[Linux-ia64] A start at large file and filesystem support

[Linux-ia64] A start at large file and filesystem support

[peterc]

I've been looking at getting large file system support to work
properly (there's currently a 1TB or 2TB limit on the size of a block device
in 2.5.9, depending on how you use it).

This patch enables large filesystems, and discs bigger than 2TB.
Only the EFI GUID partitioning scheme can be used with enormous discs,
as all the others are constrained by 32-bit number of cylinders.

Next step is to audit the filesystems that claim to be able to support
large filesystems (ext2 may support up to 32TB; reiserfs version 4,
jfs and xfs much more) and make sure that they can.

The patch is a bit big for inclusion in this email.
You can get it from
 http://www.gelato.unsw.edu.au/patches/2.5.9-largefile-patch.txt.bz2
(which also includes the changes David Mosberger put up as the diffs
from 2.5.8)

If you're bitkeeper-enabled, you can pull the patch from
bk://gelato.unsw.edu.au:2022

The change set log is as follows:

ChangeSet@1.602, 2002-04-24 12:35:11+10:00,
peterc@redback.cse.unsw.edu.au
  Merged up to v2.5.9 from local parent.

ChangeSet@1.489.5.9, 2002-04-24 12:22:42+10:00,
peterc@redback.cse.unsw.edu.au
  Many files:
    Use sector_t to count sectors
  blkpg.c:
    Use sector_t to measure sectors; avoid useless checks on a 64-bit
    machine
  
  The aim is to begin a transition to 64-bit sector numbers, even on 
  32-bit architectures, and to allow 64-bit architectures to use huge
  partitions
  and filesystems (where the filesystems support it).

ChangeSet@1.489.5.8, 2002-04-24 12:20:15+10:00,
peterc@redback.cse.unsw.edu.au
  Minor changes to fix compilation warnings on IA64



Peter C

Re: [Linux-ia64] A start at large file and filesystem support

[David Mosberger]

>>>>> On Wed, 24 Apr 2002 15:25:21 +1000, peterc@gelato.unsw.edu.au said:

  Peter> I've been looking at getting large file system support to
  Peter> work properly (there's currently a 1TB or 2TB limit on the
  Peter> size of a block device in 2.5.9, depending on how you use
  Peter> it).

  Peter> This patch enables large filesystems, and discs bigger than
  Peter> 2TB.  Only the EFI GUID partitioning scheme can be used with
  Peter> enormous discs, as all the others are constrained by 32-bit
  Peter> number of cylinders.

  Peter> Next step is to audit the filesystems that claim to be able
  Peter> to support large filesystems (ext2 may support up to 32TB;
  Peter> reiserfs version 4, jfs and xfs much more) and make sure that
  Peter> they can.

Good work!  Since this really doesn't touch any ia64 files: will you
work directly with the respective maintainers to get this into the
official 2.5 tree?

  Peter> The patch is a bit big for inclusion in this email.  You can
  Peter> get it from
  Peter> http://www.gelato.unsw.edu.au/patches/2.5.9-largefile-patch.txt.bz2
  Peter> (which also includes the changes David Mosberger put up as
  Peter> the diffs from 2.5.8)

  Peter> If you're bitkeeper-enabled, you can pull the patch from
  Peter> bk://gelato.unsw.edu.au:2022

Not sure how others feel about this, but I for one still like to get
diff -u-style patches.  It's just so much faster to get a quick idea
of what the patch does.

Thanks,

	--david

Re: [Linux-ia64] A start at large file and filesystem support

[Peter Chubb]

>>>>> "David" = David Mosberger <davidm@napali.hpl.hp.com> writes:

David> Good work!  Since this really doesn't touch any ia64 files:
David> will you work directly with the respective maintainers to get
David> this into the official 2.5 tree?

Sure.  I'll need to split off the changes from the IA64 changes, and
the patch currently will crash on systems that don't use 64-bit
addressing (because of the way the page cache works).  I'd like to get
some people who have enormous filesystems and 64-bit machines to give
the patch a go.

Peter> If you're bitkeeper-enabled, you can pull the patch from
Peter> bk://gelato.unsw.edu.au:2022

David> Not sure how others feel about this, but I for one still like
David> to get diff -u-style patches.  It's just so much faster to get
David> a quick idea of what the patch does.

That's why I provided both -- but 1.8M of diffs is a bit much for an
email.  (I had to change the definition of a couple of interface
functions, which meant lots of files have two line changes)

Peter C

Re: [Linux-ia64] A start at large file and filesystem support

[David Mosberger]

>>>>> On Wed, 24 Apr 2002 20:26:21 +1000, Peter Chubb <peter@chubb.wattle.id.au> said:

>>>>> "David" = David Mosberger <davidm@napali.hpl.hp.com> writes:

  David> Not sure how others feel about this, but I for one still like
  David> to get diff -u-style patches.  It's just so much faster to
  David> get a quick idea of what the patch does.

  Peter> That's why I provided both -- but 1.8M of diffs is a bit much
  Peter> for an email.  (I had to change the definition of a couple of
  Peter> interface functions, which meant lots of files have two line
  Peter> changes)

But you said the patch included all the ia64 changes as well:

  (which also includes the changes David Mosberger put up as the diffs
  from 2.5.8)

This makes the patch fairly useless from the point of view of seeing
what you changed. ;-)

	--david

Re: [Linux-ia64] A start at large file and filesystem support

[Peter Chubb]

David> This makes the patch fairly useless from the point of view of
David> seeing what you changed. ;-)

I'll try to get a just-what-I-changed patch out today.

Peter C