Can O_SYNC be implemented by using fsync?

From: Ulrich Kunitz (gefm21@uumail.de)
Date: Sat May 13 2000 - 07:36:59 EDT

  • Next message: Stephen C. Tweedie: "Re: [patch] balanced highmem subsystem under pre7-9"

    Hi,

    this discusses item 5.1 in Alan's list of outstanding fixes before 2.4.

    I'm using fsync to write out all dirty pages/blocks of a file in
    generic_file_write to get O_SYNC behaviour. Patch is included below. It
    is the wrong way to do O_SYNC? I've checked this and it seems to do what
    would be expected and it caused no harm to me.

    Another idea would be to use sync_page() in generic_file_write() that
    calls block_sync_page() for ext2. block_sync_page() calls only
    run_rask_queue(tq_disk). IMHO adding of sync_page_buffers() to
    block_sync_page() would be required and there is still the inode and
    superblock infos, that need to be written to disk before
    generic_file_write() returns.

    Which kernel version did support O_SYNC for ext2? AFAIK 2.2 kernels
    don't support O_SYNC.

    This is my patch:

    --- linux-2.3.99-pre8/mm/filemap.c Sat May 13 10:13:32 2000
    +++ linux-2.3.99-pre8-uk1/mm/filemap.c Sat May 13 10:43:21 2000
    @@ -39,6 +39,8 @@
      * page-cache, 21.05.1999, Ingo Molnar <mingo@redhat.com>
      *
      * SMP-threaded pagemap-LRU 1999, Andrea Arcangeli <andrea@suse.de>
    + *
    + * O_SYNC support, 29. 4. 2000, Ulrich Kunitz <gefm21@uumail,de>
      */
     
     atomic_t page_cache_size = ATOMIC_INIT(0);
    @@ -2521,7 +2523,14 @@
             if (cached_page)
                     page_cache_free(cached_page);
     
    + if ((file->f_flags & O_SYNC) && file->f_op && file->f_op->fsync) {
    + err = file->f_op->fsync(file, file->f_dentry);
    + if (err < 0)
    + status = err;
    + }
    +
             err = written ? written : status;
    +
     out:
             up(&inode->i_sem);
             return err;

    Ciao,

    Uli Kunitz 

    -- 
Ulrich Kunitz (gefm21@uumail.de)

    -
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 May 13 2000 - 07:39:11 EDT