[uml-devel] Separate COW patch.

[uml-devel] Separate COW patch.

[Montgomery, Kendal L]

[-- Attachment #1: Type: text/plain, Size: 1134 bytes --]

Hello,

I was wondering if there was anyone who is trying to use the COW filesystem
in a normal linux kernel?  Is this possible, or is it tied exclusively to
the rest of the UML patch?  How difficult would it be to make a COW only
patch?  I have a certain project that I'm working on that the COW filesystem
(or really, any other "union" filesystem, but I can't find any that are
actively developed or fully cooked) would be perfect for.  I have a
read-only filesystem that, occasionally needs to temoroarily have some
changed made to it, but then the need to be undone.  I've looked into
various ways of doing this from "snapshotting," to user-controlled
transactions on a journalling filesystem (no such thing seems to exist for
ext3 or reiserfs) where I could explicitly begin a transaction, then roll
back the transaction when I'm done (kinda like a database txn), etc.  The
COW filesystem seems like it would be a perfect fit, but before I go trying
to merge things into a non-UML kernel on my own I just want to see if anyone
else has already done this or not.  Is anyone else even interested in such a
thing?

Thanks.

Kendal.

[-- Attachment #2: Type: text/html, Size: 1648 bytes --]

Re: [uml-devel] Separate COW patch.

[Goetz Bock]

On Fri, Sep 26 '03 at 18:45, Montgomery, Kendal L wrote:
> I was wondering if there was anyone who is trying to use the COW
> filesystem in a normal linux kernel? [ ... ]
The COW is not a filesystem, it is a block level thingy.

If you only have a small number of changes, the LVM / DM snapshot
feature might be usefill. OTOH it is designed to be merged afterwards,
not discarded.
-- 
/"\ Goetz Bock at blacknet dot de  --  secure mobile Linux everNETting
\ /                     (c) 2003 as GNU FDL 1.1
 X   [ 1. Use descriptive subjects - 2. Edit a reply for brevity -  ]
/ \  [ 3. Reply to the list - 4. Read the archive *before* you post ]


-------------------------------------------------------
This sf.net email is sponsored by:ThinkGeek
Welcome to geek heaven.
http://thinkgeek.com/sf
_______________________________________________
User-mode-linux-devel mailing list
User-mode-linux-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/user-mode-linux-devel

Re: [uml-devel] Separate COW patch.

[Jeff Dike]

kendal.montgomery@qwest.com said:
> I was wondering if there was anyone who is trying to use the COW
> filesystem in a normal linux kernel?  Is this possible, or is it tied
> exclusively to the rest of the UML patch?

I started making a standalone COW driver that would be usable in any kernel,
UML or otherwise.  The results are visible in the patch in 
arch/um/drivers/cow*.c.

I got sidetracked, but this is something that I'm going to finish up at
some point.

				Jeff



-------------------------------------------------------
This sf.net email is sponsored by:ThinkGeek
Welcome to geek heaven.
http://thinkgeek.com/sf
_______________________________________________
User-mode-linux-devel mailing list
User-mode-linux-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/user-mode-linux-devel

Re: [uml-devel] Separate COW patch.

[Michael Richardson]

-----BEGIN PGP SIGNED MESSAGE-----


If you are going to spend time on this, go look at the "union" file system
type available in BSDs and the "Translucent" in Solaris.

I think that some people have tried to do this before... usually, it is
a variation of the "null" mount.

It permits one to layer one filesystem on top of another file system, and
have all modifications go to the upper one. This is done on a *per-file*
basis, not on a block file system. 

COW is about blocks, not files. This may be more appropriate for your use,
or it might not be.

] Train travel features AC outlets with no take-off restrictions|  firewalls  [
]   Michael Richardson, Sandelman Software Works, Ottawa, ON    |net architect[
] mcr@sandelman.ottawa.on.ca http://www.sandelman.ottawa.on.ca/ |device driver[
] panic("Just another Debian/notebook using, kernel hacking, security guy");  [







-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.2.2 (GNU/Linux)
Comment: Finger me for keys - custom hacks make this fully PGP2 compat

iQCVAwUBP3YVoYqHRg3pndX9AQEWjQP9ESUVMOeyR/KSlWf5h2Wj5e8yJal5lGKJ
RoW26/KmvTZXmNaHr2lpf9aM9rJWgvLPZp6c20DKgKVSkWEN+at8egZLi5N5ksUc
Nh5dzG++IidzlqxxAMxpL0TxSseuzQjEzUOGCmc+/qpjAIoS6XDs8BoraEfkYkcg
/F/EJAQvFpk=
=tPN1
-----END PGP SIGNATURE-----


-------------------------------------------------------
This sf.net email is sponsored by:ThinkGeek
Welcome to geek heaven.
http://thinkgeek.com/sf
_______________________________________________
User-mode-linux-devel mailing list
User-mode-linux-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/user-mode-linux-devel

Re: [uml-devel] Separate COW patch.

[Paul Sladen]

On Fri, 26 Sep 2003, Montgomery, Kendal L wrote:
> I was wondering if there was anyone who is trying to use the COW filesystem
> in a normal linux kernel?

The UML COW is block-based and isn't really what you want.  Try:

  http://vserver.13thfloor.at/TBVFS/

	-Paul
-- 
War is inconsistent with Truth.  Nottingham, GB





-------------------------------------------------------
This sf.net email is sponsored by:ThinkGeek
Welcome to geek heaven.
http://thinkgeek.com/sf
_______________________________________________
User-mode-linux-devel mailing list
User-mode-linux-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/user-mode-linux-devel

Re: [uml-devel] Separate COW patch.

[Matt Zimmerman]

On Fri, Sep 26, 2003 at 06:45:26PM -0400, Montgomery, Kendal L wrote:

> I was wondering if there was anyone who is trying to use the COW
> filesystem in a normal linux kernel?  Is this possible, or is it tied
> exclusively to the rest of the UML patch?  How difficult would it be to
> make a COW only patch?  I have a certain project that I'm working on that
> the COW filesystem (or really, any other "union" filesystem, but I can't
> find any that are actively developed or fully cooked) would be perfect
> for.  I have a read-only filesystem that, occasionally needs to
> temoroarily have some changed made to it, but then the need to be undone.
> I've looked into various ways of doing this from "snapshotting," to
> user-controlled transactions on a journalling filesystem (no such thing
> seems to exist for ext3 or reiserfs) where I could explicitly begin a
> transaction, then roll back the transaction when I'm done (kinda like a
> database txn), etc.  The COW filesystem seems like it would be a perfect
> fit, but before I go trying to merge things into a non-UML kernel on my
> own I just want to see if anyone else has already done this or not.  Is
> anyone else even interested in such a

You want the snapshot feature in device-mapper, with lvm2 or evms.  It does
exactly what you've described, with any filesystem.

-- 
 - mdz


-------------------------------------------------------
This sf.net email is sponsored by:ThinkGeek
Welcome to geek heaven.
http://thinkgeek.com/sf
_______________________________________________
User-mode-linux-devel mailing list
User-mode-linux-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/user-mode-linux-devel

RE: [uml-devel] Separate COW patch.

[Montgomery, Kendal L]

Thanks everyone for the feedback.  I'm looking further into using lvm2
or evms to solve this problem.  However, I'm quite certain that a
standalone COW driver would easily solve this problem as well, so if the
patch set mentioned before by Jeff Dike is ever completed, I'd be
interested in trying it out.

Jeff,

Where can I get what you've worked on so far?  I'd be willing to give it
a try and possibly help to sort out the remaining pieces that need to
make it work.

Kendal.

-----Original Message-----
From: Matt Zimmerman [mailto:mdz@debian.org] 
Sent: Saturday, September 27, 2003 9:45 PM
To: 'user-mode-linux-devel@lists.sourceforge.net'
Subject: Re: [uml-devel] Separate COW patch.


On Fri, Sep 26, 2003 at 06:45:26PM -0400, Montgomery, Kendal L wrote:

> I was wondering if there was anyone who is trying to use the COW 
> filesystem in a normal linux kernel?  Is this possible, or is it tied 
> exclusively to the rest of the UML patch?  How difficult would it be 
> to make a COW only patch?  I have a certain project that I'm working 
> on that the COW filesystem (or really, any other "union" filesystem, 
> but I can't find any that are actively developed or fully cooked) 
> would be perfect for.  I have a read-only filesystem that, 
> occasionally needs to temoroarily have some changed made to it, but 
> then the need to be undone. I've looked into various ways of doing 
> this from "snapshotting," to user-controlled transactions on a 
> journalling filesystem (no such thing seems to exist for ext3 or 
> reiserfs) where I could explicitly begin a transaction, then roll back

> the transaction when I'm done (kinda like a database txn), etc.  The 
> COW filesystem seems like it would be a perfect fit, but before I go 
> trying to merge things into a non-UML kernel on my own I just want to 
> see if anyone else has already done this or not.  Is anyone else even 
> interested in such a

You want the snapshot feature in device-mapper, with lvm2 or evms.  It
does exactly what you've described, with any filesystem.

-- 
 - mdz


-------------------------------------------------------
This sf.net email is sponsored by:ThinkGeek
Welcome to geek heaven.
http://thinkgeek.com/sf _______________________________________________
User-mode-linux-devel mailing list
User-mode-linux-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/user-mode-linux-devel


-------------------------------------------------------
This sf.net email is sponsored by:ThinkGeek
Welcome to geek heaven.
http://thinkgeek.com/sf
_______________________________________________
User-mode-linux-devel mailing list
User-mode-linux-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/user-mode-linux-devel

Re: [uml-devel] Separate COW patch.

[Jeff Dike]

kendal.montgomery@qwest.com said:
> Where can I get what you've worked on so far?  I'd be willing to give
> it a try and possibly help to sort out the remaining pieces that need
> to make it work. 

It's in arch/um/drivers/cow_kern.c.  The bit that got me hung up was ensuring
that COW bitmap updates happen before the actual I/O.  This is necessary
in case the UML crashes during a write.  You want the bitmap update to
happen only after the data has been written.

				Jeff



-------------------------------------------------------
This sf.net email is sponsored by:ThinkGeek
Welcome to geek heaven.
http://thinkgeek.com/sf
_______________________________________________
User-mode-linux-devel mailing list
User-mode-linux-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/user-mode-linux-devel