[parisc-linux] HFS linux implementation

[parisc-linux] HFS linux implementation

[Nick Cabatoff]

(I sent this to Alex deVries and didn't get a response, so I figured I'd
give this list a try instead.)

I'm interested in the possibility of doing a linux module to mount HFS
filesystems (read-only would satisfy me I think, and is a reasonable
target.)  

I realize that the port is still in its very early stages, so
I intend to develop the module on the i386 platform and then worry
about getting it working on pa-risc.

I'm hoping you can tell me

- if anyone else is already working on this

- if you think the man pages and the white paper might be insufficient
  for the implementation; they look adequate to me, but I've never
  written a file-system driver before

- if you think it's worth the effort, given that I expect HP-UX systems
  are using LVM now and I don't feel up to the task of trying to handle
  that

Re: [parisc-linux] HFS linux implementation

[Grant Grundler]

Nick Cabatoff wrote:
> (I sent this to Alex deVries and didn't get a response, so I figured I'd
> give this list a try instead.)

Poor Alex is pretty busy right now...

> I'm interested in the possibility of doing a linux module to mount HFS
> filesystems (read-only would satisfy me I think, and is a reasonable
> target.)  
> 
> I realize that the port is still in its very early stages, so
> I intend to develop the module on the i386 platform and then worry
> about getting it working on pa-risc.

Should work.

> 
> I'm hoping you can tell me
> 
> - if anyone else is already working on this

I didn't see it listed on:
	http://www.thepuffingroup.com/parisc/status.html
> 
> - if you think the man pages and the white paper might be insufficient
>   for the implementation; they look adequate to me, but I've never
>   written a file-system driver before

no clue. But someone has taken a look at this before and
concluded it was do-able. And you might check if other open
source parisc ports have already done it (eg OpenBSD or mklinux).

As a side note, don't confused HP's HFS with Apple's.
I'm not sure of what to call HP's since Apple HFS support was first.
Perhaps the other ports have set precedence for this.

> - if you think it's worth the effort, given that I expect HP-UX systems
>   are using LVM now and I don't feel up to the task of trying to handle
>   that

Yes. Most older workstations use HFS on wholedisk. HFS can be used
on disks > 4GB without LVM. I avoid LVM whenever I can. It's easier
to physically move disks from one host to another without LVM.
I only use LVM when the boot file system is on a disk > 2GB.


thanks,
grant

Grant Grundler
Unix Development Lab
+1.408.447.7253

Re: [parisc-linux] HFS linux implementation

[Nick Cabatoff]

On Mar 09, Grant Grundler wrote:
> Nick Cabatoff wrote:
> > (I sent this to Alex deVries and didn't get a response, so I figured I'd
> > give this list a try instead.)
> 
> Poor Alex is pretty busy right now...

That's what I gathered.  In any event the list was a better target for
my question I think.

> concluded it was do-able. And you might check if other open
> source parisc ports have already done it (eg OpenBSD or mklinux).

I didn't know there were any *BSD ports; I'm pretty sure mklinux didn't
get that far, based on what I looked at last year.  That's a good
thought though, thanks.  I'll see what I can find.

> As a side note, don't confused HP's HFS with Apple's.
> I'm not sure of what to call HP's since Apple HFS support was first.
> Perhaps the other ports have set precedence for this.

Actually, it looks like just stock UFS with ACL support, as far as I can
see.  Even that's kind of optional; many sites don't use ACLs at all
(hell, dump/restore don't know about them), and it looks like a
read-only implementation that just ignored them would work fine.  I
think that would be almost too easy given the existing linux UFS module
though, so once that much is working I'll probably do a writeable
version that preserves ACLs, even if it doesn't allow you to work with
them.

I think it also may be compatible with OSF/1's UFS (based on an
include-file comment), so this implementation would kill two birds with
one stone.

> > - if you think it's worth the effort, given that I expect HP-UX
> > systems are using LVM now and I don't feel up to the task of trying
> > to handle that
> 
> Yes. Most older workstations use HFS on wholedisk. HFS can be used on
> disks > 4GB without LVM. I avoid LVM whenever I can. It's easier to
> physically move disks from one host to another without LVM.  I only
> use LVM when the boot file system is on a disk > 2GB.

Ah, glad to see I was overgeneralizing from my own experience.  I've had
some bad luck with LVM on root disks myself, but I'm philosophically
opposed to having single-filesystem machines (i.e., I want seperate /var
and /tmp directories), so I stick with it.  <sigh>  I'm kind of
mystified why HP would've waited for LVM to allow multiple filesystems
on a single disk; what's so bad about partition tables?

Re: [parisc-linux] HFS linux implementation

[Grant Grundler]

getting a bit off track here...

Nick Cabatoff wrote:
> <sigh>  I'm kind of
> mystified why HP would've waited for LVM to allow multiple filesystems
> on a single disk; what's so bad about partition tables?

You have to know about HP's "hard" partition scheme.
It was a pain to support. Every new disk required a new entry
in the kernel's partition table. yes, I said Kernel. It's not
stored on the disk. And that was delivered as a patch. 'nuf said.
That's why LVM is better than "partion tables".

Since LVM was "on the roadmap" anyway, HP decided to depend LVM
to provide "soft" portitioning (ie partition tables stored on the
media) instead of something like the PC's use.

grant

Grant Grundler
Unix Development Lab
+1.408.447.7253

Re: [parisc-linux] HFS linux implementation

[Dominik Kubla]

On Fri, Mar 10, 2000 at 01:09:49AM -0500, Nick Cabatoff wrote:

> Actually, it looks like just stock UFS with ACL support, as far as I can
> see.  Even that's kind of optional; many sites don't use ACLs at all
> (hell, dump/restore don't know about them), and it looks like a
> read-only implementation that just ignored them would work fine.  I
> think that would be almost too easy given the existing linux UFS module
> though, so once that much is working I'll probably do a writeable
> version that preserves ACLs, even if it doesn't allow you to work with
> them.

Well, just talk to the linux-acl folks... they are just pushing the
(former ext2) ACL stuff up into the VFS, in order to have a generic
API to access the filesystem specific ACL implementations.  Unfortunately
the ACL development still lingers in 2.2 territory, but that's no excuse
not to think ahead and at least add the hooks for it when one touches
a file system anyway.

Web pages are at:    http://acl.bestbits.at/
Mailing list is at:  http://acl.bestbits.at/acl-devel/

Dominik
-- 
  Networking Group,  Hospital of Johannes Gutenberg-University                  
  Obere Zahlbacher Straße 69, 55101 Mainz, Germany                              
  Tel: +49 (0)6131 17-2482   FAX: +49 (0)6131 17-5521                           

Re: [parisc-linux] HFS linux implementation

[Martin K. Petersen]

>>>>> "Nick" == Nick Cabatoff <ncc@cs.mcgill.ca> writes:

Hi Nick,

Nick> I'm interested in the possibility of doing a linux module to
Nick> mount HFS filesystems (read-only would satisfy me I think, and
Nick> is a reasonable target.)

I implemented read only hfs support in the early days of the PA-RISC
port.  It appears my ufs changes didn't get merged in when we moved
from 2.2 to 2.3, though.  Will fix...

One of Paul Bame's colleagues looked into implementing write support
at some point, but I don't know status on that project.  Paul?

-- 
Martin K. Petersen, Principal Linux Consultant, Linuxcare, Inc.
mkp@mkp.net, http://www.linuxcare.com/
Linuxcare. Support for the revolution.

Re: [parisc-linux] HFS linux implementation

[Nick Cabatoff]

On Mar 10, Martin K. Petersen wrote:
> Nick> I'm interested in the possibility of doing a linux module to
> Nick> mount HFS filesystems (read-only would satisfy me I think, and
> Nick> is a reasonable target.)
> 
> I implemented read only hfs support in the early days of the PA-RISC
> port.  It appears my ufs changes didn't get merged in when we moved
> from 2.2 to 2.3, though.  Will fix...

Could I get a copy of the 2.2 version to satisfy my curiosity?

> One of Paul Bame's colleagues looked into implementing write support
> at some point, but I don't know status on that project.  Paul?

Since (expirimental) writeable UFS exists in the current kernel, and I
assume your implementation is a patch to the existing UFS code, I
don't imagine it would be difficult.  I'll try and do it if the person
you mention doesn't have time.

Is there any point in doing ACLs if you haven't already?

Re: [parisc-linux] HFS linux implementation

[willy]

On Fri, Mar 10, 2000 at 10:59:21AM -0500, Nick Cabatoff wrote:
> Could I get a copy of the 2.2 version to satisfy my curiosity?

sure.  cvs co linux gets you the linux 2.2.12 tree we were working on.

> Is there any point in doing ACLs if you haven't already?

linux doesn't currently support acls.  there's no interface for setting
or accessing acls.

Re: [parisc-linux] HFS linux implementation

[Nick Cabatoff]

On Mar 10, willy@thepuffingroup.com wrote:
> On Fri, Mar 10, 2000 at 10:59:21AM -0500, Nick Cabatoff wrote:
> > Could I get a copy of the 2.2 version to satisfy my curiosity?
> 
> sure.  cvs co linux gets you the linux 2.2.12 tree we were working on.

Okay, I'll take a look, thanks.

> > Is there any point in doing ACLs if you haven't already?
> 
> linux doesn't currently support acls.  there's no interface for setting
> or accessing acls.

There is one that hasn't made it into the stock kernel yet at
http://acl.bestbits.at.