Possibly silly Q?

Possibly silly Q?

[Gene Heskett]

Greetings;

Do we have a utility that can force the kernel to re-read, and re-initialize 
itself to a given drives partition tables without having to reboot if one is 
working with a drive that is not part of the required kernel directory tree?

The reason I ask, is that I've just spent about 20 hours fighting with amanda 
after repartitioning a drive for use by amanda as virtual tapes, 
making /dev/sdc1 into a primary partition of nearly 400GB.  /dev/sdc1 started 
out as a /boot2 labeled partition of 200mb, and because the kernels data was 
stale, amanda upchucked all over itself when the kernel thought the original 
200Mb had been filled when it had been umounted from /mnt/bootf8 
repartitioned, mke2fs'd, a journal added and a new label written and then 
mounted to /amandatapes.

Something it seems to me, should have forced the re-init, but didn't.  So is 
there a tool that can force that?

Thanks

-- 
Cheers, Gene
"There are four boxes to be used in defense of liberty:
 soap, ballot, jury, and ammo. Please use in that order."
-Ed Howdershelt (Author)
Your mode of life will be changed for the better because of new developments.

Re: Possibly silly Q?

[Robert P. J. Day]

On Mon, 14 Jan 2008, Gene Heskett wrote:

> Greetings;
>
> Do we have a utility that can force the kernel to re-read, and
> re-initialize itself to a given drives partition tables without
> having to reboot if one is working with a drive that is not part of
> the required kernel directory tree?

i would try "partprobe".

rday
--
========================================================================
Robert P. J. Day
Linux Consulting, Training and Annoying Kernel Pedantry
Waterloo, Ontario, CANADA

Home page:                                         http://crashcourse.ca
Fedora Cookbook:    http://crashcourse.ca/wiki/index.php/Fedora_Cookbook
========================================================================

Re: Possibly silly Q?

[Sven-Haegar Koch]

On Mon, 14 Jan 2008, Robert P. J. Day wrote:

> On Mon, 14 Jan 2008, Gene Heskett wrote:
> 
> > Greetings;
> >
> > Do we have a utility that can force the kernel to re-read, and
> > re-initialize itself to a given drives partition tables without
> > having to reboot if one is working with a drive that is not part of
> > the required kernel directory tree?
> 
> i would try "partprobe".

or "sfdisk -R /dev/sdX"

c'ya
sven

-- 

The Internet treats censorship as a routing problem, and routes around
it. (John Gilmore on http://www.cygnus.com/~gnu/)

Re: Possibly silly Q?

[Jan Engelhardt]

On Jan 14 2008 14:34, Gene Heskett wrote:
>
>Do we have a utility that can force the kernel to re-read, and re-initialize 
>itself to a given drives partition tables without having to reboot if one is 
>working with a drive that is not part of the required kernel directory tree?

fdisk issues an ioctl to reread the drive after you issued 'w'(rite).
Works.

Re: Possibly silly Q?

[Dagfinn Ilmari Mannsåker]

Gene Heskett <gene.heskett@gmail.com> writes:

> Greetings;
>
> Do we have a utility that can force the kernel to re-read, and re-initialize 
> itself to a given drives partition tables without having to reboot if one is 
> working with a drive that is not part of the required kernel directory tree?
[…]
> Something it seems to me, should have forced the re-init, but didn't.  So is 
> there a tool that can force that?

fdisk or similar should have issued an ioctl to reread the partition
table after writing the new one, but you can do it manually with
'blockdev --rereadpt <device>'.

-- 
ilmari
"A disappointingly low fraction of the human race is,
 at any given time, on fire." - Stig Sandbeck Mathisen

Re: Possibly silly Q?

[Robert P. J. Day]

[-- Attachment #1: Type: TEXT/PLAIN, Size: 1325 bytes --]

On Mon, 14 Jan 2008, Dagfinn Ilmari Mannsåker wrote:

> Gene Heskett <gene.heskett@gmail.com> writes:
>
> > Greetings;
> >
> > Do we have a utility that can force the kernel to re-read, and re-initialize
> > itself to a given drives partition tables without having to reboot if one is
> > working with a drive that is not part of the required kernel directory tree?
> […]
> > Something it seems to me, should have forced the re-init, but didn't.  So is
> > there a tool that can force that?
>
> fdisk or similar should have issued an ioctl to reread the partition
> table after writing the new one, but you can do it manually with
> 'blockdev --rereadpt <device>'.

i remember bringing up this very issue quite some time ago and, IIRC,
the consensus was that *primary* partition changes would be re-read by
the kernel, but not *logical* partition changes.  or something sort of
like that.

rday
--

========================================================================
Robert P. J. Day
Linux Consulting, Training and Annoying Kernel Pedantry
Waterloo, Ontario, CANADA

Home page:                                         http://crashcourse.ca
Fedora Cookbook:    http://crashcourse.ca/wiki/index.php/Fedora_Cookbook
========================================================================

Re: Possibly silly Q?

[Jan Engelhardt]

On Jan 14 2008 15:53, Robert P. J. Day wrote:
>>
>> fdisk or similar should have issued an ioctl to reread the partition
>> table after writing the new one, but you can do it manually with
>> 'blockdev --rereadpt <device>'.
>
>i remember bringing up this very issue quite some time ago and, IIRC,
>the consensus was that *primary* partition changes would be re-read by
>the kernel, but not *logical* partition changes.  or something sort of
>like that.

Nah, it dumps the whole partition structure.
Note in fs/partition/check.c:rescan_partitions:

	for (p = 1; p < disk->minors; p++)
		delete_partition(disk, p);

Re: Possibly silly Q?

[Anton Altaparmakov]

Hi,

On 14 Jan 2008, at 19:34, Gene Heskett wrote:
> Do we have a utility that can force the kernel to re-read, and re- 
> initialize
> itself to a given drives partition tables without having to reboot  
> if one is
> working with a drive that is not part of the required kernel  
> directory tree?


Yes, the blockdev utility.  You want the --rereadpt option, i.e.

blockdev --rereadpt /dev/sdc

It is part of the util-linux package.

Best regards,

	Anton
-- 
Anton Altaparmakov <aia21 at cam.ac.uk> (replace at with @)
Unix Support, Computing Service, University of Cambridge, CB2 3QH, UK
Linux NTFS maintainer, http://www.linux-ntfs.org/

Re: Possibly silly Q?

[Olaf Titz]

> 200Mb had been filled when it had been umounted from /mnt/bootf8
> repartitioned, mke2fs'd, a journal added and a new label written and then
> mounted to /amandatapes.
>
> Something it seems to me, should have forced the re-init, but didn't.  So is

Something I stumbled across just a few days ago: fdisk does cause the
kernel to re-read the new partitions, but nothing causes udev to
create any required new device nodes...

Olaf

Re: Possibly silly Q?

[Jan Engelhardt]

On Jan 14 2008 22:29, Olaf Titz wrote:
>> 200Mb had been filled when it had been umounted from /mnt/bootf8
>> repartitioned, mke2fs'd, a journal added and a new label written and then
>> mounted to /amandatapes.
>>
>> Something it seems to me, should have forced the re-init, but didn't.  So is
>
>Something I stumbled across just a few days ago: fdisk does cause the
>kernel to re-read the new partitions, but nothing causes udev to
>create any required new device nodes...

- kobject_del - will cause udev to delete /dev/sd??
- kernel rescans partitions
- kobject_add - will cause udev to add /dev/sd??

simple :)

Re: Possibly silly Q?

[Lennart Sorensen]

On Mon, Jan 14, 2008 at 02:34:00PM -0500, Gene Heskett wrote:
> Do we have a utility that can force the kernel to re-read, and re-initialize 
> itself to a given drives partition tables without having to reboot if one is 
> working with a drive that is not part of the required kernel directory tree?
> 
> The reason I ask, is that I've just spent about 20 hours fighting with amanda 
> after repartitioning a drive for use by amanda as virtual tapes, 
> making /dev/sdc1 into a primary partition of nearly 400GB.  /dev/sdc1 started 
> out as a /boot2 labeled partition of 200mb, and because the kernels data was 
> stale, amanda upchucked all over itself when the kernel thought the original 
> 200Mb had been filled when it had been umounted from /mnt/bootf8 
> repartitioned, mke2fs'd, a journal added and a new label written and then 
> mounted to /amandatapes.
> 
> Something it seems to me, should have forced the re-init, but didn't.  So is 
> there a tool that can force that?

In my experience fdisk (whichever one you use) will ask the kernel to
reread the partition table on write.

The kernel refuses to reread the partition table on any disk where it
has anything mounted.  So if any part of the disk is in use then it
refuses to change anything about it's view of that disk.

So if you only have partitions mounted from sda and sdb then you can
repartition sdc and it will reload the partition table just fine.  If
any part of sdc is mounted or otherwise in use then you can't.

I have often wondered how hard it would be to change the kernel
behaviour so that it only refuses to reload the partition table if the
partitions that are in use on the disk are being changed, but let it
reload the partition table if only unused partitions are being changed.

For example:
sdc1 200MB mounted
sdc2 500MB not mounted
sdc3 300MB not mounted

Delete sdc2 and sdc3 and create a new sdc2 that is 800MB

Since sdc1 is not changed (no change in size or start and end locations
in partition table), then the kernel should let the partition table be
reloaded since there is no affect on the in use partition.

After all it seems the disk manager in windows is capable of this, so
perhaps linux ought to be as well.  It may not be relevant very often,
but it sure is annoying having to go through multiple reboots when doing
certain disk layout changes just because you are using and not changing
one partition on a disk.

--
Len Sorensen