Re: Questions : disk partition re-reading

[Rogier Wolff <R.E.Wolff@BitWizard.nl>]

On Tue, Apr 20, 2004 at 09:46:50AM +0200, Andries Brouwer wrote:
> You must distinguish the on-disk partition table and the kernel
> partition table. You can change the on-disk partition table just by
> writing to it, but that does not change the kernel's ideas.
> You can change the kernel partition table using the right ioctls
> but that does not change the bits on disk.
> 
> Usually one does
> 	blockdev --rereadpt /dev/something
> after changing media, or after writing to the partition table,
> but that will fail if the disk is busy.

Which reminds me: Too bad the kernel says "Busy" without really
thinking aboutit. 

Assume that I have a 
	<hda1> swap 1G    (*)
	<hda2> root 19G   (active)
        <free space> 

disk, and want to add an extra partition. Maybe I have a 20G image
which I want to copy onto 20, 40, 60, 80, 120 and 200Gb disks. I can
then add the third partition as a /data or a /tmp partition. However
as the active root is on that disk I have to reboot. But in fact I
don't intend to change the active partition. 

So, there should be objections if I want the new partitioning scheme
to be: 

	<hda1> swap 1.5G 
	<hda2> root 19G   (active)
        <free space> 

then there is a problem and I can understand "Busy".  But if the new
scheme is: 

	<hda1> swap 1G  
	<hda2> root 19G   (active)
        <hda3> data 20G   <unformatted>

then I don't understand the reason for refusing the rereadpt request.

Anybody want to code this up?


	Roger. 


(*) usually active as well, but easily deactivated. So lets pretend
it's inactive for this discussion. 

-- 
** R.E.Wolff@BitWizard.nl ** http://www.BitWizard.nl/ ** +31-15-2600998 **
*-- BitWizard writes Linux device drivers for any device you may have! --*
**** "Linux is like a wigwam -  no windows, no gates, apache inside!" ****