From owner-freebsd-security Wed Jun 7 9:14:12 2000 Delivered-To: freebsd-security@freebsd.org Received: from apollo.backplane.com (apollo.backplane.com [216.240.41.2]) by hub.freebsd.org (Postfix) with ESMTP id 5068A37BA89 for ; Wed, 7 Jun 2000 09:14:04 -0700 (PDT) (envelope-from dillon@apollo.backplane.com) Received: (from dillon@localhost) by apollo.backplane.com (8.9.3/8.9.1) id JAA01082; Wed, 7 Jun 2000 09:13:42 -0700 (PDT) (envelope-from dillon) Date: Wed, 7 Jun 2000 09:13:42 -0700 (PDT) From: Matthew Dillon Message-Id: <200006071613.JAA01082@apollo.backplane.com> To: Valentin Nechayev Cc: freebsd-security@FreeBSD.ORG Subject: Re: FreeBSDDEATH.c.txt (mmap dirty page no check bug) References: <200006070424.e574Od303232@cwsys.cwsent.com> <200006070655.XAA97086@apollo.backplane.com> <20000607144421.A82711@lucky.net> Sender: owner-freebsd-security@FreeBSD.ORG Precedence: bulk X-Loop: FreeBSD.org :Of course, of course. :It is general problem of any public-accessable resource. :Do you think you can really fix this world? Or do you try to emit /tmp :as philosophical category? : :> MFS is a terrible idea for /tmp. Each page in an MFS filesystem eats :> *TWO* pages of physical memory (until swapped). This means that the : :It is problem of one broken realization, isn't it? : :-- :NVA No, I think its a problem with concept. Modern systems like ours have a VM page cache which can cache anything -- including normal filesystems. Normal filesystems aren't quite as fast as MFS (even with an async mount) because we assume, rightly, that certain pieces of data can be flushed more quickly in order to leave more memory available to other things. The concept of 'MFS' in general could be rephrased to be 'forcing the system to use more memory to cache a filesytem at the cost of making less memory available to programs and other filesystems'. If a person wants that tradeoff, then it would be just as easy to implement with a normal filesystem simply by turning off the update daemon (for that filesystem), turning off write-behind (for that filesystem), and either increasing the size of the buffer cache to accomodate more dirty pages or disassociating the dirty filesystem pages from the buffer cache (making them look like dirty mmap'd pages). And then using an async mount. A memory filesystem, even the best implemented one you could write, is best used in situations where you don't *have* a local disk in the first place. -Matt Matthew Dillon To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-security" in the body of the message