From owner-freebsd-security Wed Sep 1 21:32:14 1999 Delivered-To: freebsd-security@freebsd.org Received: from khavrinen.lcs.mit.edu (khavrinen.lcs.mit.edu [18.24.4.193]) by hub.freebsd.org (Postfix) with ESMTP id 9B42214CB0 for ; Wed, 1 Sep 1999 21:32:12 -0700 (PDT) (envelope-from wollman@khavrinen.lcs.mit.edu) Received: (from wollman@localhost) by khavrinen.lcs.mit.edu (8.9.1/8.9.1) id AAA08736; Thu, 2 Sep 1999 00:31:35 -0400 (EDT) (envelope-from wollman) Date: Thu, 2 Sep 1999 00:31:35 -0400 (EDT) From: Garrett Wollman Message-Id: <199909020431.AAA08736@khavrinen.lcs.mit.edu> To: Darren Reed Cc: freebsd-security@FreeBSD.ORG Subject: Re: socket buffer limits (was: Re: FW: Local DoS in FreeBSD) In-Reply-To: <199909020347.NAA23859@cheops.anu.edu.au> References: <199909020335.XAA08433@khavrinen.lcs.mit.edu> <199909020347.NAA23859@cheops.anu.edu.au> Sender: owner-freebsd-security@FreeBSD.ORG Precedence: bulk X-Loop: FreeBSD.org < said: > How about failing setsockopt's when they try to increase buffer space > if it would mean buffer space commitments would exceed a high water mark ? That's no different from what I dismissed in my reply to Don. We don't want to restrict the potential TCP window a user may offer, just how much actual kernel memory he may tie up at once. > Also, what if mbufs are dropped and/or send/write fails (ENOBUFS) if there > is nobody waiting to receive data and a high water mark has been reached ? POLA violation for sure. (TCP should be rewritten, however, to use the receiver's advertised window for user push-back, but that's at least a full-time person-year's worth of work to write, test, and tune. That would mean that loopback connections could only use one window's-worth of kernel memory for buffering.) -GAWollman -- Garrett A. Wollman | O Siem / We are all family / O Siem / We're all the same wollman@lcs.mit.edu | O Siem / The fires of freedom Opinions not those of| Dance in the burning flame MIT, LCS, CRS, or NSA| - Susan Aglukark and Chad Irschick To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-security" in the body of the message