⟦344289524⟧

TextFile

RECOVERY
 
The recovery program is used to prepare disks which are to be 
used with an R1000.  The process consists of several steps which
are listed here in the order required.  All of these steps
must be performed prior to using the disk in an R1000 system.
 
Formatting -
 
    The formatting process writes information onto the disk which
    is needed by the controller to transfer data.  The format of
    this information can be found in appendix A of the Spectra
    Logic 121 manual.  The RECOVERY program uses the controller's
    format drive command to format the disk.  Formatting is optional
    when invoking the RECOVERY program unless the specified disk
    has no labels.
 
Flagging bad blocks -
 
    An integral part of building a disk for an R1000 processor is to
    build a bad block map which records the location of each defective
    area on the disk.  The R1000 will not attempt to read or write to
    any block recorded in the bad block map.  In addition to building
    the bad block map the RECOVERY program re-formats each bad block
    with the BAD SECTOR bit set in the format data.  This will cause
    a BAD SECTOR error if any system software attempts to access these
    bad blocks.  All bad blocks will be flagged when the RECOVERY program
    is invoked.  This insures that blocks added to the bad block map
    by the system get flagged.
 
Surface Analysis -
 
    The surface analysis portion of the RECOVERY program is optional
    unless the disk was formatted.  The surface analysis is intended
    to insure the reliablity of the disk.  From one to three passes
    are allowed, each taking several minutes (45 for a Fujitsu 2351
    EAGLE).  Each pass consists of writing to every block on the disk
    and then reading back each block and checking the data.  A pass
    consists of both a write and read phase.  If the disk has just
    been formatted the write phase of the first pass is eliminated
    because the process of formatting writes the data.
 
Writing defect map -
 
    The defect map contains information about each defective block on
    the disk.  This data is stored on cylinder zero and is pointed to
    by the shared label (see below).  As many as 2048 defects may be
    recorded with the current defect map format.  The defect map is
    always written to the disk.
\f


Boot Label -
 
    The boot label is located on block 1 (cylinder 0, head 0, sectors 2-3)
    of every disk.  The boot label contains pointers to DFS files which
    contain the I/O Processor's Kernel and initial PASCAL programs to be
    executed at boot time.  An empty boot label is always written to the
    disk.  It is modified as needed if a DFS is actually built on the
    disk.
 
DFS Label -
 
    The DFS label contains information about the location of the DFS
    directory and DFS free list.  An empty DFS label is always written
    to the disk.  It is modified as needed if a DFS is actually built on
    the disk.  The DFS label is located on block 4 of every disk.
 
Shared (volume) Label -
 
    The shared label contains information about the location of several
    disk structures.  Some of the structures are maintained solely by
    the R1000 File System (RFS) and will not be mentioned here; others
    are shared by the DFS and the R1000 file system.  These are:
 
    o   The size of the disk. (number of cylinders, heads, sectors)
    o   The location of the bad block map.
    o   The location of the retarget map.
    o   The location of the DFS.
    o   The location of the R1000 file system.
    o   The location of the read/write diagnostic portion of the disk.
    o   The serial number of the disk's HDA.
    o   A boolean used to indicate the presence of a DFS.
    o   A boolean used to indicate the presence of an RFS.
 
    The shared label is always written to a disk and the portions
    mentioned here are only changed by the RECOVERY program.  The
    shared label is stored using the R1000 stable storage mechanism
    and is located in block 2 with a copy in block 3.
 
Building the DFS
 
    The DFS is only built if requested.  There are several stages
    reported to the terminal when building a diagnostic file system.
    They are:
 
    o   Constructing free list
 
        The DFS retains information about free disk space as a linked
        list of disk extents.  The free list is first constructed in
        memory by discarding defective blocks from the area allocated
        to the DFS.
\f


    o   Writing free list
 
        The free list just constructed is written to the disk and
        its head is recorded in the DFS label.
        
    o   Allocating and initializing directory
 
        The fixed size DFS directory is allocated from the free list
        and initialized to be empty.  At this time no files exist.
        Pointers to the directory are recorded in the DFS label.
        
    o   Allocating predefined files
 
        All disk structures which must have fixed disk locations
        or are referenced by the boot label are pre-created by the
        RECOVERY program.  These include:
        
        o   The disk bootstrap (located at disk block 0).
        o   All bootable I/O processor kernels.
        o   All bootable PASCAL programs.
        o   All bootable file systems (series 200 only).
        o   The DFS error log.
        
    o   Loading the DFS
 
        After its creation the DFS may be loaded with files from an MT
        format tape.  This step is optional.
        
To run the RECOVERY program you may boot it from tape or invoke it
by typing:
 
    CLI> x recovery
 
The program will first ask which disk drive you wish to format/build.
Answer with the appropriate disk unit number.  RECOVERY will then
attempt to read the disk's labels and bad block map into memory.  If
this step fails the disk must be re-formatted and defect information
read in from a tape created by the RDM program (see RDM documentation).
If the labels are recovered from the disk successfully then you will be
asked if the data contained in the labels should be used for the remainder
of the formatting process.  If you answer yes to this question then
disk defect map data will be read from the disk and the disk will get built
with a DFS only if it had one already.
 
You will be asked if you want to format the disk.  Formatting the disk will
destroy all data.
 
You will be asked if you want to perform surface analysis.  Surface analysis
will destroy all data present on the disk.  A read-only surface analysis
program (CHECKDISK) can be used to check disk integrity.  The RECOVERY
program is not a disk test or exerciser.
\f


Next, the disk labels will be created.  If the disk had readable labels
when the RECOVERY program was invoked and retained the information
contained in them, that data will be used to re-build the labels.  If
not you will be asked to enter some information about how the disk
will be used.  These questions include:
 
    Do you want to build a diagnostic file system on this unit [Y] ?
 
Answering yes to this question will cause space to be allocated on
the disk for a DFS.
 
    Enter last cylinder to be used by the DOS :
 
This question will only be asked if the disk will contain a DFS.  The
DFS will occupy all disk space between cylinder 1 and the cylinder
used here.  The correct answer depends on the type of disk used but
should be no less than 20,000 disk blocks.  It can be calculated as
follows:
 
    (20000 / ((H * S) / 2) + 1)
 
    where:
        H is the number of heads on the disk
        S is the number of sectors on the disk
    
    Enter first cylinder to be used for read/write diagnostics :
 
Hints for space limits are now included.

The read/write diagnostic portion of the disk starts at this cylinder
and extends to the last cylinder of the disk.  At least two cylinders
must be allocated for read/write diagnostics.  Remember that cylinder
numbers start at zero, not at one; so if a disk has 842 cylinders,
numbered 0 .. 841, the largest value which should be used is 840.
This will cause cylinders 840 and 841 to be reserved for read/write
diagnostics.
 
Once the disk labels have been generated the RECOVERY program is
finished building the disk.  If a DFS has been built on the disk
you will be asked if you want to load files into the DFS.  Files
may be loaded from an MT format tape.
 
When the RECOVERY program is done, or if any unrecoverable errors
occur during disk building, it will restart.  The only way to
terminate the RECOVERY program is by re-booting the system.

A new menu selection has been added. This is option 5, install new
DFS only. Older versions of recovery would perform this function if several
questions about formatting and changing limits were properly answered. Option
5 can now be used without having to worry about destroying the state of the
'environment' space on the disk. This does erase all dfs files and as such
should only be done when installing a full dfs backup or when rebuilding
a system using dfs and ak release tapes.
DataMuseum.dk

Rational R1000/400 DFS Tapes

⟦344289524⟧ HLP, TextFile

Derivation

TextFile
