From owner-freebsd-hackers Thu Aug 1 09:20:55 1996 Return-Path: owner-hackers Received: (from root@localhost) by freefall.freebsd.org (8.7.5/8.7.3) id JAA23102 for hackers-outgoing; Thu, 1 Aug 1996 09:20:55 -0700 (PDT) Received: from root.com (implode.root.com [198.145.90.17]) by freefall.freebsd.org (8.7.5/8.7.3) with ESMTP id JAA23095 for ; Thu, 1 Aug 1996 09:20:52 -0700 (PDT) Received: from localhost (localhost [127.0.0.1]) by root.com (8.7.5/8.6.5) with SMTP id JAA08842; Thu, 1 Aug 1996 09:20:18 -0700 (PDT) Message-Id: <199608011620.JAA08842@root.com> X-Authentication-Warning: implode.root.com: Host localhost [127.0.0.1] didn't use HELO protocol To: Thomas Graichen cc: hackers@FreeBSD.org Subject: Re: vm questions In-reply-to: Your message of "Thu, 01 Aug 1996 18:01:33 +0200." <199608011601.SAA01016@mordillo> From: David Greenman Reply-To: dg@root.com Date: Thu, 01 Aug 1996 09:20:18 -0700 Sender: owner-hackers@FreeBSD.org X-Loop: FreeBSD.org Precedence: bulk >i'm currently reading "the design & implementation ..." as some kind of >relaxation to all the learning for my physics examinations and have some >questions about the vm stuff - can someone - john or david - or maybe someone >else please look at them ? > >* on page 145 is written: "a limitation of the implementation is that it >cannot collaps two objects if either of them has allocated a pager. this >limitation is serious, since pagers are allocated when the system begins >running short of memory -- precisely the time when reclaiming of memory from >collapsed objects is most necesarry." - is this the swap leaking bug in the >original 4.4 BSD vm system ? - how is this problem avoided by the FreeBSD vm >system ? (it was from the chapter "collapsing of shadow chains") It was in the original system. It's fixed in FreeBSD by changing the code to allow objects with pagers to be collapsed. >* on page 158 is written: "a more consistent interface can be obtained by >using a common cache for both the virtual-memory system and the >filesystem. three approaches to merging the two caches are being >undertaken. one approach is to have the filesystem use objects in the >virtual-memory system as its cache; a second approach is to have the >virtual-memory objects that map files use the existing filesystem cache; the >third approach is to create a new cache that is a merger of the two existing >caches, and to convert both the virtual memory and the filesystems to use this >new cache. each of these approaches has its merits and drawbacks; it is not >yet clear which approach will work best." (this is from the chapter "vnode >pager") - my questions here are: which of the three ways is implemented in >FreeBSD ? - which are implemented in other operating systems (linux, sunos, >digital unix, etc.) ? We use the first approach. I can't speak about what other systems do. >* can anyone please write some lines about the basic changes to the vm-system >in FreeBSD (so that i can understand it :-) - i don't know much about the >kernel stuff other than reading the mailinglists and commit-messages - but i >think i understand whats written in the above book - i hope at least partially >:-) - it would also be interesting to get some feeling for the fork/exec/pipe >changes ? > >would be nice if someone could scribble something down about it - so that my >couriosity after reading the book is satsified :-) - a lot of thanks in >advance Yeah, we could write a book about this...and perhaps someday we will. A description of the changes is far beyond what can be written in email. >p.s.: and one last question - what is the state of 4.4 BSD ? - in the book it >is said that lite 2 was really the last one - but from some mailinglists it >sounds that kirk and maybe others are heavily working on filesystems etc. - >will there be something like lite 3 ? There are no plans for a lite-3 that I know of right now. Kirk is doing consulting work for BSDI and is working on some FFS related performance issues. -DG David Greenman Core-team/Principal Architect, The FreeBSD Project