Skip site navigation (1)Skip section navigation (2)
Date:      Tue, 6 Jun 2000 23:55:03 -0700 (PDT)
From:      Matthew Dillon <dillon@apollo.backplane.com>
To:        Cy Schubert - ITSD Open Systems Group <Cy.Schubert@uumail.gov.bc.ca>
Cc:        "mouss" <usebsd@free.fr>, "Peter van Dijk" <petervd@vuurwerk.nl>, freebsd-security@FreeBSD.ORG
Subject:   Re: FreeBSDDEATH.c.txt (mmap dirty page no check bug) 
Message-ID:  <200006070655.XAA97086@apollo.backplane.com>
References:   <200006070424.e574Od303232@cwsys.cwsent.com>

next in thread | previous in thread | raw e-mail | index | archive | help
:> also, the BSD/OS mfs proposal is not that god. This limits the size of /tmp,
:> and uses mfs for things that do not need to be in mfs.
:> the first thing I used to do on BSD/OS was to remove the mfs mount and to
:> softlink /var/tmp to /tmp.
:
:I disagree with this.  /tmp is cleared at boot while /var/tmp is not.  
:The reason for this is to have files remain across boot.

    Maybe on your system it is, but try running a multi-user system that
    way and you will quickly find your /var/tmp filled up to the brim.  Or,
    worse, you will find one of the two tmp directories filling up while
    the other remains entirely empty, or vise-versa depending on which
    programs your users run.

    The argument that we should have two tmp's because one should be
    treated differently from the other doesn't hold any water.  There
    should be one tmp, period.  Since programs tend to use /var/tmp 
    and /tmp interchangeably these days, one has to be a symlink to the
    other.  But trying to classify the two as having to have different
    characteristics only creates sysadmin headaches.

    What it comes down to is that it is far easier and far more robust to
    have a single (larger) temporary filesystem to maintain then to have 
    two.

:mfs is generally (arguably on some O/S's) faster than writing data to 
:disk.  (If writing to disk is faster than using mfs, assuming there's 

    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
    active dataset (what processes have accessed recently) eats twice
    as much physical memory as it needs to.  MFS also needlessly loads
    down the VM system when it does start to stress memory.   Simple
    things like someone tar xvf'ing a large distribution in /tmp can bring
    the machine to its knees with an MFS partition where the same operation
    on a normal filesystem would barely glitch most of the resource meters.

    MFS would be my *LAST* choice for a tmp.

    Like it or not, /tmp and /var/tmp are here to stay.  /usr/tmp doesn't
    exist on most systems (thank god!), so it would be stupid of us to add
    it in. 

					-Matt
					Matthew Dillon 
					<dillon@backplane.com>


To Unsubscribe: send mail to majordomo@FreeBSD.org
with "unsubscribe freebsd-security" in the body of the message




Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?200006070655.XAA97086>