[Cluster-devel] [GFS2 & DLM] Pull request

[Cluster-devel] [GFS2 & DLM] Pull request

[Steven Whitehouse]

Hi,

Please consider pulling the following changes from the GFS2 -fixes git tree,

Steve.

-----------------------------------------------------------------------------------------
The following changes since commit ce9e3d9953c8cb67001719b5516da2928e956be4:
  Linus Torvalds:
        Merge branch 'ubuntu-updates' of master.kernel.org:/.../bcollins/ubuntu-2.6

are found in the git repository at:

  git://git.kernel.org/pub/scm/linux/kernel/git/steve/gfs2-2.6-fixes.git

Adrian Bunk:
      [GFS2] fs/gfs2/dir.c:gfs2_dir_write_data(): remove dead code
      [GFS2] fs/gfs2/ops_fstype.c:gfs2_get_sb_meta(): remove unused variable
      [GFS2] fs/gfs2/dir.c:gfs2_dir_write_data(): don't use an uninitialized variable
      [GFS2] fs/gfs2/ops_fstype.c:fill_super_meta(): fix NULL dereference
      [GFS2] gfs2_dir_read_data(): fix uninitialized variable usage

Al Viro:
      [GFS2] gfs2 endianness bug: be16 assigned to be32 field

Patrick Caulfield:
      [DLM] fix iovec length in recvmsg

Steven Whitehouse:
      [GFS2] Fix bmap to map extents properly

 fs/dlm/lowcomms.c     |    1 +
 fs/gfs2/bmap.c        |   13 +++++++------
 fs/gfs2/bmap.h        |    2 +-
 fs/gfs2/dir.c         |   10 +++-------
 fs/gfs2/log.c         |    6 ++++--
 fs/gfs2/ops_address.c |    6 +++---
 fs/gfs2/ops_fstype.c  |    7 ++-----
 fs/gfs2/quota.c       |    5 +++--
 fs/gfs2/recovery.c    |    5 +++--
 9 files changed, 27 insertions(+), 28 deletions(-)

[Cluster-devel] [GFS2 & DLM] Pull request

[Steven Whitehouse]

Hi,

Please consider pulling the following GFS2 & DLM bug fixes,

Steve.

The following changes since commit d1ed6a3ea10aa7b199c434f6ffd1b6761896567a:
  Linus Torvalds:
        Merge master.kernel.org:/.../davem/net-2.6

are found in the git repository at:

  git://git.kernel.org/pub/scm/linux/kernel/git/steve/gfs2-2.6-fixes.git

Alexey Dobriyan:
      [GFS2] don't panic needlessly

Patrick Caulfield:
      [DLM] Fix kref_put oops
      [DLM] fix oops in kref_put when removing a lockspace

Steven Whitehouse:
      [GFS2] Fix incorrect fs sync behaviour.
      [GFS2] Fix OOM error handling

 fs/dlm/lockspace.c  |   14 +++++++++++++-
 fs/gfs2/inode.c     |    3 +++
 fs/gfs2/main.c      |    4 ++--
 fs/gfs2/ops_super.c |   44 ++++++++++++++++++++++++++++----------------
 4 files changed, 46 insertions(+), 19 deletions(-)

[Cluster-devel] [GFS2 & DLM] Pull request

[Steven Whitehouse]

Hi,

Please consider pulling the following GFS2 & DLM bug fixes. The DLM fixes have
all been very well tested and in addition have been in -mm for some time. The
GFS2 fixes are both trivial "one liners" and have also been in -mm for a little
while,

Steve.

The following changes since commit 1abbfb412b1610ec3a7ec0164108cee01191d9f5:
  Mel Gorman:
        x86_64: fix bad page state in process 'swapper'

are found in the git repository at:

  git://git.kernel.org/pub/scm/linux/kernel/git/steve/gfs2-2.6-fixes.git

David Teigland:
      [DLM] res_recover_locks_count not reset when recover_locks is aborted
      [DLM] status messages ping-pong between unmounted nodes
      [DLM] fix requestqueue race
      [DLM] fix aborted recovery during node removal
      [DLM] fix stopping unstarted recovery
      [DLM] do full recover_locks barrier
      [DLM] clear sbflags on lock master

Steven Whitehouse:
      [GFS2] Fix Kconfig wrt CRC32
      [GFS2] Fix memory allocation in glock.c

 fs/dlm/lock.c         |   16 ++++++++++++----
 fs/dlm/member.c       |    8 ++++++++
 fs/dlm/rcom.c         |    7 ++++---
 fs/dlm/recover.c      |    1 +
 fs/dlm/recoverd.c     |   20 +++++++++++++++++++-
 fs/dlm/requestqueue.c |   21 +++++++++++++++++----
 fs/dlm/requestqueue.h |    2 +-
 fs/gfs2/Kconfig       |    1 +
 fs/gfs2/glock.c       |    2 +-
 9 files changed, 64 insertions(+), 14 deletions(-)

[Cluster-devel] [GFS2] Don't flush everything on fdatasync [70/70]

[Steven Whitehouse]

From 33c3de32872ef3c075e4dac04c0de8f86ac39f6f Mon Sep 17 00:00:00 2001
From: Steven Whitehouse <swhiteho@redhat.com>
Date: Thu, 30 Nov 2006 10:14:32 -0500
Subject: [PATCH] [GFS2] Don't flush everything on fdatasync

The gfs2_fsync() function was doing a journal flush on each
and every call. While this is correct, its also a lot of
overhead. This patch means that on fdatasync flushes we
rely on the VFS to flush the data for us and we don't do
a journal flush unless we really need to.

We have to do a journal flush for stuffed files though because
they have the data and the inode metadata in the same block.
Journaled files also need a journal flush too of course.

Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>
---
 fs/gfs2/ops_file.c |   30 +++++++++++++++++++++++++++---
 1 files changed, 27 insertions(+), 3 deletions(-)

diff --git a/fs/gfs2/ops_file.c b/fs/gfs2/ops_file.c
index c2be216..7bd971b 100644
--- a/fs/gfs2/ops_file.c
+++ b/fs/gfs2/ops_file.c
@@ -22,6 +22,7 @@ #include <linux/gfs2_ondisk.h>
 #include <linux/ext2_fs.h>
 #include <linux/crc32.h>
 #include <linux/lm_interface.h>
+#include <linux/writeback.h>
 #include <asm/uaccess.h>
 
 #include "gfs2.h"
@@ -503,16 +504,39 @@ static int gfs2_close(struct inode *inod
  * @file: the file that points to the dentry (we ignore this)
  * @dentry: the dentry that points to the inode to sync
  *
+ * The VFS will flush "normal" data for us. We only need to worry
+ * about metadata here. For journaled data, we just do a log flush
+ * as we can't avoid it. Otherwise we can just bale out if datasync
+ * is set. For stuffed inodes we must flush the log in order to
+ * ensure that all data is on disk.
+ *
  * Returns: errno
  */
 
 static int gfs2_fsync(struct file *file, struct dentry *dentry, int datasync)
 {
-	struct gfs2_inode *ip = GFS2_I(dentry->d_inode);
+	struct inode *inode = dentry->d_inode;
+	int sync_state = inode->i_state & (I_DIRTY_SYNC|I_DIRTY_DATASYNC);
+	int ret = 0;
+	struct writeback_control wbc = {
+		.sync_mode = WB_SYNC_ALL,
+		.nr_to_write = 0,
+	};
+
+	if (gfs2_is_jdata(GFS2_I(inode))) {
+		gfs2_log_flush(GFS2_SB(inode), GFS2_I(inode)->i_gl);
+		return 0;
+	}
 
-	gfs2_log_flush(ip->i_gl->gl_sbd, ip->i_gl);
+	if (sync_state != 0) {
+		if (!datasync)
+			ret = sync_inode(inode, &wbc);
 
-	return 0;
+		if (gfs2_is_stuffed(GFS2_I(inode)))
+			gfs2_log_flush(GFS2_SB(inode), GFS2_I(inode)->i_gl);
+	}
+
+	return ret;
 }
 
 /**
-- 
1.4.1

[Cluster-devel] Re: [GFS2] Don't flush everything on fdatasync [70/70]

[Andrew Morton]

On Thu, 30 Nov 2006 12:24:08 +0000
Steven Whitehouse <swhiteho@redhat.com> wrote:

>  static int gfs2_fsync(struct file *file, struct dentry *dentry, int datasync)
>  {
> -	struct gfs2_inode *ip = GFS2_I(dentry->d_inode);
> +	struct inode *inode = dentry->d_inode;
> +	int sync_state = inode->i_state & (I_DIRTY_SYNC|I_DIRTY_DATASYNC);
> +	int ret = 0;
> +	struct writeback_control wbc = {
> +		.sync_mode = WB_SYNC_ALL,
> +		.nr_to_write = 0,
> +	};
> +
> +	if (gfs2_is_jdata(GFS2_I(inode))) {
> +		gfs2_log_flush(GFS2_SB(inode), GFS2_I(inode)->i_gl);
> +		return 0;
> +	}
>  
> -	gfs2_log_flush(ip->i_gl->gl_sbd, ip->i_gl);
> +	if (sync_state != 0) {
> +		if (!datasync)
> +			ret = sync_inode(inode, &wbc);

filemap_fdatawrite() would be simpler.

[Cluster-devel] Re: [GFS2] Don't flush everything on fdatasync [70/70]

[Steven Whitehouse]

Hi,

On Thu, 2006-11-30 at 23:01 -0800, Andrew Morton wrote:
> On Thu, 30 Nov 2006 12:24:08 +0000
> Steven Whitehouse <swhiteho@redhat.com> wrote:
> 
> >  static int gfs2_fsync(struct file *file, struct dentry *dentry, int datasync)
> >  {
> > -	struct gfs2_inode *ip = GFS2_I(dentry->d_inode);
> > +	struct inode *inode = dentry->d_inode;
> > +	int sync_state = inode->i_state & (I_DIRTY_SYNC|I_DIRTY_DATASYNC);
> > +	int ret = 0;
> > +	struct writeback_control wbc = {
> > +		.sync_mode = WB_SYNC_ALL,
> > +		.nr_to_write = 0,
> > +	};
> > +
> > +	if (gfs2_is_jdata(GFS2_I(inode))) {
> > +		gfs2_log_flush(GFS2_SB(inode), GFS2_I(inode)->i_gl);
> > +		return 0;
> > +	}
> >  
> > -	gfs2_log_flush(ip->i_gl->gl_sbd, ip->i_gl);
> > +	if (sync_state != 0) {
> > +		if (!datasync)
> > +			ret = sync_inode(inode, &wbc);
> 
> filemap_fdatawrite() would be simpler.

I was taking my cue here from ext3 which does something similar. The
filemap_fdatawrite() is done by the VFS before this is called with a
filemap_fdatawait() afterwards. This was intended to flush the metadata
via (eventually) ->write_inode() although I guess I should be calling
write_inode_now() instead?

Steve.

[Cluster-devel] Re: [GFS2] Don't flush everything on fdatasync [70/70]

[Andrew Morton]

On Fri, 01 Dec 2006 10:58:58 +0000
Steven Whitehouse <swhiteho@redhat.com> wrote:

> On Thu, 2006-11-30 at 23:01 -0800, Andrew Morton wrote:
> > On Thu, 30 Nov 2006 12:24:08 +0000
> > Steven Whitehouse <swhiteho@redhat.com> wrote:
> > 
> > >  static int gfs2_fsync(struct file *file, struct dentry *dentry, int datasync)
> > >  {
> > > -	struct gfs2_inode *ip = GFS2_I(dentry->d_inode);
> > > +	struct inode *inode = dentry->d_inode;
> > > +	int sync_state = inode->i_state & (I_DIRTY_SYNC|I_DIRTY_DATASYNC);
> > > +	int ret = 0;
> > > +	struct writeback_control wbc = {
> > > +		.sync_mode = WB_SYNC_ALL,
> > > +		.nr_to_write = 0,
> > > +	};
> > > +
> > > +	if (gfs2_is_jdata(GFS2_I(inode))) {
> > > +		gfs2_log_flush(GFS2_SB(inode), GFS2_I(inode)->i_gl);
> > > +		return 0;
> > > +	}
> > >  
> > > -	gfs2_log_flush(ip->i_gl->gl_sbd, ip->i_gl);
> > > +	if (sync_state != 0) {
> > > +		if (!datasync)
> > > +			ret = sync_inode(inode, &wbc);
> > 
> > filemap_fdatawrite() would be simpler.
> 
> I was taking my cue here from ext3 which does something similar. The
> filemap_fdatawrite() is done by the VFS before this is called with a
> filemap_fdatawait() afterwards. This was intended to flush the metadata
> via (eventually) ->write_inode() although I guess I should be calling
> write_inode_now() instead?

oh I see, you're jsut trying to write the inode itself, not the pages.

write_inode_now() will write the pages, which you seem to not want to do.
Whatever.  The APIs here are a bit awkward.

[Cluster-devel] Re: [GFS2] Don't flush everything on fdatasync [70/70]

[Steven Whitehouse]

Hi,

On Fri, 2006-12-01 at 11:09 -0800, Andrew Morton wrote:
> On Fri, 01 Dec 2006 10:58:58 +0000
> Steven Whitehouse <swhiteho@redhat.com> wrote:
> 
> > On Thu, 2006-11-30 at 23:01 -0800, Andrew Morton wrote:
> > > On Thu, 30 Nov 2006 12:24:08 +0000
> > > Steven Whitehouse <swhiteho@redhat.com> wrote:
> > > 
> > > >  static int gfs2_fsync(struct file *file, struct dentry *dentry, int datasync)
> > > >  {
> > > > -	struct gfs2_inode *ip = GFS2_I(dentry->d_inode);
> > > > +	struct inode *inode = dentry->d_inode;
> > > > +	int sync_state = inode->i_state & (I_DIRTY_SYNC|I_DIRTY_DATASYNC);
> > > > +	int ret = 0;
> > > > +	struct writeback_control wbc = {
> > > > +		.sync_mode = WB_SYNC_ALL,
> > > > +		.nr_to_write = 0,
> > > > +	};
> > > > +
> > > > +	if (gfs2_is_jdata(GFS2_I(inode))) {
> > > > +		gfs2_log_flush(GFS2_SB(inode), GFS2_I(inode)->i_gl);
> > > > +		return 0;
> > > > +	}
> > > >  
> > > > -	gfs2_log_flush(ip->i_gl->gl_sbd, ip->i_gl);
> > > > +	if (sync_state != 0) {
> > > > +		if (!datasync)
> > > > +			ret = sync_inode(inode, &wbc);
> > > 
> > > filemap_fdatawrite() would be simpler.
> > 
> > I was taking my cue here from ext3 which does something similar. The
> > filemap_fdatawrite() is done by the VFS before this is called with a
> > filemap_fdatawait() afterwards. This was intended to flush the metadata
> > via (eventually) ->write_inode() although I guess I should be calling
> > write_inode_now() instead?
> 
> oh I see, you're jsut trying to write the inode itself, not the pages.
> 
> write_inode_now() will write the pages, which you seem to not want to do.
> Whatever.  The APIs here are a bit awkward.

Well its not so much that we want to avoid it, but since the VFS has
already done that, it should be pretty much a no-op aside from the
waiting which will happen within the call (so that presumably the VFS's
wait after the call will be more or less a no-op).

I notice that fs/sync.c:file_fsync() uses write_inode_now(), so maybe it
is a better choice as it at least means that I don't have to fill in a
struct write_back control for a mapping that I don't really want to
write again,

Steve.

[Cluster-devel] Re: [GFS2] Don't flush everything on fdatasync [70/70]

[Steven Whitehouse]

Hi,

On Fri, 2006-12-01 at 11:09 -0800, Andrew Morton wrote:
> On Fri, 01 Dec 2006 10:58:58 +0000
> Steven Whitehouse <swhiteho@redhat.com> wrote:
> 
> > On Thu, 2006-11-30 at 23:01 -0800, Andrew Morton wrote:
> > > On Thu, 30 Nov 2006 12:24:08 +0000
> > > Steven Whitehouse <swhiteho@redhat.com> wrote:
> > > 
> > > >  static int gfs2_fsync(struct file *file, struct dentry *dentry, int datasync)
> > > >  {
> > > > -	struct gfs2_inode *ip = GFS2_I(dentry->d_inode);
> > > > +	struct inode *inode = dentry->d_inode;
> > > > +	int sync_state = inode->i_state & (I_DIRTY_SYNC|I_DIRTY_DATASYNC);
> > > > +	int ret = 0;
> > > > +	struct writeback_control wbc = {
> > > > +		.sync_mode = WB_SYNC_ALL,
> > > > +		.nr_to_write = 0,
> > > > +	};
> > > > +
> > > > +	if (gfs2_is_jdata(GFS2_I(inode))) {
> > > > +		gfs2_log_flush(GFS2_SB(inode), GFS2_I(inode)->i_gl);
> > > > +		return 0;
> > > > +	}
> > > >  
> > > > -	gfs2_log_flush(ip->i_gl->gl_sbd, ip->i_gl);
> > > > +	if (sync_state != 0) {
> > > > +		if (!datasync)
> > > > +			ret = sync_inode(inode, &wbc);
> > > 
> > > filemap_fdatawrite() would be simpler.
> > 
> > I was taking my cue here from ext3 which does something similar. The
> > filemap_fdatawrite() is done by the VFS before this is called with a
> > filemap_fdatawait() afterwards. This was intended to flush the metadata
> > via (eventually) ->write_inode() although I guess I should be calling
> > write_inode_now() instead?
> 
> oh I see, you're jsut trying to write the inode itself, not the pages.
> 
> write_inode_now() will write the pages, which you seem to not want to do.
> Whatever.  The APIs here are a bit awkward.

I've added a comment to explain whats going on here, and also the
following patch. I know it could be better, but its still an improvement
on what was there before,

Steve.

--------------------------------------------------------------------------------------------

From 34126f9f41901ca9d7d0031c2b11fc0d6c07b72d Mon Sep 17 00:00:00 2001
From: Steven Whitehouse <swhiteho@redhat.com>
Date: Thu, 7 Dec 2006 09:13:14 -0500
Subject: [PATCH] [GFS2] Change gfs2_fsync() to use write_inode_now()

This is a bit better than the previous version of gfs2_fsync()
although it would be better still if we were able to call a
function which only wrote the inode & metadata. Its no big deal
though that this will potentially write the data as well since
the VFS has already done that before calling gfs2_fsync(). I've
also added a comment to explain whats going on here.

Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>
Cc: Andrew Morton <akpm@osdl.org>
---
 fs/gfs2/ops_file.c |   11 ++++++-----
 1 files changed, 6 insertions(+), 5 deletions(-)

diff --git a/fs/gfs2/ops_file.c b/fs/gfs2/ops_file.c
index 7bd971b..b3f1e03 100644
--- a/fs/gfs2/ops_file.c
+++ b/fs/gfs2/ops_file.c
@@ -510,6 +510,11 @@ static int gfs2_close(struct inode *inod
  * is set. For stuffed inodes we must flush the log in order to
  * ensure that all data is on disk.
  *
+ * The call to write_inode_now() is there to write back metadata and
+ * the inode itself. It does also try and write the data, but thats
+ * (hopefully) a no-op due to the VFS having already called filemap_fdatawrite()
+ * for us.
+ *
  * Returns: errno
  */
 
@@ -518,10 +523,6 @@ static int gfs2_fsync(struct file *file,
 	struct inode *inode = dentry->d_inode;
 	int sync_state = inode->i_state & (I_DIRTY_SYNC|I_DIRTY_DATASYNC);
 	int ret = 0;
-	struct writeback_control wbc = {
-		.sync_mode = WB_SYNC_ALL,
-		.nr_to_write = 0,
-	};
 
 	if (gfs2_is_jdata(GFS2_I(inode))) {
 		gfs2_log_flush(GFS2_SB(inode), GFS2_I(inode)->i_gl);
@@ -530,7 +531,7 @@ static int gfs2_fsync(struct file *file,
 
 	if (sync_state != 0) {
 		if (!datasync)
-			ret = sync_inode(inode, &wbc);
+			ret = write_inode_now(inode, 0);
 
 		if (gfs2_is_stuffed(GFS2_I(inode)))
 			gfs2_log_flush(GFS2_SB(inode), GFS2_I(inode)->i_gl);
-- 
1.4.1

[Cluster-devel] [GFS2 & DLM] Pull request

[Steven Whitehouse]

Hi,

All the outstanding issues have now been resolved. Please consider
pulling the following GFS2 & DLM patches,

Steve.

---------------------------------------------------------------------------------------

The following changes since commit 0215ffb08ce99e2bb59eca114a99499a4d06e704:
  Linus Torvalds:
        Linux 2.6.19

are found in the git repository at:

  git://git.kernel.org/pub/scm/linux/kernel/git/steve/gfs2-2.6-nmw.git

Al Viro:
      [GFS2] split gfs2_dinode into on-disk and host variants
      [GFS2] gfs2_dinode_host fields are host-endian
      [GFS2] split gfs2_sb
      [GFS2] fields of gfs2_sb_host are host-endian
      [GFS2] split and annotate gfs2_rgrp
      [GFS2] split and annotate gfs2_inum_range
      [GFS2] split and annotate gfs2_log_head
      [GFS2] split and annotate gfs2_meta_header
      [GFS2] split and annotate gfs_rindex
      [GFS2] split and annotate gfs2_inum
      [GFS2] split and annotate gfs2_quota
      [GFS2] split and annotate gfs2_statfs_change
      [GFS2] split and annotate gfs2_quota_change
      [GFS2] gfs2 misc endianness annotations
      [GFS2] gfs2 __user misannotation fix

David Teigland:
      [DLM] res_recover_locks_count not reset when recover_locks is aborted
      [DLM] status messages ping-pong between unmounted nodes
      [DLM] fix requestqueue race
      [DLM] fix aborted recovery during node removal
      [DLM] fix stopping unstarted recovery
      [DLM] do full recover_locks barrier
      [DLM] clear sbflags on lock master
      [DLM] fix add_requestqueue checking nodes list
      [DLM] fix size of STATUS_REPLY message
      [DLM] don't accept replies to old recovery messages

Patrick Caulfield:
      [DLM] Add support for tcp communications
      [DLM] Fix DLM config
      [DLM] Clean up lowcomms

Randy Dunlap:
      [GFS2] lock function parameter

Russell Cattelan:
      [GFS2] Fix race in logging code
      [GFS2] Remove unused zero_readpage from stuffed_readpage

Ryusuke Konishi:
      [GFS2] fs/gfs2/log.c:log_bmap() fix printk format warning
      [DLM] fix format warnings in rcom.c and recoverd.c

Srinivasa Ds:
      [GFS2] Mount problem with the GFS2 code

Steven Whitehouse:
      [GFS2] Fix crc32 calculation in recovery.c
      [GFS2] Change argument of gfs2_dinode_out
      [GFS2] Change argument to gfs2_dinode_in
      [GFS2] Move gfs2_dinode_in to inode.c
      [GFS2] Change argument to gfs2_dinode_print
      [GFS2] Shrink gfs2_inode (1) - di_header/di_num
      [GFS2] Shrink gfs2_inode (2) - di_major/di_minor
      [GFS2] Shrink gfs2_inode (3) - di_mode
      [GFS2] Shrink gfs2_inode (4) - di_uid/di_gid
      [GFS2] Shrink gfs2_inode (5) - di_nlink
      [GFS2] Shrink gfs2_inode (6) - di_atime/di_mtime/di_ctime
      [GFS2] Shrink gfs2_inode (7) - di_payload_format
      [GFS2] Shrink gfs2_inode (8) - i_vn
      [GFS2] Tidy up 0 initialisations in inode.c
      [GFS2] Don't copy meta_header for rgrp in and out
      [GFS2] Remove unused GL_DUMP flag
      [GFS2] Fix page lock/glock deadlock
      [GFS2] Only set inode flags when required
      [GFS2] Inode number is constant
      [GFS2] Remove gfs2_inode_attr_in
      [GFS2] Fix memory allocation in glock.c
      [GFS2] Tidy up bmap & fix boundary bug
      [GFS2] Remove unused sysfs files
      [GFS2] Remove unused function from inode.c
      [GFS2] Make sentinel dirents compatible with gfs1
      [GFS2] Fix Kconfig wrt CRC32
      [GFS2] Simplify glops functions
      [GFS2] Fix glock ordering on inode creation
      [GFS2] mark_inode_dirty after write to stuffed file
      [GFS2] Fix journal flush problem
      [GFS2] Move gfs2_meta_syncfs() into log.c
      [GFS2] Reduce number of arguments to meta_io.c:getbuf()
      [GFS2] Fix recursive locking in gfs2_permission
      [GFS2] Fix recursive locking in gfs2_getattr
      [GFS2] Remove gfs2_check_acl()
      [GFS2] Add a comment about reading the super block
      [GFS2] Don't flush everything on fdatasync
      [GFS2] Fix indent in recovery.c
      [GFS2] Change gfs2_fsync() to use write_inode_now()

 fs/dlm/Kconfig              |   20 +
 fs/dlm/Makefile             |    4 
 fs/dlm/dlm_internal.h       |    4 
 fs/dlm/lock.c               |   16 -
 fs/dlm/lockspace.c          |    4 
 fs/dlm/lowcomms-sctp.c      | 1227 +++++++++++++++++++++++++++++++++++++++++++
 fs/dlm/lowcomms-tcp.c       | 1189 +++++++++++++++++++++++++++++++++++++++++
 fs/dlm/lowcomms.c           | 1239 -------------------------------------------
 fs/dlm/lowcomms.h           |    2 
 fs/dlm/main.c               |   10 
 fs/dlm/member.c             |    8 
 fs/dlm/rcom.c               |   58 ++
 fs/dlm/recover.c            |    1 
 fs/dlm/recoverd.c           |   44 +-
 fs/dlm/requestqueue.c       |   26 +
 fs/dlm/requestqueue.h       |    2 
 fs/gfs2/Kconfig             |    1 
 fs/gfs2/acl.c               |   39 -
 fs/gfs2/acl.h               |    1 
 fs/gfs2/bmap.c              |  179 +++---
 fs/gfs2/daemon.c            |    7 
 fs/gfs2/dir.c               |   93 ++-
 fs/gfs2/dir.h               |    8 
 fs/gfs2/eaops.c             |    2 
 fs/gfs2/eattr.c             |   66 +-
 fs/gfs2/eattr.h             |    6 
 fs/gfs2/glock.c             |   36 -
 fs/gfs2/glock.h             |    3 
 fs/gfs2/glops.c             |  138 +----
 fs/gfs2/incore.h            |   43 +
 fs/gfs2/inode.c             |  406 +++++---------
 fs/gfs2/inode.h             |   20 -
 fs/gfs2/log.c               |   41 +
 fs/gfs2/log.h               |    2 
 fs/gfs2/lops.c              |   40 +
 fs/gfs2/lops.h              |    2 
 fs/gfs2/meta_io.c           |   46 +-
 fs/gfs2/meta_io.h           |    1 
 fs/gfs2/ondisk.c            |  138 +----
 fs/gfs2/ops_address.c       |   52 +-
 fs/gfs2/ops_dentry.c        |    4 
 fs/gfs2/ops_export.c        |   38 +
 fs/gfs2/ops_export.h        |    2 
 fs/gfs2/ops_file.c          |   66 ++
 fs/gfs2/ops_file.h          |    2 
 fs/gfs2/ops_fstype.c        |    4 
 fs/gfs2/ops_inode.c         |  134 ++---
 fs/gfs2/ops_super.c         |   11 
 fs/gfs2/ops_vm.c            |    2 
 fs/gfs2/quota.c             |   15 -
 fs/gfs2/recovery.c          |   29 +
 fs/gfs2/recovery.h          |    2 
 fs/gfs2/rgrp.c              |   13 
 fs/gfs2/super.c             |   50 +-
 fs/gfs2/super.h             |    6 
 fs/gfs2/sys.c               |    8 
 fs/gfs2/util.h              |    6 
 include/linux/gfs2_ondisk.h |  138 ++++-
 58 files changed, 3442 insertions(+), 2312 deletions(-)
 create mode 100644 fs/dlm/lowcomms-sctp.c
 create mode 100644 fs/dlm/lowcomms-tcp.c
 delete mode 100644 fs/dlm/lowcomms.c

[Cluster-devel] Re: [GFS2] Don't flush everything on fdatasync [70/70]

[Wendy Cheng]

Steven Whitehouse wrote:
> Hi,
>
> On Fri, 2006-12-01 at 11:09 -0800, Andrew Morton wrote:
>   
>>> I was taking my cue here from ext3 which does something similar. The
>>> filemap_fdatawrite() is done by the VFS before this is called with a
>>> filemap_fdatawait() afterwards. This was intended to flush the metadata
>>> via (eventually) ->write_inode() although I guess I should be calling
>>> write_inode_now() instead?
>>>       
>> oh I see, you're jsut trying to write the inode itself, not the pages.
>>
>> write_inode_now() will write the pages, which you seem to not want to do.
>> Whatever.  The APIs here are a bit awkward.
>>     
>
> I've added a comment to explain whats going on here, and also the
> following patch. I know it could be better, but its still an improvement
> on what was there before,
>
>
>   
Steve,

I'm in the middle of something else and don't have upstream kernel 
source handy at this moment. But I read akpm's comment as 
"write_inode_now" would do writepage and that is *not* what you want (?) 
(since vfs has done that before this call is invoked). I vaguely 
recalled I did try write_inode_now() on GFS1 once but had to replace it 
with "sync_inode" on RHEL4 (for the reason that I can't remember at this 
moment). I suggest you keep "sync_inode" (at least for a while until we 
can prove other call can do better). This "sync_inode" has been well 
tested out (with GFS1's fsync call).

There is another issue. It is a gray area. Note that you don't grab any 
glock here ... so if someone *has* written something in other nodes, 
this sync could miss it (?). This depends on how people expects a 
fsync/fdatasync should behave in a cluster filesystem. GFS1 asks for a 
shared lock here so it will force other node to flush the data (I 
personally think this is a more correct behavior). Your call though.

-- Wendy

> --------------------------------------------------------------------------------------------
>
> >From 34126f9f41901ca9d7d0031c2b11fc0d6c07b72d Mon Sep 17 00:00:00 2001
> From: Steven Whitehouse <swhiteho@redhat.com>
> Date: Thu, 7 Dec 2006 09:13:14 -0500
> Subject: [PATCH] [GFS2] Change gfs2_fsync() to use write_inode_now()
>
> This is a bit better than the previous version of gfs2_fsync()
> although it would be better still if we were able to call a
> function which only wrote the inode & metadata. Its no big deal
> though that this will potentially write the data as well since
> the VFS has already done that before calling gfs2_fsync(). I've
> also added a comment to explain whats going on here.
>
> Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>
> Cc: Andrew Morton <akpm@osdl.org>
> ---
>  fs/gfs2/ops_file.c |   11 ++++++-----
>  1 files changed, 6 insertions(+), 5 deletions(-)
>
> diff --git a/fs/gfs2/ops_file.c b/fs/gfs2/ops_file.c
> index 7bd971b..b3f1e03 100644
> --- a/fs/gfs2/ops_file.c
> +++ b/fs/gfs2/ops_file.c
> @@ -510,6 +510,11 @@ static int gfs2_close(struct inode *inod
>   * is set. For stuffed inodes we must flush the log in order to
>   * ensure that all data is on disk.
>   *
> + * The call to write_inode_now() is there to write back metadata and
> + * the inode itself. It does also try and write the data, but thats
> + * (hopefully) a no-op due to the VFS having already called filemap_fdatawrite()
> + * for us.
> + *
>   * Returns: errno
>   */
>  
> @@ -518,10 +523,6 @@ static int gfs2_fsync(struct file *file,
>  	struct inode *inode = dentry->d_inode;
>  	int sync_state = inode->i_state & (I_DIRTY_SYNC|I_DIRTY_DATASYNC);
>  	int ret = 0;
> -	struct writeback_control wbc = {
> -		.sync_mode = WB_SYNC_ALL,
> -		.nr_to_write = 0,
> -	};
>  
>  	if (gfs2_is_jdata(GFS2_I(inode))) {
>  		gfs2_log_flush(GFS2_SB(inode), GFS2_I(inode)->i_gl);
> @@ -530,7 +531,7 @@ static int gfs2_fsync(struct file *file,
>  
>  	if (sync_state != 0) {
>  		if (!datasync)
> -			ret = sync_inode(inode, &wbc);
> +			ret = write_inode_now(inode, 0);
>  
>  		if (gfs2_is_stuffed(GFS2_I(inode)))
>  			gfs2_log_flush(GFS2_SB(inode), GFS2_I(inode)->i_gl);
>   

[Cluster-devel] Re: [GFS2] Don't flush everything on fdatasync [70/70]

[Steven Whitehouse]

Hi,

On Thu, 2006-12-07 at 14:05 -0500, Wendy Cheng wrote:
> Steven Whitehouse wrote:
> > Hi,
> >
> > On Fri, 2006-12-01 at 11:09 -0800, Andrew Morton wrote:
> >   
> >>> I was taking my cue here from ext3 which does something similar. The
> >>> filemap_fdatawrite() is done by the VFS before this is called with a
> >>> filemap_fdatawait() afterwards. This was intended to flush the metadata
> >>> via (eventually) ->write_inode() although I guess I should be calling
> >>> write_inode_now() instead?
> >>>       
> >> oh I see, you're jsut trying to write the inode itself, not the pages.
> >>
> >> write_inode_now() will write the pages, which you seem to not want to do.
> >> Whatever.  The APIs here are a bit awkward.
> >>     
> >
> > I've added a comment to explain whats going on here, and also the
> > following patch. I know it could be better, but its still an improvement
> > on what was there before,
> >
> >
> >   
> Steve,
> 
> I'm in the middle of something else and don't have upstream kernel 
> source handy at this moment. But I read akpm's comment as 
> "write_inode_now" would do writepage and that is *not* what you want (?) 
> (since vfs has done that before this call is invoked). I vaguely 
> recalled I did try write_inode_now() on GFS1 once but had to replace it 
> with "sync_inode" on RHEL4 (for the reason that I can't remember at this 
> moment). I suggest you keep "sync_inode" (at least for a while until we 
> can prove other call can do better). This "sync_inode" has been well 
> tested out (with GFS1's fsync call).
> 
Ok. Its gone upstream now, but I'm happy to revert that change if it
turns out to be a problem. My tests show identical performance between
the two calls, but maybe there is a corner case I missed?

Both calls do writepage() but since the VFS has already done it for us,
the chances of there being any more dirty pages to write is fairly
small, so its unlikely to be much of a problem I think, but I'm willing
to be proved wrong if there is a good reason.

> There is another issue. It is a gray area. Note that you don't grab any 
> glock here ... so if someone *has* written something in other nodes, 
> this sync could miss it (?). This depends on how people expects a 
> fsync/fdatasync should behave in a cluster filesystem. GFS1 asks for a 
> shared lock here so it will force other node to flush the data (I 
> personally think this is a more correct behavior). Your call though.
> 
> -- Wendy
> 
Its a tricky one to deal with. I would expect that the chances of an
application relying on an fsync on one node to cause a cross-cluster
flush is relatively unlikely. It would mean that there would have to be
another communication channel between the processes on the different
nodes through which the node that was writing data could request a flush
and then receive notification that it has finished, otherwise it would
not seem to make any sense. It would seem an odd way to write an
application, but maybe one does exist which does this somewhere.

Delving back into the history it looks like this is a change (with
respect to gfs1) made by Ken rather than myself. I don't mind adding
this feature though, but even so what we have now is still a marked
improvement on what was there previously I think,

Steve.

[Cluster-devel] Re: [GFS2] Don't flush everything on fdatasync [70/70]

[Steven Whitehouse]

Hi,

On Thu, 2006-12-07 at 14:05 -0500, Wendy Cheng wrote:
> Steven Whitehouse wrote:
> > Hi,
> >
> > On Fri, 2006-12-01 at 11:09 -0800, Andrew Morton wrote:
> >   
> >>> I was taking my cue here from ext3 which does something similar. The
> >>> filemap_fdatawrite() is done by the VFS before this is called with a
> >>> filemap_fdatawait() afterwards. This was intended to flush the metadata
> >>> via (eventually) ->write_inode() although I guess I should be calling
> >>> write_inode_now() instead?
> >>>       
> >> oh I see, you're jsut trying to write the inode itself, not the pages.
> >>
> >> write_inode_now() will write the pages, which you seem to not want to do.
> >> Whatever.  The APIs here are a bit awkward.
> >>     
> >
> > I've added a comment to explain whats going on here, and also the
> > following patch. I know it could be better, but its still an improvement
> > on what was there before,
> >
> >
> >   
> Steve,
> 
> I'm in the middle of something else and don't have upstream kernel 
> source handy at this moment. But I read akpm's comment as 
> "write_inode_now" would do writepage and that is *not* what you want (?) 
> (since vfs has done that before this call is invoked). I vaguely 
> recalled I did try write_inode_now() on GFS1 once but had to replace it 
> with "sync_inode" on RHEL4 (for the reason that I can't remember at this 
> moment). I suggest you keep "sync_inode" (at least for a while until we 
> can prove other call can do better). This "sync_inode" has been well 
> tested out (with GFS1's fsync call).
> 
Ok. Its gone upstream now, but I'm happy to revert that change if it
turns out to be a problem. My tests show identical performance between
the two calls, but maybe there is a corner case I missed?

Both calls do writepage() but since the VFS has already done it for us,
the chances of there being any more dirty pages to write is fairly
small, so its unlikely to be much of a problem I think, but I'm willing
to be proved wrong if there is a good reason.

> There is another issue. It is a gray area. Note that you don't grab any 
> glock here ... so if someone *has* written something in other nodes, 
> this sync could miss it (?). This depends on how people expects a 
> fsync/fdatasync should behave in a cluster filesystem. GFS1 asks for a 
> shared lock here so it will force other node to flush the data (I 
> personally think this is a more correct behavior). Your call though.
> 
> -- Wendy
> 
Its a tricky one to deal with. I would expect that the chances of an
application relying on an fsync on one node to cause a cross-cluster
flush is relatively unlikely. It would mean that there would have to be
another communication channel between the processes on the different
nodes through which the node that was writing data could request a flush
and then receive notification that it has finished, otherwise it would
not seem to make any sense. It would seem an odd way to write an
application, but maybe one does exist which does this somewhere.

Delving back into the history it looks like this is a change (with
respect to gfs1) made by Ken rather than myself. I don't mind adding
this feature though, but even so what we have now is still a marked
improvement on what was there previously I think,

Steve.


-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo at vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/

[Cluster-devel] [GFS2 & DLM] Pull request

[Steven Whitehouse]

Hi,

Please consider pulling the following two compile fixes for GFS2/DLM,

Steve.

------------------------------------------------------------------------------------------
The following changes since commit d1998ef38a13c4e74c69df55ccd38b0440c429b2:
  Ben Collins:
        ib_verbs: Use explicit if-else statements to avoid errors with do-while macros

are found in the git repository at:

  git://git.kernel.org/pub/scm/linux/kernel/git/steve/gfs2-2.6-fixes.git

Patrick Caulfield:
      [DLM] fix compile warning

Steven Whitehouse:
      [GFS2] Fix Kconfig

 fs/dlm/lowcomms-tcp.c |    2 +-
 fs/gfs2/Kconfig       |    4 +++-
 2 files changed, 4 insertions(+), 2 deletions(-)

[Cluster-devel] [GFS2 & DLM] Pull request

[Steven Whitehouse]

Hi,

Please consider pulling the following GFS2 & DLM changes. They are as per
the patches posted recently on lkml, except for three minor changes
(two small bug fixes and a function which should have been static) which
are marked with [*] in the list below. All the other patches have
been in -mm, most of them for a number of weeks. None of the patches touch
any code outside the gfs2 and dlm directories,

Steve.

The following changes since commit 62d0cfcb27cf755cebdc93ca95dabc83608007cd:
  Linus Torvalds (1):
        Linux 2.6.20

are found in the git repository at:

  git://git.kernel.org/pub/scm/linux/kernel/git/steve/gfs2-2.6-nmw.git

Adrian Bunk (4):
      [DLM] fs/dlm/lowcomms-tcp.c: remove 2 functions
      [GFS2] make gfs2_change_nlink_i() static
      [GFS2/DLM] fix GFS2 circular dependency
      [GFS2] make gfs2_writepages() static [*]

David Teigland (18):
      [GFS2] don't try to lockfs after shutdown
      [DLM] fix resend rcom lock
      [DLM] fix old rcom messages
      [DLM] add version check
      [DLM] fix send_args() lvb copying
      [DLM] fix receive_request() lvb copying
      [DLM] fix lost flags in stub replies
      [DLM] change some log_error to log_debug
      [DLM] rename dlm_config_info fields
      [DLM] add config entry to enable log_debug
      [DLM] expose dlm_config_info fields in configfs
      [DLM] fix user unlocking
      [DLM] fix master recovery
      [DLM] saved dlm message can be dropped
      [DLM] can miss clearing resend flag
      [GFS2] increase default lock limit
      [GFS2] make lock_dlm drop_count tunable in sysfs
      [DLM] zero new user lvbs

Eric Sandeen (2):
      [GFS2] use CURRENT_TIME_SEC instead of get_seconds in gfs2
      [GFS2] more CURRENT_TIME_SEC

Patrick Caulfield (7):
      [DLM] Fix schedule() calls
      [DLM] Fix spin lock already unlocked bug
      [DLM] Use workqueues for dlm lowcomms
      [DLM] lowcomms tidy
      [DLM] fix lowcomms receiving
      [DLM] Make sock_sem into a mutex
      [DLM] fix softlockup in dlm_recv

Randy Dunlap (2):
      [GFS2/DLM] use sysfs
      [DLM/GFS2] indent help text

Robert Peterson (1):
      [GFS2] gfs2 knows of directories which it chooses not to display

Russell Cattelan (2):
      [GFS2] BZ 217008 fsfuzzer fix.
      [GFS2] Fix unlink deadlocks

S. Wendy Cheng (2):
      [GFS2] Fix change nlink deadlock
      [GFS2] Fix gfs2_rename deadlock

Steven Whitehouse (18):
      [GFS2] Fix DIO deadlock
      [GFS2] Fail over to readpage for stuffed files
      [GFS2] Fix ordering of page disposal vs. glock_dq
      [GFS2] Add writepages for "data=writeback" mounts
      [GFS2] Clean up/speed up readdir
      [GFS2] Remove max_atomic_write tunable
      [GFS2] Shrink gfs2_inode memory by half
      [GFS2] Remove the "greedy" function from glock.[ch]
      [GFS2] Remove unused go_callback operation
      [GFS2] Remove local exclusive glock mode
      [GFS2] Tidy up glops calls
      [GFS2] Remove queue_empty() function
      [GFS2] Compile fix for glock.c
      [GFS2] Fix typo in glock.c
      [GFS2] Fix recursive locking attempt with NFS
      [GFS2] Fix list corruption in lops.c
      [GFS2] Put back semaphore to avoid umount problem
      [GFS2] Unlock page on prepare_write try lock failure [*]

Wendy Cheng (1):
      [GFS2] nfsd readdirplus assertion failure [*]

 fs/dlm/Kconfig                 |   18 +-
 fs/dlm/config.c                |  154 ++++++++++++++++--
 fs/dlm/config.h                |   17 +-
 fs/dlm/dlm_internal.h          |   20 ++-
 fs/dlm/lock.c                  |   87 +++++++---
 fs/dlm/lockspace.c             |   10 +-
 fs/dlm/lowcomms-sctp.c         |  151 ++++++++---------
 fs/dlm/lowcomms-tcp.c          |  361 +++++++++++-----------------------------
 fs/dlm/midcomms.c              |    4 +-
 fs/dlm/rcom.c                  |   85 ++++++----
 fs/dlm/recover.c               |    8 +-
 fs/dlm/recoverd.c              |   22 ++--
 fs/dlm/user.c                  |    9 +
 fs/dlm/util.c                  |    4 +
 fs/gfs2/Kconfig                |   47 +++---
 fs/gfs2/bmap.c                 |   10 +-
 fs/gfs2/dir.c                  |   25 ++--
 fs/gfs2/dir.h                  |   21 +--
 fs/gfs2/eattr.c                |    8 +-
 fs/gfs2/glock.c                |  316 +++++++++---------------------------
 fs/gfs2/glock.h                |   11 --
 fs/gfs2/glops.c                |  136 +++++----------
 fs/gfs2/incore.h               |   18 +--
 fs/gfs2/inode.c                |   61 ++++----
 fs/gfs2/lm.c                   |    8 +-
 fs/gfs2/locking/dlm/lock_dlm.h |    2 +-
 fs/gfs2/locking/dlm/main.c     |    6 -
 fs/gfs2/locking/dlm/mount.c    |    6 +-
 fs/gfs2/locking/dlm/sysfs.c    |   13 ++
 fs/gfs2/lops.c                 |   14 ++-
 fs/gfs2/ops_address.c          |  134 +++++++++------
 fs/gfs2/ops_dentry.c           |   16 ++-
 fs/gfs2/ops_export.c           |   15 +-
 fs/gfs2/ops_file.c             |   52 +------
 fs/gfs2/ops_inode.c            |   55 +++++--
 fs/gfs2/ops_super.c            |   11 +-
 fs/gfs2/ops_vm.c               |   24 +---
 fs/gfs2/super.c                |   16 +--
 fs/gfs2/sys.c                  |   10 -
 39 files changed, 866 insertions(+), 1119 deletions(-)