Combined storage tree

Combined storage tree

[James Bottomley]

One of the requests from LSF10 in August was the production of a
combined storage tree.  This is now ready at

git://git.kernel.org/pub/scm/linux/kernel/git/jejb/storage-tree

It's actually a nightly built merge tree consisting of

scsi-misc; scsi-rc-fixes
libata#upstream-fixes, libata#upstream
block#for-linus, block#for-next
and the dm quilt (which is empty at the moment).

I haven't yet added vfs or any of the fs trees, but if necessary, I can.

Note, because it's built nightly, like linux-next, it's hard (but not
impossible) to use it as a basis for git trees (it is much easier to use
it as a basis for quilts).

If there are any cross-tree patch sets, I can add them in for this too.

The trees are labelled <linux-tree>-stor<n>

You can find the combined diffs to the base linus tree here:

http://www.kernel.org/pub/linux/kernel/people/jejb/storage-tree

(The build script and driving file are in the parent directory).

Any build cockups or missing trees, please let me know.

James

Re: Combined storage tree

[Dave Chinner]

On Fri, Sep 10, 2010 at 01:27:27PM -0500, James Bottomley wrote:
> One of the requests from LSF10 in August was the production of a
> combined storage tree.  This is now ready at
> 
> git://git.kernel.org/pub/scm/linux/kernel/git/jejb/storage-tree
> 
> It's actually a nightly built merge tree consisting of
> 
> scsi-misc; scsi-rc-fixes
> libata#upstream-fixes, libata#upstream
> block#for-linus, block#for-next
> and the dm quilt (which is empty at the moment).
> 
> I haven't yet added vfs or any of the fs trees, but if necessary, I can.
> 
> Note, because it's built nightly, like linux-next, it's hard (but not
> impossible) to use it as a basis for git trees (it is much easier to use
> it as a basis for quilts).

Hmmm. I was kind of hoping for an upstream maintainer tree, kind of
like the netdev tree.  I really don't see a tree like this getting
wide use - if I enjoyed the pain of rebasing against throw-away
merge trees every day, then I'd already be using linux-next....

Cheers,

Dave.
-- 
Dave Chinner
david@fromorbit.com

Re: Combined storage tree

[James Bottomley]

On Sat, 2010-09-11 at 18:20 +1000, Dave Chinner wrote:
> On Fri, Sep 10, 2010 at 01:27:27PM -0500, James Bottomley wrote:
> > One of the requests from LSF10 in August was the production of a
> > combined storage tree.  This is now ready at
> > 
> > git://git.kernel.org/pub/scm/linux/kernel/git/jejb/storage-tree
> > 
> > It's actually a nightly built merge tree consisting of
> > 
> > scsi-misc; scsi-rc-fixes
> > libata#upstream-fixes, libata#upstream
> > block#for-linus, block#for-next
> > and the dm quilt (which is empty at the moment).
> > 
> > I haven't yet added vfs or any of the fs trees, but if necessary, I can.
> > 
> > Note, because it's built nightly, like linux-next, it's hard (but not
> > impossible) to use it as a basis for git trees (it is much easier to use
> > it as a basis for quilts).
> 
> Hmmm. I was kind of hoping for an upstream maintainer tree, kind of
> like the netdev tree.  I really don't see a tree like this getting
> wide use - if I enjoyed the pain of rebasing against throw-away
> merge trees every day, then I'd already be using linux-next....

Well, to be honest, that's what people wanted when the issue was raised
at LSF10.  However, unlike net, storage has never had a single
maintainer, so it's a bit more political than just doing that by fiat,
plus not all of the current maintainers with storage trees were there.
So we agreed (reluctantly) to start with an automatic tree and see how
much of the current problem that solved.  If the automatic tree turns
out not to be very useful, we can revisit the idea of a storage
maintainer.

The reason for being storage only rather than saying just use linux-next
(which was mentioned) is that next is a much bigger tree, so it's harder
to follow.  The diffs to mainline in the current storage tree are still
under a megabyte.

James

Re: Combined storage tree

[Dave Chinner]

On Sat, Sep 11, 2010 at 08:55:16AM -0500, James Bottomley wrote:
> On Sat, 2010-09-11 at 18:20 +1000, Dave Chinner wrote:
> > On Fri, Sep 10, 2010 at 01:27:27PM -0500, James Bottomley wrote:
> > > One of the requests from LSF10 in August was the production of a
> > > combined storage tree.  This is now ready at
> > > 
> > > git://git.kernel.org/pub/scm/linux/kernel/git/jejb/storage-tree
> > > 
> > > It's actually a nightly built merge tree consisting of
> > > 
> > > scsi-misc; scsi-rc-fixes
> > > libata#upstream-fixes, libata#upstream
> > > block#for-linus, block#for-next
> > > and the dm quilt (which is empty at the moment).
> > > 
> > > I haven't yet added vfs or any of the fs trees, but if necessary, I can.
> > > 
> > > Note, because it's built nightly, like linux-next, it's hard (but not
> > > impossible) to use it as a basis for git trees (it is much easier to use
> > > it as a basis for quilts).
> > 
> > Hmmm. I was kind of hoping for an upstream maintainer tree, kind of
> > like the netdev tree.  I really don't see a tree like this getting
> > wide use - if I enjoyed the pain of rebasing against throw-away
> > merge trees every day, then I'd already be using linux-next....
> 
> Well, to be honest, that's what people wanted when the issue was raised
> at LSF10.  However, unlike net, storage has never had a single
> maintainer, so it's a bit more political than just doing that by fiat,

Bah. Technical arguments win here, not politics. Besides, what
possible political concern can anyone have about using a different
upstream tree for development? A storage maintainer tree would not
replace anyone's little fiefdom; what we need is an integration point
long before stuff gets to Linus....

> plus not all of the current maintainers with storage trees were there.

If that's the barrier to discussion, then where else but a dedicated
storage workshop are you going to get a more representative sample
of storage developers in the same room?

Cheers,

Dave.
-- 
Dave Chinner
david@fromorbit.com

Re: Combined storage tree

[James Bottomley]

On Mon, 2010-09-13 at 12:58 +1000, Dave Chinner wrote: 
> On Sat, Sep 11, 2010 at 08:55:16AM -0500, James Bottomley wrote:
> > On Sat, 2010-09-11 at 18:20 +1000, Dave Chinner wrote:
> > > On Fri, Sep 10, 2010 at 01:27:27PM -0500, James Bottomley wrote:
> > > > One of the requests from LSF10 in August was the production of a
> > > > combined storage tree.  This is now ready at
> > > > 
> > > > git://git.kernel.org/pub/scm/linux/kernel/git/jejb/storage-tree
> > > > 
> > > > It's actually a nightly built merge tree consisting of
> > > > 
> > > > scsi-misc; scsi-rc-fixes
> > > > libata#upstream-fixes, libata#upstream
> > > > block#for-linus, block#for-next
> > > > and the dm quilt (which is empty at the moment).
> > > > 
> > > > I haven't yet added vfs or any of the fs trees, but if necessary, I can.
> > > > 
> > > > Note, because it's built nightly, like linux-next, it's hard (but not
> > > > impossible) to use it as a basis for git trees (it is much easier to use
> > > > it as a basis for quilts).
> > > 
> > > Hmmm. I was kind of hoping for an upstream maintainer tree, kind of
> > > like the netdev tree.  I really don't see a tree like this getting
> > > wide use - if I enjoyed the pain of rebasing against throw-away
> > > merge trees every day, then I'd already be using linux-next....
> > 
> > Well, to be honest, that's what people wanted when the issue was raised
> > at LSF10.  However, unlike net, storage has never had a single
> > maintainer, so it's a bit more political than just doing that by fiat,
> 
> Bah. Technical arguments win here, not politics.

Heh, who runs a tree is both politics and technical.

> Besides, what
> possible political concern can anyone have about using a different
> upstream tree for development? A storage maintainer tree would not
> replace anyone's little fiefdom; what we need is an integration point
> long before stuff gets to Linus....

For a storage git tree to be usable as a stable base, it has to be the
upstream tree.  That would mean that, like net, the way upstream would
be via this tree.  That's quite a shift in the way we currently work.

> > plus not all of the current maintainers with storage trees were there.
> 
> If that's the barrier to discussion, then where else but a dedicated
> storage workshop are you going to get a more representative sample
> of storage developers in the same room?

This sort of thing doesn't get decided by fiat.  If you can't get all of
the relevant parties to agree, you have to demonstrate the need.  So
doing a rollup tree to test how much of the problem is solvable that way
seems like a reasonable first step.

James