Date: Sun, 09 Feb 1997 16:48:54 +0900 From: Kenjiro Cho <kjc@csl.sony.co.jp> To: Garrett Wollman <wollman@lcs.mit.edu> Cc: Eka Kelana <eka@werty.wasantara.net.id>, freebsd-hackers@FreeBSD.ORG, freebsd-bugs@FreeBSD.ORG, dg@root.com Subject: Re: Need quick answer Message-ID: <199702090749.QAA29283@hotaka.csl.sony.co.jp> In-Reply-To: Your message of "Fri, 07 Feb 1997 07:29:24 EST." <9702071229.AA18092@halloran-eldar.lcs.mit.edu>
next in thread | previous in thread | raw e-mail | index | archive | help
<<On Fri, 7 Feb 1997 05:28:25 +0700, Eka Kelana <eka@werty.wasantara.net.id> said:
> I need a good explanation about this code in in_rmx.c:
> ...
> if (!rt->rt_rmx.rmx_sendpipe && !(rt->rt_rmx.rmx_locks & RTV_SPIPE))
rt-> rt_rmx.rmx_sendpipe = tcp_sendspace;
> if (!rt->rt_rmx.rmx_recvpipe && !(rt->rt_rmx.rmx_locks & RTV_RPIPE))
rt-> rt_rmx.rmx_recvpipe = tcp_recvspace;
> ...
>>>>> On Fri, 7 Feb 1997 07:29:24 -0500, Garrett Wollman <wollman@lcs.mit.edu> said:
>> The reason why that is there is because otherwise tcp_close() won't
>> save the ssthresh.
If this is the only reason to set the pipe sizes, I would suggest to
make modifications to
- tcp_close to update ssthresh even when the pipe sizes are not
specified.
- in_addroute to leave the pipe sizes unspecified.
Then, ssthresh can be updated in the same manner, but no side-effect.
IMO, if the pipe size is ever set automatically, it should reflect the
measured throughput * RTT.
--kj
---
Kenjiro Cho
Sony Computer Science Laboratory Inc.
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?199702090749.QAA29283>