On Tue, 2009-11-03 at 15:16 +0100, Hans de Goede wrote:
> Hi,
>
> On 10/29/2009 05:56 PM, David Lehman wrote:
> > Use Request and Chunk instances to calculate growth for all partitions,
> > then recreates and add the partitions to disk with new geometries.
> > ---
> > storage/partitioning.py | 591 ++++++++++++++++++++++++++++++++--------------
> > 1 files changed, 411 insertions(+), 180 deletions(-)
> >
> > diff --git a/storage/partitioning.py b/storage/partitioning.py
> > index 9471513..1c02b6a 100644
> > --- a/storage/partitioning.py
> > +++ b/storage/partitioning.py
>
> <snip>
>
> > +class Chunk(object):
> > + """ A free region on disk from which partitions will be allocated """
> > + def __init__(self, geometry, requests=None):
> > + """ Create a Chunk instance.
> > +
> > + Arguments:
> > +
> > + geometry -- parted.Geometry instance describing the free space
> > +
> > +
> > + Keyword Arguments:
> > +
> > + requests -- list of Request instances allocated from this chunk
> > +
> > + """
> > + self.geometry = geometry # parted.Geometry
> > + self.pool = self.geometry.length # free sector count
> > + self.sectorSize = self.geometry.device.physicalSectorSize
>
> This is wrong, parted's physicialSectorSize is for informational purposes only,
> the geometry's alignemnts, etc. Are all expressed in logical sector sizes
> (logical == the addressing which goes over the wire to the disk, physical ==
> addressing on the platter, so not really interesting).
>
> Note that parted calls the logical sector size just plain sector size.
>
> Note that your patch has this issue in multiple places, please fix them all.
Ok -- s/physicalSectorSize/sectorSize/g -- will do.
>
> > + self.base = 0 # sum of growable requests' base
> > + # sizes, in sectors
> > + self.requests = [] # list of Request instances
> > + if isinstance(requests, list):
> > + for req in requests:
> > + self.addRequest(req)
> > +
>
> <snip>
>
> > + chunk.growRequests()
> > +
> > + # recalculate partition geometries
> > + disklabel = disk.format
> > +
> > + start = chunk.geometry.start
> > + #if start == 0 or start == getattr(extended_geometry, "start", 0):
> > + # start += 1
> > +
> > + first_logical = False
> > + if extended_geometry and chunk.contains(extended_geometry):
> > + first_logical = True
> > +
> > + new_partitions = []
> > + for p in chunk.requests:
> > + ptype = p.partition.partedPartition.type
> > + log.debug("partition %s (%d): %s" % (p.partition.name,
> > + p.partition.id, ptype))
> > + if ptype == parted.PARTITION_EXTENDED:
> > + continue
> > +
> > + #if ptype == parted.PARTITION_LOGICAL:
> > + # # As you wish, parted.
> > + # start += 1
> > +
> > + # XXX if the start of the extended is aligned then we must
> > + # burn one logical block to align the first logical
> > + # partition
> > + if ptype == parted.PARTITION_LOGICAL and first_logical:
> > + start += _a.grainSize
> > + first_logical = False
> > +
>
> Unfortunately, with extended partitions, we loose a grain size for every logical
> partition. An extended partition looks like this (for example):
>
> 1 sector 1 entry partition table for 1st logical, with a pointer to the next table
> xx sector logical
> 1 sector 1 entry partition table for 2nd logical, with a pointer to the next table
> xx sector logical
> 1 sector 1 entry partition table for 3th logical, with a zero pointer indicating no more logicals follow
> xx sector logical
>
> So we need atleast 1 sector between each logical, now we can make the ending of the logicals
> not aligned, but adding things like filesystem block size into the equasion this probably just means
> we will still be loosing the space.
>
> So I think this first_logical flag is not needed.
Ok -- this is how I had it before I added in the alignment stuff. Will
change it back.
>
> Regards,
>
> Hans
>
> _______________________________________________
> Anaconda-devel-list mailing list
> Anaconda-devel-list@xxxxxxxxxx
> https://www.redhat.com/mailman/listinfo/anaconda-devel-list
_______________________________________________
Anaconda-devel-list mailing list
Anaconda-devel-list@xxxxxxxxxx
https://www.redhat.com/mailman/listinfo/anaconda-devel-list
