (with
above code) is still 1460...
(not *1024*, as I would expect).

Does anyone know if there is something else to be done to decrease the
MSS value

I have tried this on Linux 2.6.5-1.358 and Linux 2.4.20.

Thanks in advance,
Vishal.

vishal...@gmail.com <(E-Mail Removed)> wrote:
> maxseg = 0;
> maxseg_len = sizeof (maxseg);
> getsockopt(sockfd, IPPROTO_TCP, TCP_MAXSEG, &maxseg, &maxseg_len);

> maxseg = 1024;
> setsockopt(sockfd, IPPROTO_TCP, TCP_MAXSEG, &maxseg, maxseg_len)

I know it is a pedantic question, but are you sure the setsockopt()
returned success?

Do you have another getsockopt() call after it to retrieve the value
again?

Just a really wild guess, and likely wrong, perhaps setting the MSS is
a priviledged operation and regular userids cannot do it?

The manpage for TCP doesn't say anything about priviledge though:

TCP_MAXSEG
The maximum segment size for outgoing TCP packets. If this
option is set before connection establishment, it also changes
the MSS value announced to the other end in the initial packet.
Values greater than the (eventual) interface MTU have no effect.
TCP will also impose its minimum and maximum bounds over the
value provided.


Also, if you intend your code to be portable, keep in mind that not
all stacks actually allow the application to set the MSS, only to
retrieve it.

rick jones
--
The glass is neither half-empty nor half-full. The glass has a leak.
The real question is "Can it be patched?"
these opinions are mine, all mine; HP might not want them anyway...
feel free to post, OR email to rick.jones2 in hp.com but NOT BOTH...

I didn't cut-paste the whole thing...but yes:
I do check for the return status of setsockopt () and it is success.
Also, getsockopt(MAXSEG) (after setsockopt(MAXSEG)) returns the 1024.
I am running the program as "root".

Not sure if there is anything else missing ... could somone please
provide
some pointers.

Thanks,
Vishal.

vishal...@gmail.com <(E-Mail Removed)> wrote:
> I didn't cut-paste the whole thing...but yes:
> I do check for the return status of setsockopt () and it is success.
> Also, getsockopt(MAXSEG) (after setsockopt(MAXSEG)) returns the 1024.
> I am running the program as "root".

> Not sure if there is anything else missing ... could somone please
> provide some pointers.

Well, the manpage did suggest there were limits applied, and there is
always the prospect of a bug, so an examination of the stack's source
code may be inorder.

Lets back-up a bit first though - why do you want to set the MSS to
1024 bytes? What issue are you trying to resolve?

rick jones
--
oxymoron n, commuter in a gas-guzzling luxury SUV with an American flag
these opinions are mine, all mine; HP might not want them anyway...
feel free to post, OR email to rick.jones2 in hp.com but NOT BOTH...

I want to test whether my TCP stack abides by MSS advertised by the
peer. Hence, thought of decreasing MSS from the peer to see how my TCP
behaves.

vishal...@gmail.com <(E-Mail Removed)> wrote:
> I want to test whether my TCP stack abides by MSS advertised by the
> peer. Hence, thought of decreasing MSS from the peer to see how my TCP
> behaves.

You might try setting, on the linux system, a route PMTU on a specific
hostroute for the IP on which your TCP stack resides. Or, if you know
that it is only for testing and there isn't much else going-on, you
could simply change the interface MTU on the linux system for the
duration of your test.

rick jones
--
oxymoron n, commuter in a gas-guzzling luxury SUV with an American flag
these opinions are mine, all mine; HP might not want them anyway...
feel free to post, OR email to rick.jones2 in hp.com but NOT BOTH...