[Q] file cache behaviour

From: Bernd Melchers (melchers@fu-berlin.de)
Date: Sat Feb 26 2000 - 18:14:36 EST

  • Next message: Werner Almesberger: "Re: How to survive in a Micro$oft environment??"

    I have problems to understand the file cache algorithms.
    The following behaviour is observed with kernel 2.2.15pre5 + ide20000127
    patch.
    My Computer has 256 MB RAM and initially a big part of them is free.
    If i generate a big file, most of the free memory is used for the page
    cache, as expected:

    % dd if=/dev/zero of=xtmp bs=1000000 count=300
    % free
                 total used free shared buffers
    cached
    Mem: 258248 255208 3040 6608 188100
    11016
    -/+ buffers/cache: 56092 202156

    If i remove this file, the file file cache is further used:
    % rm xtmp
    % free
                 total used free shared buffers
    cached
    Mem: 258248 255156 3092 6968 188008
    11040
    -/+ buffers/cache: 56108 202140

    But what is in the file cache? There is no corresponding file system
    data for the file cache, because 'xtmp' vanished.

    My other problem:
    Now if i *read* the same file instead of removing the file like above,
    the file
    cache is not longer used after the unlink command:
    % dd if=/dev/zero of=xtmp bs=1000000 count=300
    % free
    Mem: 258248 255252 2996 5892 195252
    4956
    -/+ buffers/cache: 55044 203204
    % wc xtmp
    % free
                 total used free shared buffers
    cached
    Mem: 258248 255540 2708 4852 5244
    216700
    -/+ buffers/cache: 33596 224652
    % rm xtmp
    % free
                 total used free shared buffers
    cached
    Mem: 258248 44304 213944 4864 5272
    5448
    -/+ buffers/cache: 33584 224664

    It seems to me that the file cache ...
    - is used longer than the live time of a file, if the corresponding file
    is *written* but
    - vanishes if the file is *read* and then deleted.

    I won't say that this is an error in the cache code, i don't know it!
    But i simply don't understand whats happen. Could someone explain?
    The described behaviour seems to me a little inconsistent, out of the
    view
    of a simple user.

    Bernd Melchers 

    -- 
melchers@fu-berlin.de
Freie Universitaet Berlin

    -
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu
Please read the FAQ at http://www.tux.org/lkml/

    This archive was generated by hypermail 2b29 : Sat Feb 26 2000 - 18:33:37 EST