[lkml]   [1997]   [Sep]   [24]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
Messages in this thread
    SubjectRe: New feature
    On Tue, 23 Sep 1997, David S. Miller wrote:

    > Date: Tue, 23 Sep 1997 21:49:12 -0400 (EDT)
    > From: "Richard B. Johnson" <>
    > Correct! Now, is a directory access a file access? I don't think it
    > should be.
    > ATIME is the access time for an inode, when you read an inode which
    > happens to be a directory you are accessing it.

    The directory, yes. After all, I did read (access) all the directories.
    However, some directory files do not contain other directory files.
    They contain "normal" files which were not accessed at all. These files
    should not have had their ATIME updated because they were not accessed.

    I am aware that this is happening. I am not aware that this is correct.
    Further, once these dirtied buffers were written back to the disk, both
    the buffers and the physical media were now in sync. The fs must know
    that they are in sync, the fs wrote them. The next access forces the
    fs to re-read everything from the physical media again.

    Now, this is not, repeat not, a complaint. I just test this stuff. I
    have a SMP machine that is dedicated entirely to testing. Every piece
    of code that runs on that machine has been built from sources on that
    machine including trivial stuff like `cat`.

    So that I didn't raise any feathers, I explained a test method that
    will demonstrate a potential problem. I did not do anything else.

    My Suns do not show ATIME updates on all the files just because the
    directories were accessed. Further, they don't reread everything after
    a sync(). I don't know if this is the correct behavior or not.

    I have no known way of accessing ATIME on my SGI machines because
    `ls` doesn't seem to support it. The SGI machines are used for
    schematic capture and have no programming tools so I can't write
    something to check it out although some day I may build the gcc
    compiler on it.

    Richard B. Johnson
    Analogic Corporation
    Penguin : Linux version 2.1.55 on an i586 machine (66.15 BogoMips).
    Warning : It's hard to stay on the trailing edge of technology.
    Linux : Engineering tool
    Spam : sync@localhost, daemon@loghost

     \ /
      Last update: 2005-03-22 13:40    [W:0.048 / U:172.436 seconds]
    ©2003-2016 Jasper Spaans. hosted at Digital OceanAdvertise on this site