[Ocfs2-tools-commits] branch, dups, created. ocfs2-tools-1.4.0-357-gfa85910

Thu Jul 30 12:16:00 PDT 2009

This is an automated email from the git hooks/post-receive script. It was
generated because a ref change was pushed to the repository containing
the project "Tools to manage the ocfs2 filesystem.".

The branch, dups has been created
        at  fa8591043db17c5d141d1892ed71ac05816d2479 (commit)

- Log -----------------------------------------------------------------
commit fa8591043db17c5d141d1892ed71ac05816d2479
Author: Joel Becker <joel.becker at oracle.com>
Date:   Sun Jul 19 01:20:18 2009 -0700

    fsck.ocfs2: Implement Pass 1D resolution of multiply-claimed clusters.

    Inodes that share multiply-claimed clusters can be duplicated or cloned.
    By copying the inode data off to other clusters, this removes the
    multiple claims on the original clusters.  We then remove the cloned
    inode, which points to the originally shared clusters.  We use
    ocfs2_truncate_full() to make sure we don't free our multiply-claimed
    clusters until they're really unused.

    We also implement deleting inodes that share multiply claimed clusters,
    because it can use the same ocfs2_truncate_full() code.  Here we also
    mark the inode as unused.  Pass 2 will let the user remove the directory
    entry (behavior copied from e2fsck).

    Signed-off-by: Joel Becker <joel.becker at oracle.com>

commit fa049fdf2c0fda387725bcb65eaf10208e2c5aa7
Author: Joel Becker <joel.becker at oracle.com>
Date:   Wed Jul 22 17:15:50 2009 -0700

    libocfs2: Allow callers to specify the free_clusters routine in ocfs2_truncate

    o2fsck wants to call ocfs2_truncate but handle freeing the clusters
    itself.  Rather than reimplementing the truncate code in fsck, we create
    ocfs2_truncate_full() that can take a free_clusters() routine.
    ocfs2_truncate() just calls ocfs2_truncate_full() with a NULL routine
    pointer.

    Signed-off-by: Joel Becker <joel.becker at oracle.com>

commit 3f94ade22c3b15d84eb20a7ab3a95e3f5dcc062a
Author: Joel Becker <joel.becker at oracle.com>
Date:   Sat Jun 13 19:34:47 2009 -0700

    fsck.ocfs2: Implement Pass 1D's prompt for dup fixes.

    For each inode claiming shared clusters, ask about cloning or deleting
    them.

    Chain allocators are skipped, as they are not safe to clone or delete.
    They are very location specific.  If they are sharing with another chain
    allocator, pass0 should have handled it.  If they are sharing with a
    regular inode, that inode can be cloned or deleted.

    The other system files can be cloned but not deleted.

    Signed-off-by: Joel Becker <joel.becker at oracle.com>

commit eaff1fe8e5d947c91b55d4cbf08da60b482f4b77
Author: Joel Becker <joel.becker at oracle.com>
Date:   Sat Jun 13 17:04:54 2009 -0700

    fsck.ocfs2: Implement the reporting portion of pass 1D.

    If a cluster has more than one owner, print out the files that own
    it.  This is information prior to correction of the problem.

    How can a cluster we determined to be multiply-claimed have only one
    owner?  If reconciling a different cluster pass has removed another
    owner, this one may be left with only one.

    Signed-off-by: Joel Becker <joel.becker at oracle.com>

commit bcfd68d97bc77ba26c31782e99cd4e12c80404d5
Author: Joel Becker <joel.becker at oracle.com>
Date:   Thu Jun 11 21:22:56 2009 -0700

    fsck.ocfs2: Implement Pass 1C, scanning for inode names.

    Pass 1C scans the directory tree to provide names for any inode owning
    multiply-claimed clusters.  This allows Pass 1D to print readable names
    when asking questions.

    Signed-off-by: Joel Becker <joel.becker at oracle.com>

commit 9d53f3548047af1fbd288e25a14a45ea6ae6d9d2
Author: Joel Becker <joel.becker at oracle.com>
Date:   Tue Jun 9 15:02:56 2009 -0700

    fsck.ocfs2: Implement Pass 1B to find multiply-claimed clusters.

    Pass 1B builds up data structures describing every cluster claimed by
    multiple objects.  It knows what objects own the cluster, and it knows
    what objects have multiply claimed clusters.  This can include data,
    chain groups, and extended attributes.

    Signed-off-by: Joel Becker <joel.becker at oracle.com>

commit f497c1e85dfc096268dc173bede20bfa202bd3d7
Author: Joel Becker <joel.becker at oracle.com>
Date:   Tue Jun 9 10:18:13 2009 -0700

    fsck.ocfs2: Add o2fsck_abort()

    When a void function wants to exit o2fsck, it can't just exit(1).  o2fsck
    may have cluster state and locks that need cleaning.  o2fsck_abort()
    will send SIGTERM to fsck, thus triggering a clean abort.

    Signed-off-by: Joel Becker <joel.becker at oracle.com>

commit d784dcc83b07b2637578117cf19026734513ac08
Author: Joel Becker <joel.becker at oracle.com>
Date:   Mon Jun 8 16:55:03 2009 -0700

    fsck.ocfs2: Allocate and use duplicate clusters bitmap.

    This bitmap keeps track of any clusters claimed by multiple filesystem
    objects.  We will correct the multiply-claimed clusters in latter
    changes.

    Signed-off-by: Joel Becker <joel.becker at oracle.com>

commit 3096bb12f436bf05721f33fd0926984f0def85ef
Author: Joel Becker <joel.becker at oracle.com>
Date:   Tue Jun 9 11:21:59 2009 -0700

    fsck.ocfs2: Remove the unused ost_bad_inodes bitmap.

    We allocate a bitmap of bad inodes.  We test it in pass 2.  But we
    never, ever set a bit in it.

    It looks to be cloned from e2fsck.  In e2fsck, they don't fix invalid
    inode fields in pass 1.  They just mark the inode in the bad map.
    Later, in pass2, they test the bad map and fix the fields there.  The
    advantage is that they can print a pathname.

    That's a major rework for o2fsck, so we're just going to remove the
    unused bad map for now.  If someone wants to add it back later, they
    can.

    Signed-off-by: Joel Becker <joel.becker at oracle.com>

-----------------------------------------------------------------------

hooks/post-receive
-- 
Tools to manage the ocfs2 filesystem.