This is a discussion on How are partitions numbered? within the Linux General forums, part of the Linux Forums category; Are partitions numbered strictly sequentially under Linux or in the order in which they were created? Does it depend on ...
|
|||||||
| FAQ | Members List | Calendar | Search | Today's Posts | Mark Forums Read |
|
|||
|
Are partitions numbered strictly sequentially under Linux or in the order
in which they were created? Does it depend on which tool was used to create them, e.g., fdisk, FDISK.EXE, Partition Magic? For example, if I initially partition a drive as: Partition 1 *Free space* Partition 2 but later go back and partition the Free space, would I get: Partition 1 Partition 2 Partition 3 or Partition 1 Partition 3 Partition 2 Would the same answer(s) be true for logical partitions in an extended partition? Thanks for your help. |
|
|||
|
Charles Sullivan wrote:
> > Are partitions numbered strictly sequentially under Linux or in the order > in which they were created? Generally, in the order in which they were created unless your partitioning tool asks you which number partition you're creating. Linux fdisk will ask you for a partition number when you create primary partitions but not for logical drives in the extended partition (which are always sequential). DOS fdisk is always sequential, IIRC. > Does it depend on which tool was used to > create them, e.g., fdisk, FDISK.EXE, Partition Magic? > > For example, if I initially partition a drive as: > Partition 1 > *Free space* > Partition 2 > > but later go back and partition the Free space, would I get: > Partition 1 > Partition 2 > Partition 3 > > or > Partition 1 > Partition 3 > Partition 2 The second layout: 1-3-2. Note that in those circumstances, fdisk will issue a warning about "partitions not in disk order" or some such. It's just an informational message, though. > Would the same answer(s) be true for logical partitions in > an extended partition? Yes, in general. See my note above. |
|
|||
|
Charles Sullivan <cwsulliv@triad.rr.com> writes:
> Are partitions numbered strictly sequentially under Linux or in the order > in which they were created? Does it depend on which tool was used to > create them, e.g., fdisk, FDISK.EXE, Partition Magic? Under Linux, the IDE disks are called hd[a-z], and the scsi disks are called sd[a-z]. So, /dev/hdb refers to the slave disk on the first IDE channel, /dev/hdc refers to the master disk on the second IDE channel, etc. It is entirely possible that you have /dev/hda and /dev/hdd but not /dev/hdb and /dev/hdc. It means that you have a primary master and a secondary slave drive. The scsi disks are named in a bit different matter. Say, you have three scsi cards, and on the first one you have one, on the second you have two, and on the third, you have one disk. So /dev/sda will be the sole disk on the first scsi card, /dev/sdb will be the scsi card with the lower scsi id on the second card, /dev/sdc will be the second (with a higher scsi id) disk on the second card, and /dev/sdd will be the sole drive on the third scsi card. The problem here is that if you remove a disk which is not the very last, then all the disks beyond that will cascade down and what used to be /dev/sdc can become /dev/sdb, etc. So you now know how the disks are named. In the PC world, the first sector of the disk is called the Master Boot Record, MBR in short. Inside this MBR is the boot loader (which is relevant if you boot from the disk), and the partition table. The partition table is 64 bytes long, and one entry inside the partition table is 16 bytes long. Therefore, you can have four primary partitions (16*4=64) on a single disk. One for each entry. Suppose we are talking about the primary slave IDE disk, which is /dev/hdb. BTW /dev/hdb refers to the whole disk. So the first entry in the partition table is /dev/hdb1, the second is /dev/hdb2, the third is /dev/hdb3, and the fourth is /dev/hdb4. It is entirely possible that there is no entry in the partition table for /dev/hdb2, the second slot, and in this case you would have /dev/hdb1, /dev/hdb3, and /dev/hdb4. One valid entry in the partition table is the so called extended partition type. In that case, the area referred by that entry can have sub partitions which are called logical partitions. They start with the number 5 (remember, 1-4 are for primary partitions), and they are numbered sequentially. The first logical partition will be /dev/hdb5, the second will be /dev/hdb6, etc. From this, it should be clear that the partition numbering scheme depends on the partitioning software but only in the way in which order the software fills the partition table. If it fills the first slot first, then you have /dev/hdb1, then /dev/hdb2, etc. If it fills backwards, then you have /dev/hdb4 first, then /dev/hdb3, etc. It is also possible to have a hole in the partition table, so you might have /dev/hdb1 and /dev/hdb3. Also, if you have only a single primary partition on a disk, and this is an extended partition, and inside that extended partition you have a single logical partition, then you will have /dev/hdb1 and /dev/hdb5 on the disk. > For example, if I initially partition a drive as: > Partition 1 > *Free space* > Partition 2 > > but later go back and partition the Free space, would I get: > Partition 1 > Partition 2 > Partition 3 > > or > Partition 1 > Partition 3 > Partition 2 It entirely depends on the order the entries in the partition table point to the disk areas. > Would the same answer(s) be true for logical partitions in > an extended partition? AFAIK, if you erase a logical partition, then the ones above it will cascade down, but I am not sure since I didn't play with many logical partitions for years. Vilmos |
|
|||
|
On Sat, 28 Jun 2003 18:25:29 -0700, Vilmos Soti wrote:
[snip] > The scsi disks are named in a bit different matter. Say, you > have three scsi cards, and on the first one you have one, on > the second you have two, and on the third, you have one disk. > So /dev/sda will be the sole disk on the first scsi card, > /dev/sdb will be the scsi card with the lower scsi id on the > second card, /dev/sdc will be the second (with a higher scsi > id) disk on the second card, and /dev/sdd will be the sole > drive on the third scsi card. The problem here is that if > you remove a disk which is not the very last, then all the > disks beyond that will cascade down and what used to be /dev/sdc > can become /dev/sdb, etc. devfs can help alleviate this - my ide-scsi cdrom is, for example: lr-xr-xr-x 1 root root 31 Jun 27 16:54 /dev/sr0 -> scsi/host0/bus0/target0/lun0/cd As long as you use the overly verbose devfs filenames (or a symlink thereof), they should remain unchanged (I think). -- Freenet distribution not available If the rich could pay the poor to die for them, what a living the poor could make! |
|
|||
|
Vilmos Soti <vilmos@vilmos.org>,
In a message on 28 Jun 2003 18:25:29 -0700, wrote : VS> Charles Sullivan <cwsulliv@triad.rr.com> writes: VS> VS> > Are partitions numbered strictly sequentially under Linux or in the order VS> > in which they were created? Does it depend on which tool was used to VS> > create them, e.g., fdisk, FDISK.EXE, Partition Magic? VS> VS> Under Linux, the IDE disks are called hd[a-z], and the scsi disks VS> are called sd[a-z]. So, /dev/hdb refers to the slave disk on the VS> first IDE channel, /dev/hdc refers to the master disk on the VS> second IDE channel, etc. It is entirely possible that you have VS> /dev/hda and /dev/hdd but not /dev/hdb and /dev/hdc. It means that VS> you have a primary master and a secondary slave drive. I don't believe the IDE interfaces allow you to have slaves without masters -- you can have masters without slaves. VS> VS> The scsi disks are named in a bit different matter. Say, you VS> have three scsi cards, and on the first one you have one, on VS> the second you have two, and on the third, you have one disk. VS> So /dev/sda will be the sole disk on the first scsi card, VS> /dev/sdb will be the scsi card with the lower scsi id on the VS> second card, /dev/sdc will be the second (with a higher scsi VS> id) disk on the second card, and /dev/sdd will be the sole VS> drive on the third scsi card. The problem here is that if VS> you remove a disk which is not the very last, then all the VS> disks beyond that will cascade down and what used to be /dev/sdc VS> can become /dev/sdb, etc. This is what Ext2 and Ext3 file system labels are for... :-) VS> VS> So you now know how the disks are named. In the PC world, VS> the first sector of the disk is called the Master Boot Record, VS> MBR in short. Inside this MBR is the boot loader (which is VS> relevant if you boot from the disk), and the partition table. VS> The partition table is 64 bytes long, and one entry inside VS> the partition table is 16 bytes long. Therefore, you can have VS> four primary partitions (16*4=64) on a single disk. One for VS> each entry. VS> VS> Suppose we are talking about the primary slave IDE disk, which VS> is /dev/hdb. BTW /dev/hdb refers to the whole disk. So the VS> first entry in the partition table is /dev/hdb1, the second VS> is /dev/hdb2, the third is /dev/hdb3, and the fourth is /dev/hdb4. VS> VS> It is entirely possible that there is no entry in the partition VS> table for /dev/hdb2, the second slot, and in this case you would VS> have /dev/hdb1, /dev/hdb3, and /dev/hdb4. One valid entry in the VS> partition table is the so called extended partition type. In VS> that case, the area referred by that entry can have sub partitions VS> which are called logical partitions. They start with the number VS> 5 (remember, 1-4 are for primary partitions), and they are VS> numbered sequentially. The first logical partition will be VS> /dev/hdb5, the second will be /dev/hdb6, etc. VS> VS> From this, it should be clear that the partition numbering VS> scheme depends on the partitioning software but only in the VS> way in which order the software fills the partition table. VS> If it fills the first slot first, then you have /dev/hdb1, VS> then /dev/hdb2, etc. If it fills backwards, then you have VS> /dev/hdb4 first, then /dev/hdb3, etc. It is also possible VS> to have a hole in the partition table, so you might have VS> /dev/hdb1 and /dev/hdb3. VS> VS> Also, if you have only a single primary partition on a disk, VS> and this is an extended partition, and inside that extended VS> partition you have a single logical partition, then you VS> will have /dev/hdb1 and /dev/hdb5 on the disk. VS> VS> > For example, if I initially partition a drive as: VS> > Partition 1 VS> > *Free space* VS> > Partition 2 VS> > VS> > but later go back and partition the Free space, would I get: VS> > Partition 1 VS> > Partition 2 VS> > Partition 3 VS> > VS> > or VS> > Partition 1 VS> > Partition 3 VS> > Partition 2 VS> VS> It entirely depends on the order the entries in the partition VS> table point to the disk areas. VS> VS> > Would the same answer(s) be true for logical partitions in VS> > an extended partition? VS> VS> AFAIK, if you erase a logical partition, then the ones above VS> it will cascade down, but I am not sure since I didn't play VS> with many logical partitions for years. VS> VS> Vilmos VS> |
|
|||
|
On Sat, 28 Jun 2003 21:25:29 -0400, Vilmos Soti wrote:
> Charles Sullivan <cwsulliv@triad.rr.com> writes: > >> Are partitions numbered strictly sequentially under Linux or in the >> order in which they were created? Does it depend on which tool was >> used to create them, e.g., fdisk, FDISK.EXE, Partition Magic? > <snip> > From this, it should be clear that the partition numbering scheme > depends on the partitioning software but only in the way in which order > the software fills the partition table. If it fills the first slot > first, then you have /dev/hdb1, then /dev/hdb2, etc. If it fills > backwards, then you have /dev/hdb4 first, then /dev/hdb3, etc. It is > also possible to have a hole in the partition table, so you might have > /dev/hdb1 and /dev/hdb3. <snip> OK, I think I got it: The partitions are numbered by Linux according to their order in the partition table, but the order of the actual partitioned spaces on a drive may or may not be the same, depending on the software used to create the partitions. Thanks Vilmos. Regards, Charles Sullivan |
|
|||
|
On Sat, 28 Jun 2003 22:25:04 -0400, Robert Heller wrote:
<snip> > I don't believe the IDE interfaces allow you to have slaves without > masters -- you can have masters without slaves. <snip> I think this may depend on the particular hardware and bios. I have my two HDDs as Master/Slave on a Promise UDMA 66 add-on card. If I unplug the master, the bios will (after a timeout) boot from the slave. I notice that some HDDs have one jumper position for both Single drive or Master in a Master/Slave configuration, whereas others have separate jumper positions for these two cases. This may make a difference. Because of flaky behavior of my particular (Phoenix) bios when all 4 onboard IDE slots are filled, I moved an ATAPI ZIP drive to the secondary channel of the Promise card. I neglected to change the ZIP drive jumper from slave to master when I did this but the drive worked fine as slave even though there was no master on that channel. |
![]() |
| Thread Tools | |
| Display Modes | |
|
|