[GIT Review] xfs: 2.6.36 fixes

[GIT Review] xfs: 2.6.36 fixes

[Dave Chinner]

Folks,

Can you please cast an eye over the the tree below to check that all
the bug fixes that need to go into 2.6.36 are there? If so, I'll
send a pull request to Linus tomorrow.

Note that these are just the outstanding bug fixes that have been
reviewed (as Linus has again decreed for pull requests after -rc1),
so it not the complete set of reviewd patches that exist.

Cheers,

Dave.


The following changes since commit 9ee47476d6734c9deb9ae9ab05d963302f6b6150:

  Merge branch 'radix-tree' of git://git.kernel.org/pub/scm/linux/kernel/git/dgc/xfsdev (2010-08-22 19:55:14 -0700)

are available in the git repository at:

  git://git.kernel.org/pub/scm/linux/kernel/git/dgc/xfsdev.git 2.6.36-fixes

Christoph Hellwig (1):
      xfs: do not discard page cache data on EAGAIN

Dave Chinner (8):
      xfs: unlock items before allowing the CIL to commit
      xfs: ensure we mark all inodes in a freed cluster XFS_ISTALE
      xfs: fix untrusted inode number lookup
      writeback: write_cache_pages doesn't terminate at nr_to_write <= 0
      xfs: handle negative wbc->nr_to_write during sync writeback
      xfs: dummy transactions should not dirty VFS state
      xfs: Reduce log force overhead for delayed logging
      xfs: don't do memory allocation under the CIL context lock

Stuart Brodsky (1):
      xfs: ensure f_ffree returned by statfs() is non-negative

 fs/xfs/linux-2.6/xfs_aops.c  |   13 ++-
 fs/xfs/linux-2.6/xfs_super.c |    9 +-
 fs/xfs/linux-2.6/xfs_sync.c  |   42 +------
 fs/xfs/xfs_fsops.c           |   31 +++--
 fs/xfs/xfs_fsops.h           |    2 +-
 fs/xfs/xfs_ialloc.c          |   16 ++-
 fs/xfs/xfs_inode.c           |   49 ++++----
 fs/xfs/xfs_log.c             |    7 +-
 fs/xfs/xfs_log_cil.c         |  263 +++++++++++++++++++++++++-----------------
 fs/xfs/xfs_log_priv.h        |   13 ++-
 fs/xfs/xfs_trans.c           |    5 +-
 fs/xfs/xfs_trans_priv.h      |    3 +-
 mm/page-writeback.c          |   26 ++---
 13 files changed, 261 insertions(+), 218 deletions(-)

-- 
Dave Chinner
david@fromorbit.com

_______________________________________________
xfs mailing list
xfs@oss.sgi.com
http://oss.sgi.com/mailman/listinfo/xfs

Re: [GIT Review] xfs: 2.6.36 fixes

[Christoph Hellwig]

On Tue, Aug 24, 2010 at 12:44:45PM +1000, Dave Chinner wrote:
> Folks,
> 
> Can you please cast an eye over the the tree below to check that all
> the bug fixes that need to go into 2.6.36 are there? If so, I'll
> send a pull request to Linus tomorrow.

The tree looks good to me.

_______________________________________________
xfs mailing list
xfs@oss.sgi.com
http://oss.sgi.com/mailman/listinfo/xfs

Re: [GIT Review] xfs: 2.6.36 fixes

[Alex Elder]

On Tue, 2010-08-24 at 12:44 +1000, Dave Chinner wrote:
> Folks,
> 
> Can you please cast an eye over the the tree below to check that all
> the bug fixes that need to go into 2.6.36 are there? If so, I'll
> send a pull request to Linus tomorrow.
> 
> Note that these are just the outstanding bug fixes that have been
> reviewed (as Linus has again decreed for pull requests after -rc1),
> so it not the complete set of reviewd patches that exist.
> 
> Cheers,
> 
> Dave.

All of these commits look good to me.  I will have these
and all the rest published on the oss.sgi.com tree later
this week.  I've been on vacation and have gotten behind.

Reviewed-by: Alex Elder <aelder@sgi.com>

> 
> The following changes since commit 9ee47476d6734c9deb9ae9ab05d963302f6b6150:
> 
>   Merge branch 'radix-tree' of git://git.kernel.org/pub/scm/linux/kernel/git/dgc/xfsdev (2010-08-22 19:55:14 -0700)
> 
> are available in the git repository at:
> 
>   git://git.kernel.org/pub/scm/linux/kernel/git/dgc/xfsdev.git 2.6.36-fixes
. . .

_______________________________________________
xfs mailing list
xfs@oss.sgi.com
http://oss.sgi.com/mailman/listinfo/xfs

Re: [GIT Review] xfs: 2.6.36 fixes

[Dave Chinner]

On Tue, Aug 24, 2010 at 05:54:58PM -0500, Alex Elder wrote:
> On Tue, 2010-08-24 at 12:44 +1000, Dave Chinner wrote:
> > Folks,
> > 
> > Can you please cast an eye over the the tree below to check that all
> > the bug fixes that need to go into 2.6.36 are there? If so, I'll
> > send a pull request to Linus tomorrow.
> > 
> > Note that these are just the outstanding bug fixes that have been
> > reviewed (as Linus has again decreed for pull requests after -rc1),
> > so it not the complete set of reviewd patches that exist.
> > 
> > Cheers,
> > 
> > Dave.
> 
> All of these commits look good to me.  I will have these
> and all the rest published on the oss.sgi.com tree later
> this week.  I've been on vacation and have gotten behind.

I'm going to send the pull to Linus, anyway, Alex.

The problem is deeper than "been on vacation" - regardless of
whether you are on vacation we can't rely on you to do anything
immediately. It might be code review, triage community reported
bugs, pulling patches into the OSS tree or sending stuff to Linus,
but it's always a week or two later than it needs to be. This is
not a new problem, either.

Just on the git aspect of this problem, I haven't been using the git
tree on oss.sgi.com now for a couple of months - instead I'm working
from a mainline tree and using topic branches and local merges to
manange separate patch sets. Having to work with a slow-to-update XFS
tree is actually quite painful, and most of that pain goes away it I
just drop the OSS tree out of the loop completely.

For example, the last pull request I sent to Linus was for the
radix-tree branch containing writeback regression fixes. Linus
merged that branch into mainline within ten minutes of me sending
the pull request.

If I contrast that to getting the same patches to Linus via the oss
tree - I'd still be waiting for you to get them into the OSS tree
and all I know is that it would be "later this week". It's just
easier to send stuff that is ready straight to Linus.  Given that
I'm already doing all the git tree work to integrate, tag and test
all the XFS patches coming in on the mailing list, adding an extra
tree hop with an unknown latency to get the commits to Linus is
distinctly sub-optimal.

So, give me some good reasons why I should continue to send XFS
kernel code through the SGI controlled tree on oss.sgi.com rather
than directly to Linus.

Cheers,

Dave.

-- 
Dave Chinner
david@fromorbit.com

_______________________________________________
xfs mailing list
xfs@oss.sgi.com
http://oss.sgi.com/mailman/listinfo/xfs

Re: [GIT Review] xfs: 2.6.36 fixes

[Dave Chinner]

On Wed, Aug 25, 2010 at 11:10:28AM +1000, Dave Chinner wrote:
> On Tue, Aug 24, 2010 at 05:54:58PM -0500, Alex Elder wrote:
> > On Tue, 2010-08-24 at 12:44 +1000, Dave Chinner wrote:
> > > Folks,
> > > 
> > > Can you please cast an eye over the the tree below to check that all
> > > the bug fixes that need to go into 2.6.36 are there? If so, I'll
> > > send a pull request to Linus tomorrow.
> > > 
> > > Note that these are just the outstanding bug fixes that have been
> > > reviewed (as Linus has again decreed for pull requests after -rc1),
> > > so it not the complete set of reviewd patches that exist.
> > > 
> > > Cheers,
> > > 
> > > Dave.
> > 
> > All of these commits look good to me.  I will have these
> > and all the rest published on the oss.sgi.com tree later
> > this week.  I've been on vacation and have gotten behind.
> 
> I'm going to send the pull to Linus, anyway, Alex.
> 
> The problem is deeper than "been on vacation" - regardless of
> whether you are on vacation we can't rely on you to do anything
> immediately. It might be code review, triage community reported
> bugs, pulling patches into the OSS tree or sending stuff to Linus,
> but it's always a week or two later than it needs to be. This is
> not a new problem, either.

So this doesn't get taken the wrong way, this is not a comment on
the quality of the work you do, Alex. My concern is the limited
amount of time you have available to do the required work and
avoiding the bottleneck it causes.

Cheers,

Dave.
-- 
Dave Chinner
david@fromorbit.com

_______________________________________________
xfs mailing list
xfs@oss.sgi.com
http://oss.sgi.com/mailman/listinfo/xfs

Re: [GIT Review] xfs: 2.6.36 fixes

[Alex Elder]

On Wed, 2010-08-25 at 14:50 +1000, Dave Chinner wrote:
> On Wed, Aug 25, 2010 at 11:10:28AM +1000, Dave Chinner wrote:
> > On Tue, Aug 24, 2010 at 05:54:58PM -0500, Alex Elder wrote:
> > > On Tue, 2010-08-24 at 12:44 +1000, Dave Chinner wrote:
> > > > Folks,
> > > >
> > > > Can you please cast an eye over the the tree below to check that
all
> > > > the bug fixes that need to go into 2.6.36 are there? If so, I'll
> > > > send a pull request to Linus tomorrow.
> > > >
> > > > Note that these are just the outstanding bug fixes that have
been
> > > > reviewed (as Linus has again decreed for pull requests after
-rc1),
> > > > so it not the complete set of reviewd patches that exist.
> > > >
> > > > Cheers,
> > > >
> > > > Dave.
> > >
> > > All of these commits look good to me.  I will have these
> > > and all the rest published on the oss.sgi.com tree later
> > > this week.  I've been on vacation and have gotten behind.
> >
> > I'm going to send the pull to Linus, anyway, Alex.
> >
> > The problem is deeper than "been on vacation" - regardless of
> > whether you are on vacation we can't rely on you to do anything
> > immediately. It might be code review, triage community reported
> > bugs, pulling patches into the OSS tree or sending stuff to Linus,
> > but it's always a week or two later than it needs to be. This is
> > not a new problem, either.
>
> So this doesn't get taken the wrong way, this is not a comment on
> the quality of the work you do, Alex. My concern is the limited
> amount of time you have available to do the required work and
> avoiding the bottleneck it causes.

This has been an ongoing issue, and I concur that it's been a matter
of giving XFS maintenance the time and attention it requires.

In July I raised this issue internally and got agreement to give
this more priority versus other competing demands.  That helped
some, but there still were some other things that needed my
dedicated focus, and that disrupted the continuity of my attention
to XFS maintenance.

My vacation last week, directly on the heels of my trip to Boston
the week before (during which I had limited network access) resulted
in an extended delay whose cause was actually pretty unusual (and
unfortunately timed). 

I have now been directed to spend 100% of my time on XFS maintenance
for the community.  This will not result in 10 minute turnaround but
it will allow me to offer everything I've got to the cause, without
interruption.

SGI continues to be committed to supporting XFS (for more than just
its own customers), and in the coming month or so is going to be
evaluating what more can be done to serve and benefit the community.
In the mean time XFS maintenance has my full attention.

                                        -Alex

_______________________________________________
xfs mailing list
xfs@oss.sgi.com
http://oss.sgi.com/mailman/listinfo/xfs

Re: [GIT Review] xfs: 2.6.36 fixes

[Dave Chinner]

On Wed, Aug 25, 2010 at 09:17:26PM -0500, Alex Elder wrote:
> On Wed, 2010-08-25 at 14:50 +1000, Dave Chinner wrote:
> > On Wed, Aug 25, 2010 at 11:10:28AM +1000, Dave Chinner wrote:
> > > On Tue, Aug 24, 2010 at 05:54:58PM -0500, Alex Elder wrote:
> > > > On Tue, 2010-08-24 at 12:44 +1000, Dave Chinner wrote:
> > > > > Folks,
> > > > >
> > > > > Can you please cast an eye over the the tree below to check that
> all
> > > > > the bug fixes that need to go into 2.6.36 are there? If so, I'll
> > > > > send a pull request to Linus tomorrow.
> > > > >
> > > > > Note that these are just the outstanding bug fixes that have
> been
> > > > > reviewed (as Linus has again decreed for pull requests after
> -rc1),
> > > > > so it not the complete set of reviewd patches that exist.
> > > > >
> > > > > Cheers,
> > > > >
> > > > > Dave.
> > > >
> > > > All of these commits look good to me.  I will have these
> > > > and all the rest published on the oss.sgi.com tree later
> > > > this week.  I've been on vacation and have gotten behind.
> > >
> > > I'm going to send the pull to Linus, anyway, Alex.
> > >
> > > The problem is deeper than "been on vacation" - regardless of
> > > whether you are on vacation we can't rely on you to do anything
> > > immediately. It might be code review, triage community reported
> > > bugs, pulling patches into the OSS tree or sending stuff to Linus,
> > > but it's always a week or two later than it needs to be. This is
> > > not a new problem, either.
> >
> > So this doesn't get taken the wrong way, this is not a comment on
> > the quality of the work you do, Alex. My concern is the limited
> > amount of time you have available to do the required work and
> > avoiding the bottleneck it causes.
> 
> This has been an ongoing issue, and I concur that it's been a matter
> of giving XFS maintenance the time and attention it requires.
> 
> In July I raised this issue internally and got agreement to give
> this more priority versus other competing demands.  That helped
> some, but there still were some other things that needed my
> dedicated focus, and that disrupted the continuity of my attention
> to XFS maintenance.
> 
> My vacation last week, directly on the heels of my trip to Boston
> the week before (during which I had limited network access) resulted
> in an extended delay whose cause was actually pretty unusual (and
> unfortunately timed). 
> 
> I have now been directed to spend 100% of my time on XFS maintenance
> for the community.  This will not result in 10 minute turnaround but
> it will allow me to offer everything I've got to the cause, without
> interruption.

Wonderful news, Alex! I can't think of a better way to address my
concerns.

/me does a happy dance

> SGI continues to be committed to supporting XFS (for more than just
> its own customers), and in the coming month or so is going to be
> evaluating what more can be done to serve and benefit the community.
> In the mean time XFS maintenance has my full attention.

I'd suggest that one thing we really need to do first is work out a
semi-coherent development roadmap so that we all understand what the
big issues are that we need to solve. I've been noticing the lack of
a definite direction for XFS recently - we've been scratching itches
but not really having an idea of what the next itch is going to be.

I can see some interesting requirements starting to coalesce from
RedHat customers that are just starting to use XFS, and I'm sure
that SGI is in the same boat. I also know that there are many
different lurkers in the community that would like certain features
in XFS to differentiate their products, so perhaps it is time to
flush them all out and work out a future direction that we can work
towards together.....

Cheers,

Dave.
-- 
Dave Chinner
david@fromorbit.com

_______________________________________________
xfs mailing list
xfs@oss.sgi.com
http://oss.sgi.com/mailman/listinfo/xfs