Subject: Re: Kernel resident memory size is quite large
To: None <port-amd64@netbsd.org>
From: None <antiright@gmail.com>
List: port-amd64
Date: 03/30/2006 07:57:54
--tKW2IUtsqtDRztdT
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline


--tKW2IUtsqtDRztdT
Content-Type: message/rfc822
Content-Disposition: inline

Date: Wed, 29 Mar 2006 16:06:46 -0500
From: jefbed
To: Havard Eidnes <he@netbsd.org>
Subject: Re: Kernel resident memory size is quite large
Message-ID: <20060329210646.GA7762@antiright.dyndns.org>
References: <20060329105225.GA18685@antiright.dyndns.org> <20060329.152028.04464439.he@uninett.no> <20060329134932.GA24749@beta.martani.repy.czf> <20060329.160858.06082150.he@uninett.no>
Mime-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <20060329.160858.06082150.he@uninett.no>
User-Agent: Mutt/1.4.2.1i

On Wed, Mar 29, 2006 at 04:08:58PM +0200, Havard Eidnes wrote:
> > > > Is there any reason why the memory used by the kernel is so large?
> > > > Top is showing that kernel processes use 233MB of memory.
> > > > The system has 1GB, so the amount used by the kernel is a large percentage.
> > > > Does this memory include the buffer cache, or is it occupied by the kernel
> > > > itself?
> > >
> > > First I have to ask how you read top's output to come to the
> > > conclusion you do.  E.g. you can't simply add the "RES" sizes of
> > > the processes with [] around their names; as far as I know, these
> > > processes all share the same virtual address space.
> > > 
Yes
> > > However, with that said, it is quite normal for the kernel's data
> > > structures to occupy a major portion of physical memory, since this
> > > includes the area for buffering file data, vnodes etc. etc.
> >                         ^^^^^^^^^^^^^^^^^^^
> >
> > I don't think that buffered file data will show as the RES size of kernel
> > threads.
> 

Buffered data is separate in top's accounting (my system shows 484MB of
file data in memory).  

> No, I agree, but that's why I specifically asked how he came to the
> conclusion that "kernel processes use 233MB of memory", and since no
> raw data was given to support this conclusion, it's worth double-
> checking and asking for that raw data, e.g. a cut+paste from top's
> output.
> 
Paste of top output, as requested, sorted by RES:

load averages:  0.14,  0.16,  0.16                                     16:05:55
72 processes:  1 runnable, 69 sleeping, 1 stopped, 1 on processor
CPU states:  1.5% user,  0.0% nice,  0.5% system,  0.0% interrupt, 98.0% idle
Memory: 452M Act, 151M Inact, 5252K Wired, 21M Exec, 484M File, 54M Free
Swap: 512M Total, 512M Free

  PID USERNAME PRI NICE   SIZE   RES STATE      TIME   WCPU    CPU COMMAND
   11 root      18    0     0K  230M syncer     3:30  0.00%  0.00% [ioflush]
    2 root      10    0     0K  230M usbevt     0:00  0.00%  0.00% [usb0]
    3 root      10    0     0K  230M usbtsk     0:00  0.00%  0.00% [usbtask]
    4 root      10    0     0K  230M usbevt     0:00  0.00%  0.00% [usb1]
    5 root      -6    0     0K  230M atath      0:00  0.00%  0.00% [atabus0]
    6 root      -6    0     0K  230M atath      0:00  0.00%  0.00% [atabus1]
    7 root      -6    0     0K  230M atath      0:00  0.00%  0.00% [atabus2]
    8 root      -6    0     0K  230M atath      0:00  0.00%  0.00% [atabus3]
    9 root      -6    0     0K  230M sccomp     0:00  0.00%  0.00% [atapibus0]
    0 root     -18    0     0K  230M schedule   0:00  0.00%  0.00% [swapper]
   10 root     -18    0     0K  230M pgdaemon   0:00  0.00%  0.00% [pagedaemon]
   12 root     -18    0     0K  230M aiodoned   0:00  0.00%  0.00% [aiodoned]
  399 jefbed     2    0    61M  152M select   267:18  0.15%  0.15% XFree86
  497 jefbed     2    0  4928K 8640K select     0:35  0.00%  0.00% fluxbox
  861 jefbed     2    0  2812K 6948K select     0:58  0.05%  0.05% xterm
  303 root      10    0   200K 6716K mfsidl     0:00  0.00%  0.00% mount_mfs
 8516 jefbed     2    0   644K 4724K poll       0:00  0.00%  0.00% arshell
  538 root       2    0   572K 4604K poll       0:00  0.00%  0.00% arshell
  563 root      18    0  1180K 3944K pause      0:03  0.00%  0.00% ntpd
  491 jefbed     2    0   408K 3208K select     0:00  0.00%  0.00% ssh
 1168 jefbed     2    0   368K 3140K select     0:00  0.00%  0.00% ssh
  659 jefbed     2   10   644K 3024K poll       0:03  0.00%  0.00% xscreensaver
  141 jefbed     2    0  2048K 2960K select     0:06  0.00%  0.00% screen-4.0.2
 7762 jefbed    10    0  1060K 2880K wait       0:00  0.00%  0.00% mutt
17326 jefbed     2    0   568K 2816K netio      0:00  0.00%  0.00% fetchmail


> "RES" coloumn, so if he is indeed seeing 233M for each and every

I do recognize that this memory is shared amongst the processes,
so I'm not implying that *each* uses that amount.

-Jeff


--tKW2IUtsqtDRztdT--