[git pull] device mapper fix for 4.18-rc6

[git pull] device mapper fix for 4.18-rc6

[Mike Snitzer]

Hi Linus,

The following change since commit 021c91791a5e7e85c567452f1be3e4c2c6cb6063:

  Linux 4.18-rc3 (2018-07-01 16:04:53 -0700)

is available in the Git repository at:

  git://git.kernel.org/pub/scm/linux/kernel/git/device-mapper/linux-dm.git tags/for-4.18/dm-fixes-2

for you to fetch up to d284f8248c72d0cb36a930920e60592eb455cd0d:

  dm writecache: support optional offset for start of device (2018-07-02 16:14:02 -0400)

Please pull, thanks.
Mike

----------------------------------------------------------------
- Fix DM writecache target to allow an optional offset to the start of
  the data and metadata area.  This allows userspace tools (e.g. LVM2)
  to place a header and metadata at the front of the writecache device
  for its use.

----------------------------------------------------------------
Mikulas Patocka (1):
      dm writecache: support optional offset for start of device

 Documentation/device-mapper/writecache.txt |  2 ++
 drivers/md/dm-writecache.c                 | 43 ++++++++++++++++++++----------
 2 files changed, 31 insertions(+), 14 deletions(-)

Re: [git pull] device mapper fix for 4.18-rc6

[Linus Torvalds]

On Fri, Jul 20, 2018 at 12:14 PM Mike Snitzer <snitzer@redhat.com> wrote:
>
>   Fix DM writecache target to allow an optional offset to the start of
>   the data and metadata area.  This allows userspace tools (e.g. LVM2)
>   to place a header and metadata at the front of the writecache device
>   for its use.

"fix"?

Not really.

But I'm a softie, and I took it. Don't do it again.

              Linus

Re: [git pull] device mapper fix for 4.18-rc6

[Mike Snitzer]

On Fri, Jul 20 2018 at  5:25pm -0400,
Linus Torvalds <torvalds@linux-foundation.org> wrote:

> On Fri, Jul 20, 2018 at 12:14 PM Mike Snitzer <snitzer@redhat.com> wrote:
> >
> >   Fix DM writecache target to allow an optional offset to the start of
> >   the data and metadata area.  This allows userspace tools (e.g. LVM2)
> >   to place a header and metadata at the front of the writecache device
> >   for its use.
> 
> "fix"?
> 
> Not really.

I hold it to be a "fix" because it was an embarassing oversight on my
part.  Without this, userspace support cannot be properly added.  LVM2
developers raised their inability to put their metadata anywhere and I
was like "oh shit".

Not much point releasing the DM kernel code if it cannot be properly
tied into a volume manager for persistent use across boots, etc.

> But I'm a softie, and I took it. Don't do it again.

Considering dm-writecache was merged for 4.18 I think it important to
get it right.  So thankfully you've chosen wisely ;)

But I certainly sat on this change for a couple rc releases, should've
gotten it to you a couple weeks ago.

Thanks,
Mike

Re: [git pull] device mapper fix for 4.18-rc6

[Mikulas Patocka]

On Fri, 20 Jul 2018, Mike Snitzer wrote:

> On Fri, Jul 20 2018 at  5:25pm -0400,
> Linus Torvalds <torvalds@linux-foundation.org> wrote:
> 
> > On Fri, Jul 20, 2018 at 12:14 PM Mike Snitzer <snitzer@redhat.com> wrote:
> > >
> > >   Fix DM writecache target to allow an optional offset to the start of
> > >   the data and metadata area.  This allows userspace tools (e.g. LVM2)
> > >   to place a header and metadata at the front of the writecache device
> > >   for its use.
> > 
> > "fix"?
> > 
> > Not really.
> 
> I hold it to be a "fix" because it was an embarassing oversight on my
> part.  Without this, userspace support cannot be properly added.  LVM2
> developers raised their inability to put their metadata anywhere and I
> was like "oh shit".

This is not oversight.

It was originally planned that the dm-writecache target would use lvm 
metadata just like the dm-cache target - if it were implemented this way, 
no "offset" argument would be needed. But David Teigland who was given the 
task to implement dm-writecache support in lvm refused to do it this way 
and he insists that he must put some of his own metadata at the beginning 
of the cache device before the superblock.

So this patch is needed because of him.

Mikulas

Re: [git pull] device mapper fix for 4.18-rc6

[David Teigland]

On Tue, Jul 24, 2018 at 11:18:54AM -0400, Mikulas Patocka wrote:
> > I hold it to be a "fix" because it was an embarassing oversight on my
> > part.  Without this, userspace support cannot be properly added.  LVM2
> > developers raised their inability to put their metadata anywhere and I
> > was like "oh shit".
> 
> This is not oversight.
> 
> It was originally planned that the dm-writecache target would use lvm 
> metadata just like the dm-cache target - if it were implemented this way, 
> no "offset" argument would be needed. But David Teigland who was given the 
> task to implement dm-writecache support in lvm refused to do it this way 
> and he insists that he must put some of his own metadata at the beginning 
> of the cache device before the superblock.
> 
> So this patch is needed because of him.

Mikulas, an offset would have been useful while I was experimenting with
dm-writecache.  I don't actually use or need the offset in lvm.  I don't
think it's a bad idea, but I don't care.

Dave