[parted-devel] Possible GPT r/w error

[David Cantrell]

Andrew Gaffney wrote:
> leslie.polzer at gmx.net wrote:
> >On Tue, Apr 25, 2006 at 04:59:09PM -0400, David Cantrell wrote:
> >
> >>Steps given to me to reproduce the problem:
> >>
> >>[...]
> >We need more information from them.  Is this the minimal case?  What
> >happens when mounting and unmounting occur at different times?  Does the
> >bug show up only when creating the second partition?
> >
> >  A dump of the first three sectors at different stages of the process
> >would also be helpful.
> >
> >  I'm eager to get this information because I would like to fix this
> >serious bug.
> 
> It isn't exactly smart to be partitioning a disk with mounted/active 
> partitions.

In general, yes.  But enterprise (I really don't have a better word to
use) customers sort of expect this.  Think about older Unix systems
where accepted partitioning standards left part of the physical disk
unallocated.  When a particular area of the disk was running out of
space, the admin could claim part of the remaining disk space using
format(1M) and then mount it.  All while the system is running.

There are definite instances where you do not want to modify a partition
table while the system is running with mounted and active partitions,
but this shouldn't be one of those cases.

-- 
David Cantrell
Red Hat / Westford, MA