Date: Thu, 17 Feb 2000 19:50:58 -0800 From: jayanth <jayanth@yahoo-inc.com> To: fbsdnet <freebsd-net@FreeBSD.ORG> Subject: tcpsendspace and tcprecvspace issue ? Message-ID: <38ACC1A2.A830B26C@yahoo-inc.com>
next in thread | raw e-mail | index | archive | help
There seems to be a certain condition where tcp advertises the
wrong window size when the tcpsendspace and
tcprecvspace value are set to a value = 65536(>65000 )
Assume rfc1323 is off.
Scenario:
Turn keepalive on and open a connection.
Send some data and let the connection be idle.
let the keepalive timer expire at the server end.
tcp now advertises a window size of 164 instead
of the original window size.
I went through the code and found that
in tcp_input.c in tcp_mss() the following code
bumps the value from 65536 to 65700
tcp_mss()
---------
#ifdef RTV_RPIPE
if ((bufsize = rt->rt_rmx.rmx_recvpipe) == 0)
#endif
bufsize = so->so_rcv.sb_hiwat;
if (bufsize > mss) {
bufsize = roundup(bufsize, mss);
^^^^^^^^^^^^^^^^^^^^^^^^^^^
if (bufsize > sb_max)
bufsize = sb_max;
(void)sbreserve(&so->so_rcv, bufsize);
If the keepalive timer expires the tcp_respond()
uses the sbspace(so->so_rcv) to determine the
window size.
tcp_respond()
------------
ti->ti_win = htons((u_short) (win >> tp->rcv_scale));
else
ti->ti_win = htons((u_short)win);
ti->ti_win is cast to a unsigned short resulting in the wrap around
65700 - 65336 = 164 which is the advertised window size..
The problem occurs if the roundup result is greater than 65536.
Is this a bug ?
The fix is either not do a roundup() if rfc1323 is off
or not to allow this value to exceed 65535 if rfc1323 is off.
Hope I am making sense.
cheers
jayanth
To Unsubscribe: send mail to majordomo@FreeBSD.org
with "unsubscribe freebsd-net" in the body of the message
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?38ACC1A2.A830B26C>