Tools that manage md devices can be found at
   http://www.<country>.kernel.org/pub/linux/utils/raid/....


Boot time assembly of RAID arrays
---------------------------------

You can boot with your md device with the following kernel command
lines:

for old raid arrays without persistent superblocks:
  md=<md device no.>,<raid level>,<chunk size factor>,<fault level>,dev0,dev1,...,devn

for raid arrays with persistent superblocks
  md=<md device no.>,dev0,dev1,...,devn
or, to assemble a partitionable array:
  md=d<md device no.>,dev0,dev1,...,devn
  
md device no. = the number of the md device ... 
              0 means md0, 
	      1 md1,
	      2 md2,
	      3 md3,
	      4 md4

raid level = -1 linear mode
              0 striped mode
	      other modes are only supported with persistent super blocks

chunk size factor = (raid-0 and raid-1 only)
              Set  the chunk size as 4k << n.
	      
fault level = totally ignored
			    
dev0-devn: e.g. /dev/hda1,/dev/hdc1,/dev/sda1,/dev/sdb1
			    
A possible loadlin line (Harald Hoyer <HarryH@Royal.Net>)  looks like this:

e:\loadlin\loadlin e:\zimage root=/dev/md0 md=0,0,4,0,/dev/hdb2,/dev/hdc3 ro


Boot time autodetection of RAID arrays
--------------------------------------

When md is compiled into the kernel (not as module), partitions of
type 0xfd are scanned and automatically assembled into RAID arrays.
This autodetection may be suppressed with the kernel parameter
"raid=noautodetect".  As of kernel 2.6.9, only drives with a type 0
superblock can be autodetected and run at boot time.

The kernel parameter "raid=partitionable" (or "raid=part") means
that all auto-detected arrays are assembled as partitionable.

Boot time assembly of degraded/dirty arrays
-------------------------------------------

If a raid5 or raid6 array is both dirty and degraded, it could have
undetectable data corruption.  This is because the fact that it is
'dirty' means that the parity cannot be trusted, and the fact that it
is degraded means that some datablocks are missing and cannot reliably
be reconstructed (due to no parity).

For this reason, md will normally refuse to start such an array.  This
requires the sysadmin to take action to explicitly start the array
despite possible corruption.  This is normally done with
   mdadm --assemble --force ....

This option is not really available if the array has the root
filesystem on it.  In order to support this booting from such an
array, md supports a module parameter "start_dirty_degraded" which,
when set to 1, bypassed the checks and will allows dirty degraded
arrays to be started.

So, to boot with a root filesystem of a dirty degraded raid[56], use

   md-mod.start_dirty_degraded=1


Superblock formats
------------------

The md driver can support a variety of different superblock formats.
Currently, it supports superblock formats "0.90.0" and the "md-1" format
introduced in the 2.5 development series.

The kernel will autodetect which format superblock is being used.

Superblock format '0' is treated differently to others for legacy
reasons - it is the original superblock format.


General Rules - apply for all superblock formats
------------------------------------------------

An array is 'created' by writing appropriate superblocks to all
devices.

It is 'assembled' by associating each of these devices with an
particular md virtual device.  Once it is completely assembled, it can
be accessed.

An array should be created by a user-space tool.  This will write
superblocks to all devices.  It will usually mark the array as
'unclean', or with some devices missing so that the kernel md driver
can create appropriate redundancy (copying in raid1, parity
calculation in raid4/5).

When an array is assembled, it is first initialized with the
SET_ARRAY_INFO ioctl.  This contains, in particular, a major and minor
version number.  The major version number selects which superblock
format is to be used.  The minor number might be used to tune handling
of the format, such as suggesting where on each device to look for the
superblock.

Then each device is added using the ADD_NEW_DISK ioctl.  This
provides, in particular, a major and minor number identifying the
device to add.

The array is started with the RUN_ARRAY ioctl.

Once started, new devices can be added.  They should have an
appropriate superblock written to them, and then passed be in with
ADD_NEW_DISK.

Devices that have failed or are not yet active can be detached from an
array using HOT_REMOVE_DISK.


Specific Rules that apply to format-0 super block arrays, and
       arrays with no superblock (non-persistent).
-------------------------------------------------------------

An array can be 'created' by describing the array (level, chunksize
etc) in a SET_ARRAY_INFO ioctl.  This must has major_version==0 and
raid_disks != 0.

Then uninitialized devices can be added with ADD_NEW_DISK.  The
structure passed to ADD_NEW_DISK must specify the state of the device
and it's role in the array.

Once started with RUN_ARRAY, uninitialized spares can be added with
HOT_ADD_DISK.


MD devices in sysfs
-------------------
md devices appear in sysfs (/sys) as regular block devices,
e.g.
   /sys/block/md0

Each 'md' device will contain a subdirectory called 'md' which
contains further md-specific information about the device.

All md devices contain:
  level
     a text file indicating the 'raid level'. e.g. raid0, raid1,
     raid5, linear, multipath, faulty.
     If no raid level has been set yet (array is still being
     assembled), the value will reflect whatever has been written
     to it, which may be a name like the above, or may be a number
     such as '0', '5', etc.

  raid_disks
     a text file with a simple number indicating the number of devices
     in a fully functional array.  If this is not yet known, the file
     will be empty.  If an array is being resized (not currently
     possible) this will contain the larger of the old and new sizes.
     Some raid level (RAID1) allow this value to be set while the
     array is active.  This will reconfigure the array.   Otherwise
     it can only be set while assembling an array.

  chunk_size
     This is the size if bytes for 'chunks' and is only relevant to
     raid levels that involve striping (1,4,5,6,10). The address space
     of the array is conceptually divided into chunks and consecutive
     chunks are striped onto neighbouring devices.
     The size should be at least PAGE_SIZE (4k) and should be a power
     of 2.  This can only be set while assembling an array

  layout
     The "layout" for the array for the particular level.  This is
     simply a number that is interpretted differently by different
     levels.  It can be written while assembling an array.

  reshape_position
     This is either "none" or a sector number within the devices of
     the array where "reshape" is up to.  If this is set, the three
     attributes mentioned above (raid_disks, chunk_size, layout) can
     potentially have 2 values, an old and a new value.  If these
     values differ, reading the attribute returns
        new (old)
     and writing will effect the 'new' value, leaving the 'old'
     unchanged.

  component_size
     For arrays with data redundancy (i.e. not raid0, linear, faulty,
     multipath), all components must be the same size - or at least
     there must a size that they all provide space for.  This is a key
     part or the geometry of the array.  It is measured in sectors
     and can be read from here.  Writing to this value may resize
     the array if the personality supports it (raid1, raid5, raid6),
     and if the component drives are large enough.

  metadata_version
     This indicates the format that is being used to record metadata
     about the array.  It can be 0.90 (traditional format), 1.0, 1.1,
     1.2 (newer format in varying locations) or "none" indicating that
     the kernel isn't managing metadata at all.

  resync_start
     The point at which resync should start.  If no resync is needed,
     this will be a very large number.  At array creation it will
     default to 0, though starting the array as 'clean' will
     set it much larger.

   new_dev
     This file can be written but not read.  The value written should
     be a block device number as major:minor.  e.g. 8:0
     This will cause that device to be attached to the array, if it is
     available.  It will then appear at md/dev-XXX (depending on the
     name of the device) and further configuration is then possible.

   safe_mode_delay
     When an md array has seen no write requests for a certain period
     of time, it will be marked as 'clean'.  When another write
     request arrives, the array is marked as 'dirty' before the write
     commences.  This is known as 'safe_mode'.
     The 'certain period' is controlled by this file which stores the
     period as a number of seconds.  The default is 200msec (0.200).
     Writing a value of 0 disables safemode.

   array_state
     This file contains a single word which describes the current
     state of the array.  In many cases, the state can be set by
     writing the word for the desired state, however some states
     cannot be explicitly set, and some transitions are not allowed.

     Select/poll works on this file.  All changes except between
     	active_idle and active (which can be frequent and are not
	very interesting) are notified.  active->active_idle is
	reported if the metadata is externally managed.

     clear
         No devices, no size, no level
         Writing is equivalent to STOP_ARRAY ioctl
     inactive
         May have some settings, but array is not active
            all IO results in error
         When written, doesn't tear down array, but just stops it
     suspended (not supported yet)
         All IO requests will block. The array can be reconfigured.
         Writing this, if accepted, will block until array is quiessent
     readonly
         no resync can happen.  no superblocks get written.
         write requests fail
     read-auto
         like readonly, but behaves like 'clean' on a write request.

     clean - no pending writes, but otherwise active.
         When written to inactive array, starts without resync
         If a write request arrives then
           if metadata is known, mark 'dirty' and switch to 'active'.
           if not known, block and switch to write-pending
         If written to an active array that has pending writes, then fails.
     active
         fully active: IO and resync can be happening.
         When written to inactive array, starts with resync

     write-pending
         clean, but writes are blocked waiting for 'active' to be written.

     active-idle
         like active, but no writes have been seen for a while (safe_mode_delay).


As component devices are added to an md array, they appear in the 'md'
directory as new directories named
      dev-XXX
where XXX is a name that the kernel knows for the device, e.g. hdb1.
Each directory contains:

      block
        a symlink to the block device in /sys/block, e.g.
	     /sys/block/md0/md/dev-hdb1/block -> ../../../../block/hdb/hdb1

      super
        A file containing an image of the superblock read from, or
        written to, that device.

      state
        A file recording the current state of the device in the array
	which can be a comma separated list of
	      faulty   - device has been kicked from active use due to
                         a detected fault
	      in_sync  - device is a fully in-sync member of the array
	      writemostly - device will only be subject to read
		         requests if there are no other options.
			 This applies only to raid1 arrays.
	      blocked  - device has failed, metadata is "external",
	                 and the failure hasn't been acknowledged yet.
			 Writes that would write to this device if
			 it were not faulty are blocked.
	      spare    - device is working, but not a full member.
			 This includes spares that are in the process
			 of being recovered to
	This list may grow in future.
	This can be written to.
	Writing "faulty"  simulates a failure on the device.
	Writing "remove" removes the device from the array.
	Writing "writemostly" sets the writemostly flag.
	Writing "-writemostly" clears the writemostly flag.
	Writing "blocked" sets the "blocked" flag.
	Writing "-blocked" clear the "blocked" flag and allows writes
		to complete.

	This file responds to select/poll. Any change to 'faulty'
	or 'blocked' causes an event.

      errors
	An approximate count of read errors that have been detected on
	this device but have not caused the device to be evicted from
	the array (either because they were corrected or because they
	happened while the array was read-only).  When using version-1
	metadata, this value persists across restarts of the array.

	This value can be written while assembling an array thus
	providing an ongoing count for arrays with metadata managed by
	userspace.

      slot
        This gives the role that the device has in the array.  It will
	either be 'none' if the device is not active in the array
        (i.e. is a spare or has failed) or an integer less than the
	'raid_disks' number for the array indicating which position
	it currently fills.  This can only be set while assembling an
	array.  A device for which this is set is assumed to be working.

      offset
        This gives the location in the device (in sectors from the
        start) where data from the array will be stored.  Any part of
        the device before this offset us not touched, unless it is
        used for storing metadata (Formats 1.1 and 1.2).

      size
        The amount of the device, after the offset, that can be used
        for storage of data.  This will normally be the same as the
	component_size.  This can be written while assembling an
        array.  If a value less than the current component_size is
        written, it will be rejected.


An active md device will also contain and entry for each active device
in the array.  These are named

    rdNN

where 'NN' is the position in the array, starting from 0.
So for a 3 drive array there will be rd0, rd1, rd2.
These are symbolic links to the appropriate 'dev-XXX' entry.
Thus, for example,
       cat /sys/block/md*/md/rd*/state
will show 'in_sync' on every line.


Active md devices for levels that support data redundancy (1,4,5,6)
also have

   sync_action
     a text file that can be used to monitor and control the rebuild
     process.  It contains one word which can be one of:
       resync        - redundancy is being recalculated after unclean
                       shutdown or creation
       recover       - a hot spare is being built to replace a
                       failed/missing device
       idle          - nothing is happening
       check         - A full check of redundancy was requested and is
                       happening.  This reads all block and checks
                       them. A repair may also happen for some raid
                       levels.
       repair        - A full check and repair is happening.  This is
                       similar to 'resync', but was requested by the
                       user, and the write-intent bitmap is NOT used to
		       optimise the process.

      This file is writable, and each of the strings that could be
      read are meaningful for writing.

       'idle' will stop an active resync/recovery etc.  There is no
           guarantee that another resync/recovery may not be automatically
	   started again, though some event will be needed to trigger
           this.
	'resync' or 'recovery' can be used to restart the
           corresponding operation if it was stopped with 'idle'.
	'check' and 'repair' will start the appropriate process
           providing the current state is 'idle'.

      This file responds to select/poll.  Any important change in the value
      triggers a poll event.  Sometimes the value will briefly be
      "recover" if a recovery seems to be needed, but cannot be
      achieved. In that case, the transition to "recover" isn't
      notified, but the transition away is.

   degraded
      This contains a count of the number of devices by which the
      arrays is degraded.  So an optimal array with show '0'.  A
      single failed/missing drive will show '1', etc.
      This file responds to select/poll, any increase or decrease
      in the count of missing devices will trigger an event.

   mismatch_count
      When performing 'check' and 'repair', and possibly when
      performing 'resync', md will count the number of errors that are
      found.  The count in 'mismatch_cnt' is the number of sectors
      that were re-written, or (for 'check') would have been
      re-written.  As most raid levels work in units of pages rather
      than sectors, this my be larger than the number of actual errors
      by a factor of the number of sectors in a page.

   bitmap_set_bits
      If the array has a write-intent bitmap, then writing to this
      attribute can set bits in the bitmap, indicating that a resync
      would need to check the corresponding blocks. Either individual
      numbers or start-end pairs can be written.  Multiple numbers
      can be separated by a space.
      Note that the numbers are 'bit' numbers, not 'block' numbers.
      They should be scaled by the bitmap_chunksize.

   sync_speed_min
   sync_speed_max
     This are similar to /proc/sys/dev/raid/speed_limit_{min,max}
     however they only apply to the particular array.
     If no value has been written to these, of if the word 'system'
     is written, then the system-wide value is used.  If a value,
     in kibibytes-per-second is written, then it is used.
     When the files are read, they show the currently active value
     followed by "(local)" or "(system)" depending on whether it is
     a locally set or system-wide value.

   sync_completed
     This shows the number of sectors that have been completed of
     whatever the current sync_action is, followed by the number of
     sectors in total that could need to be processed.  The two
     numbers are separated by a '/'  thus effectively showing one
     value, a fraction of the process that is complete.
     A 'select' on this attribute will return when resync completes,
     when it reaches the current sync_max (below) and possibly at
     other times.

   sync_max
     This is a number of sectors at which point a resync/recovery
     process will pause.  When a resync is active, the value can
     only ever be increased, never decreased.  The value of 'max'
     effectively disables the limit.


   sync_speed
     This shows the current actual speed, in K/sec, of the current
     sync_action.  It is averaged over the last 30 seconds.

   suspend_lo
   suspend_hi
     The two values, given as numbers of sectors, indicate a range
     within the array where IO will be blocked.  This is currently
     only supported for raid4/5/6.


Each active md device may also have attributes specific to the
personality module that manages it.
These are specific to the implementation of the module and could
change substantially if the implementation changes.

These currently include

  stripe_cache_size  (currently raid5 only)
      number of entries in the stripe cache.  This is writable, but
      there are upper and lower limits (32768, 16).  Default is 128.
  strip_cache_active (currently raid5 only)
      number of active entries in the stripe cache
  preread_bypass_threshold (currently raid5 only)
      number of times a stripe requiring preread will be bypassed by
      a stripe that does not require preread.  For fairness defaults
      to 1.  Setting this to 0 disables bypass accounting and
      requires preread stripes to wait until all full-width stripe-
      writes are complete.  Valid values are 0 to stripe_cache_size.