FDISK(8)

FDISK(8)           Linux Programmer’s Manual              FDISK(8)



NAME
       fdisk - Partition table manipulator for Linux

SYNOPSIS
       fdisk [-uc] [-b sectorsize] [-C cyls] [-H heads] [-S sects] device

       fdisk -l [-u] [device...]

       fdisk -s partition...

       fdisk -v

       fdisk -h

DESCRIPTION
       Hard  disks can be divided into one or more logical disks called parti-
       tions.  This division is described in the partition table found in sec-
       tor 0 of the disk.

       In the BSD world one talks about ‘disk slices’ and a ‘disklabel’.

       Linux  needs  at     least one partition, namely for its root file system.
       It can use swap files and/or swap partitions, but the latter  are  more
       efficient. So, usually one will want a second Linux partition dedicated
       as swap partition.  On Intel compatible hardware, the BIOS  that     boots
       the  system can often only access the first 1024 cylinders of the disk.
       For this reason people with large disks often create a third partition,
       just  a    few  MB large, typically mounted on /boot, to store the kernel
       image and a few auxiliary files needed at boot time, so as to make sure
       that  this  stuff  is  accessible to the BIOS.  There may be reasons of
       security, ease of administration and backup, or testing,     to  use  more
       than the minimum number of partitions.

       fdisk  (in  the    first form of invocation) is a menu driven program for
       creation and manipulation of partition tables.  It understands DOS type
       partition tables and BSD or SUN type disklabels.

       fdisk  doesn’t  understand  GUID     Partition  Table  (GPT) and it is not
       designed for large partitions. In particular case use more advanced GNU
       parted(8).

       The device is usually /dev/sda, /dev/sdb or so. A device name refers to
       the entire disk.     The old systems without libata (a library used inside
       the  Linux  kernel  to support ATA host controllers and devices) make a
       difference between IDE and SCSI disks. In such a case the  device  name
       will be /dev/hd* (IDE) or /dev/sd* (SCSI).

       The  partition  is  a  device name followed by a partition number.  For
       example, /dev/sda1 is the first partition on the first hard disk in the
       system.      See    also   Linux   kernel  documentation  (the  Documenta-
       tion/devices.txt file).

       A BSD/SUN type disklabel can describe 8 partitions, the third of     which
       should  be  a  ‘whole  disk’  partition.     Do not start a partition that
       actually uses its first sector (like a swap partition) at  cylinder  0,
       since that will destroy the disklabel.

       An  IRIX/SGI type disklabel can describe 16 partitions, the eleventh of
       which should be an entire ‘volume’ partition, while the ninth should be
       labeled    ‘volume header’.  The volume header will also cover the parti-
       tion table, i.e., it starts at block zero and extends by     default  over
       five  cylinders.      The remaining space in the volume header may be used
       by header directory entries.  No partitions may overlap with the volume
       header.     Also  do not change its type and make some file system on it,
       since you will lose the partition table.     Use this type of  label  only
       when  working  with  Linux on IRIX/SGI machines or IRIX/SGI disks under
       Linux.

       A DOS type partition table can describe an unlimited number  of    parti-
       tions.  In  sector  0 there is room for the description of 4 partitions
       (called ‘primary’). One of these may be an extended partition; this  is
       a  box  holding    logical partitions, with descriptors found in a linked
       list of sectors, each preceding the corresponding  logical  partitions.
       The  four primary partitions, present or not, get numbers 1-4.  Logical
       partitions start numbering from 5.

       In a DOS type partition table the starting offset and the size of  each
       partition  is  stored  in  two  ways:  as an absolute number of sectors
       (given in 32 bits) and as a Cylinders/Heads/Sectors  triple  (given  in
       10+8+6  bits).  The former is OK - with 512-byte sectors this will work
       up to 2 TB. The latter has two different problems. First of all,     these
       C/H/S fields can be filled only when the number of heads and the number
       of sectors per track are known. Secondly, even if we  know  what     these
       numbers    should be, the 24 bits that are available do not suffice.  DOS
       uses C/H/S only, Windows uses both, Linux never uses C/H/S.

       If possible, fdisk will obtain the disk geometry     automatically.      This
       is  not necessarily the physical disk geometry (indeed, modern disks do
       not really have anything like a physical geometry, certainly not     some-
       thing  that  can     be  described    in  simplistic Cylinders/Heads/Sectors
       form), but is the disk geometry that MS-DOS uses for the partition  ta-
       ble.

       Usually all goes well by default, and there are no problems if Linux is
       the only system on the disk. However, if the disk has to be shared with
       other  operating     systems, it is often a good idea to let an fdisk from
       another operating system make at least one partition. When Linux     boots
       it looks at the partition table, and tries to deduce what (fake) geome-
       try is required for good cooperation with other systems.

       Whenever a partition table is printed out, a consistency check is  per-
       formed  on  the    partition table entries.  This check verifies that the
       physical and logical start and end points are identical, and  that  the
       partition  starts and ends on a cylinder boundary (except for the first
       partition).

       Some versions of MS-DOS create a first partition which does  not     begin
       on  a cylinder boundary, but on sector 2 of the first cylinder.    Parti-
       tions beginning in cylinder 1 cannot begin on a cylinder boundary,  but
       this  is     unlikely  to  cause  difficulty  unless you have OS/2 on your
       machine.

       A sync() and a BLKRRPART ioctl() (reread partition table from disk) are
       performed  before  exiting  when     the partition table has been updated.
       Long ago it used to be necessary to reboot after the use of  fdisk.   I
       do  not    think this is the case anymore - indeed, rebooting too quickly
       might cause loss of not-yet-written data. Note that both the kernel and
       the disk hardware may buffer data.


DOS 6.x WARNING
       The DOS 6.x FORMAT command looks for some information in the first sec-
       tor of the data area of the partition, and treats this  information  as
       more  reliable than the information in the partition table.  DOS FORMAT
       expects DOS FDISK to clear the first 512 bytes of the data  area     of  a
       partition  whenever a size change occurs.  DOS FORMAT will look at this
       extra information even if the /U flag is given -- we  consider  this  a
       bug in DOS FORMAT and DOS FDISK.

       The  bottom  line is that if you use cfdisk or fdisk to change the size
       of a DOS partition table entry, then you must also use dd to  zero  the
       first 512 bytes of that partition before using DOS FORMAT to format the
       partition.  For example, if you were using cfdisk to make a DOS    parti-
       tion table entry for /dev/sda1, then (after exiting fdisk or cfdisk and
       rebooting Linux so that the partition table information is  valid)  you
       would  use the command "dd if=/dev/zero of=/dev/sda1 bs=512 count=1" to
       zero the first 512 bytes of the partition.

       BE EXTREMELY CAREFUL if you use the dd command, since a small typo  can
       make all of the data on your disk useless.

       For  best results, you should always use an OS-specific partition table
       program.     For example, you should make  DOS  partitions    with  the  DOS
       FDISK program and Linux partitions with the Linux fdisk or Linux cfdisk
       program.


OPTIONS
       -b sectorsize
          Specify the sector size of the disk. Valid values are 512, 1024,
          2048  or    4096.    (Recent kernels know the sector size. Use this
          only on old kernels or to override the  kernel’s    ideas.)     Since
          util-linux-ng  2.17  fdisk  differentiates  between  logical and
          physical sector size. This option changes both sector  sizes  to
          sectorsize.

       -h     Print help and then exit.

       -c     Switch off DOS-compatible mode. (Recommended)

       -C cyls
          Specify the number of cylinders of the disk.  I have no idea why
          anybody would want to do so.

       -H heads
          Specify the number of heads of the disk. (Not the physical  num-
          ber, of course, but the number used for partition tables.)  Rea-
          sonable values are 255 and 16.

       -S sects
          Specify the number of sectors per track of the disk.   (Not  the
          physical    number,     of  course, but the number used for partition
          tables.)    A reasonable value is 63.

       -l     List the partition tables for the     specified  devices  and  then
          exit.   If no devices are given, those mentioned in /proc/parti-
          tions (if that exists) are used.

       -u     When listing partition tables, give sizes in sectors instead  of
          cylinders.

       -s partition
          The size of the partition (in blocks) is printed on the standard
          output.

       -v     Print version number of fdisk program and exit.

BUGS
       There are several *fdisk programs around.  Each has  its     problems  and
       strengths.   Try     them  in  the    order cfdisk, fdisk, sfdisk.  (Indeed,
       cfdisk is a beautiful program that has strict requirements on the  par-
       tition  tables  it accepts, and produces high quality partition tables.
       Use it if you can.  fdisk is a buggy program that does fuzzy  things  -
       usually    it happens to produce reasonable results. Its single advantage
       is that it has some support for BSD disk labels and other non-DOS  par-
       tition  tables.    Avoid it if you can.  sfdisk is for hackers only - the
       user interface is terrible, but it is more correct than fdisk and  more
       powerful     than  both fdisk and cfdisk.  Moreover, it can be used nonin-
       teractively.)

       These days there also is parted.     The cfdisk interface  is  nicer,  but
       parted  does  much  more:  it not only resizes partitions, but also the
       filesystems that live in them.

       The IRIX/SGI type disklabel is currently not supported by  the  kernel.
       Moreover, IRIX/SGI header directories are not fully supported yet.

       The option ‘dump partition table to file’ is missing.

SEE ALSO
       cfdisk(8), sfdisk(8), mkfs(8), parted(8), partprobe(8), kpartx(8)

AVAILABILITY
       The fdisk command is part of the util-linux-ng package and is available
       from ftp://ftp.kernel.org/pub/linux/utils/util-linux-ng/.



Linux 2.0             11 June 1998                  FDISK(8)

你可能感兴趣的:(FDISK(8))