>>>>> Ryusuke Konishi writes:
>>>>> On Thu, 14 Jun 2012 11:22:17 +0700, Ivan Shmakov wrote:
>>>>> Ryusuke Konishi writes:

[…]

 >>> Implementing Off-line resizing is a bit tough task because it
 >>> requires a user space library to read internal meta data of the
 >>> nilfs file system and append updated blocks to it.

        … I guess that a user-level toolkit akin to e2fsprogs would be
        handy anyway.  (Shouldn't there be an fsck(8) for NILFS, BTW?)

        Other than resizing, there could be a feature to “split” a
        filesystem — take one or more “snapshot” checkpoints and make a
        copy of the parts of the filesystem “reachable” from these
        checkpoints on a distinct block device (perhaps of a differing
        size), which seems like one more way to do backups.

 >>> This is almost equivalent to implementing kernel functions of nilfs
 >>> in user land.

[…]

 >> I wonder, how complicated this task would be?  In particular, do I
 >> need to alter all the segments to resize filesystem, or only
 >> specific ones?  Also, are these internal metadata records just “a
 >> bunch of C structs”, or something more complex?

 > This needs at least the following tasks:

 > 1) Update two super blocks to increase/decrease the number of total
 > segments and the device size. (easy)

 > 2) Update segment usage information to adjust free segment count of
 > nilfs_sufile_header struct (sh_ncleansegs) which is written in
 > the segment usage file. (Sadly not easy)

 > 3) If truncating in-use segments, relocating blocks and updating disk
 > address translation information are also needed. (very hard)

        ACK.  (Fortunately, I'm not that interested in shrinking
        filesystems right now.)

 > For reference, a discussion on this topic is found in the archives of
 > the list:

 > http://marc.info/?t=128419843600003&r=1&w=2

        ACK, thanks!  I hope to take a closer look at this problem
        within two weeks or so.

        (JFTR: those reading via Gmane may access the thread via either
        of [1, 2]; the first message in the thread is [3].)

[1] http://comments.gmane.org/gmane.comp.file-systems.nilfs.user/397
[2] http://thread.gmane.org/gmane.comp.file-systems.nilfs.user/397/focus=1686
[3] news:baefa5609cc3ab4bae1625cad4c27e5c02986...@tdpost.td.tandbergdata.com

-- 
FSF associate member #7257

--
To unsubscribe from this list: send the line "unsubscribe linux-nilfs" in
the body of a message to [email protected]
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Reply via email to