Re: HAMMER and bad disk sectors

[Matthew Dillon <dillon@xxxxxxxxxxxxxxxxxxxx>]

:Do you think it might be useful to have a bad block table? It may take
:days for a new hard drive to arrive, and sometimes the only choice is to
:keep the box running for the time being.
:
:Petr

    I don't think it would be useful.  Even with UFS/FFS the bad block
    handling has mostly been removed over the years.  A bad block tends to be
    a precursor to a dead disk, if the system is important you need to
    replace the disk immediately and not try to baby it along.  If the bad
    block happens to appear in the middle of file data the filesystem might
    survive but if it appears anywhere else that can spell doom.

    Basically trying to implement a bad block handling feature in the
    filesystem to cover all the possible cases where a bad block can
    interfere with an operation is (almost) an exercise in futility.  It
    takes a lot of man-hours, it is hard to test to make sure it actually
    works as advertised, and there are tons of possible situations and
    cases.  It just isn't worth the effort in my view.

					-Matt
					Matthew Dillon 
					<dillon@backplane.com>