HEADS UP - PFS master/slave downgrading/upgrading bug fixed w/HAMMER

[Matthew Dillon]

    There is a bug where if you have a PFS master and then downgrade it to
    a SLAVE, then upgrade it to a MASTER again, the upgrade operation
    will wipe (probably) most of the data in the PFS.

    The bug is due to the sync-end-tid parameter for the PFS not being
    properly updated when the PFS is downgraded from a master to a slave.
    Then when it is upgraded again HAMMER winds up rolling the filesystem
    back to an older transaction id, basically destroying the PFS.

    This bug only effects PFSs going from MASTER->SLAVE.  PFS's going from
    SLAVE->MASTER (which is the most common use case) do not have this bug.
    Normal filesystem version upgrades (e.g. from version 1 to version 2)
    are also unaffected.

    Most people don't downgrade PFS masters's to slaves under normal
    operation but in case you were thinking about it, don't do it with
    kernels prior to the fix :-)

    A fix has been committed to HEAD and the DragonFly_RELEASE_2_4
    branch.

					-Matt
					Matthew Dillon 
					<dillon at backplane.com>