home *** CD-ROM | disk | FTP | other *** search
- Newsgroups: comp.unix.ultrix
- Path: sparky!uunet!decwrl!pa.dec.com!nntpd2.cxo.dec.com!nabeth!alan
- From: alan@nabeth.enet.dec.com (Alan Rollow - Alan's Home for Wayward Tumbleweeds.)
- Subject: Re: max[d|t|s]siz
- Message-ID: <1992Aug13.183731.4038@nntpd2.cxo.dec.com>
- Keywords: kernel, maxdsiz, maxtsiz, maxssiz
- Lines: 44
- Sender: alan@nabeth (Alan Rollow - Alan's Home for Wayward Tumbleweeds.)
- Reply-To: alan@nabeth.enet.dec.com (Alan Rollow - Alan's Home for Wayward Tumbleweeds.)
- Organization: Digital Equipment Corporation
- References: <1992Aug13.134919.22439@morgan.ucs.mun.ca>
- Date: Thu, 13 Aug 1992 18:37:31 GMT
-
-
- In article <1992Aug13.134919.22439@morgan.ucs.mun.ca>, kforward@morgan.ucs.mun.ca (Ken Forward) writes:
- >Newsgroups: comp.unix.ultrix
- >From: kforward@morgan.ucs.mun.ca (Ken Forward)
- >Subject: max[d|t|s]siz
- >
- >A colleague has recently upgraded his DS5000/125 w/ Ultrix 4.2 to 128MB
- >of memory. Aside from changing 'physmem' in the kernel configuration file,
- >we are wondering how best to set the maxdsiz, maxtsiz, and maxssiz
- >parameters; in particular, is there any problem with setting them all
- >very large, eg, equal to physmem ? What sort of overhead would we incur
- >by doing so ?
- >
- I haven't looked that closely at the code, but a quick look at the
- base system sources and vm sources suggest it is used for two things:
-
- 1. Limit the {data,text,stack} sizes. This is just a check a
- check against the value, when the process wants to grow.
-
- 2. Allocate dmap entries for the swap space. I'm not sure of
- implementation, but I think a dmap entry represents a swapfrag
- block of swap space. Each process will need enough dmap entries
- to swap all the potential data, text and stack space, if it is
- ever swapped out.
-
- So, if you raise one of the max[d|t|s]siz parameters much beyond the
- usual, you'll be using more kermel memory for dmap entries. A quick
- look at /usr/include/sys/dmap.h suggests that each dmap entry is 16
- bytes. Of course if you use a large swapfrag size, you can reduce
- the number of dmaps that are needed.
-
- So, I conclude that the overhead for setting them high is "not much".
- If you discover differently, let us know.
-
- What you will have to most concerned about is sufficient page/swap
- space. You need to be able to reserve as much as the process may
- need.
-
- >--
- > Ken Forward, Dept of Physics, Memorial University of Newfoundland
- >
- --
- Alan Rollow alan@nabeth.cxo.dec.com
-
-