[PATCH 1/10] merge xfs_unmount into xfs_fs_put_super / xfs_fs_fill_super

[PATCH 1/10] merge xfs_unmount into xfs_fs_put_super / xfs_fs_fill_super

[Christoph Hellwig]

xfs_unmount is small and already pretty Linux specific, so merge it into
the callers.  The real unmount path is simplified a little by doing a
WARN_ON on the xfs_unmount_flush retval directly instead of propagating
the error back to the caller, and the mout failure case in simplified
significantly by removing the forced shudown case and all the dmapi
events that shouldn't be sent because the dmapi mount event hasn't been
sent by that time either.


Signed-off-by: Christoph Hellwig <hch@lst.de>

Index: linux-2.6-xfs/fs/xfs/linux-2.6/xfs_super.c
===================================================================
--- linux-2.6-xfs.orig/fs/xfs/linux-2.6/xfs_super.c	2008-04-25 20:48:31.000000000 +0200
+++ linux-2.6-xfs/fs/xfs/linux-2.6/xfs_super.c	2008-04-25 20:50:16.000000000 +0200
@@ -1087,14 +1087,61 @@ xfs_fs_put_super(
 	struct super_block	*sb)
 {
 	struct xfs_mount	*mp = XFS_M(sb);
+	struct xfs_inode	*rip = mp->m_rootip;
+	int			unmount_event_flags = 0;
 	int			error;
 
 	kthread_stop(mp->m_sync_task);
 
 	xfs_sync(mp, SYNC_ATTR | SYNC_DELWRI);
-	error = xfs_unmount(mp, 0, NULL);
-	if (error)
-		printk("XFS: unmount got error=%d\n", error);
+
+#ifdef HAVE_DMAPI
+	if (mp->m_flags & XFS_MOUNT_DMAPI) {
+		unmount_event_flags =
+			(mp->m_dmevmask & (1 << DM_EVENT_UNMOUNT)) ?
+				0 : DM_FLAGS_UNWANTED;
+		/*
+		 * Ignore error from dmapi here, first unmount is not allowed
+		 * to fail anyway, and second we wouldn't want to fail a
+		 * unmount because of dmapi.
+		 */
+		XFS_SEND_PREUNMOUNT(mp, rip, DM_RIGHT_NULL, rip, DM_RIGHT_NULL,
+				NULL, NULL, 0, 0, unmount_event_flags);
+	}
+#endif
+
+	/*
+	 * Blow away any referenced inode in the filestreams cache.
+	 * This can and will cause log traffic as inodes go inactive
+	 * here.
+	 */
+	xfs_filestream_unmount(mp);
+
+	XFS_bflush(mp->m_ddev_targp);
+	error = xfs_unmount_flush(mp, 0);
+	WARN_ON(error);
+
+	IRELE(rip);
+
+	/*
+	 * If we're forcing a shutdown, typically because of a media error,
+	 * we want to make sure we invalidate dirty pages that belong to
+	 * referenced vnodes as well.
+	 */
+	if (XFS_FORCED_SHUTDOWN(mp)) {
+		error = xfs_sync(mp, SYNC_WAIT | SYNC_CLOSE);
+		ASSERT(error != EFSCORRUPTED);
+	}
+
+	if (mp->m_flags & XFS_MOUNT_DMAPI) {
+		XFS_SEND_UNMOUNT(mp, rip, DM_RIGHT_NULL, 0, 0,
+				unmount_event_flags);
+	}
+
+	xfs_unmountfs(mp, NULL);
+	xfs_qmops_put(mp);
+	xfs_dmops_put(mp);
+	kmem_free(mp, sizeof(xfs_mount_t));
 }
 
 STATIC void
@@ -1400,7 +1447,23 @@ fail_vnrele:
 	}
 
 fail_unmount:
-	xfs_unmount(mp, 0, NULL);
+	/*
+	 * Blow away any referenced inode in the filestreams cache.
+	 * This can and will cause log traffic as inodes go inactive
+	 * here.
+	 */
+	xfs_filestream_unmount(mp);
+
+	XFS_bflush(mp->m_ddev_targp);
+	error = xfs_unmount_flush(mp, 0);
+	WARN_ON(error);
+
+	IRELE(mp->m_rootip);
+
+	xfs_unmountfs(mp, NULL);
+	xfs_qmops_put(mp);
+	xfs_dmops_put(mp);
+	kmem_free(mp, sizeof(xfs_mount_t));
 
 fail_vfsop:
 	kmem_free(args, sizeof(*args));
Index: linux-2.6-xfs/fs/xfs/xfs_vfsops.c
===================================================================
--- linux-2.6-xfs.orig/fs/xfs/xfs_vfsops.c	2008-04-25 20:48:31.000000000 +0200
+++ linux-2.6-xfs/fs/xfs/xfs_vfsops.c	2008-04-25 20:48:32.000000000 +0200
@@ -550,93 +550,6 @@ error0:
 	return error;
 }
 
-int
-xfs_unmount(
-	xfs_mount_t	*mp,
-	int		flags,
-	cred_t		*credp)
-{
-	xfs_inode_t	*rip;
-	bhv_vnode_t	*rvp;
-	int		unmount_event_wanted = 0;
-	int		unmount_event_flags = 0;
-	int		xfs_unmountfs_needed = 0;
-	int		error;
-
-	rip = mp->m_rootip;
-	rvp = XFS_ITOV(rip);
-
-#ifdef HAVE_DMAPI
-	if (mp->m_flags & XFS_MOUNT_DMAPI) {
-		error = XFS_SEND_PREUNMOUNT(mp,
-				rip, DM_RIGHT_NULL, rip, DM_RIGHT_NULL,
-				NULL, NULL, 0, 0,
-				(mp->m_dmevmask & (1<<DM_EVENT_PREUNMOUNT))?
-					0:DM_FLAGS_UNWANTED);
-			if (error)
-				return XFS_ERROR(error);
-		unmount_event_wanted = 1;
-		unmount_event_flags = (mp->m_dmevmask & (1<<DM_EVENT_UNMOUNT))?
-					0 : DM_FLAGS_UNWANTED;
-	}
-#endif
-
-	/*
-	 * Blow away any referenced inode in the filestreams cache.
-	 * This can and will cause log traffic as inodes go inactive
-	 * here.
-	 */
-	xfs_filestream_unmount(mp);
-
-	XFS_bflush(mp->m_ddev_targp);
-	error = xfs_unmount_flush(mp, 0);
-	if (error)
-		goto out;
-
-	ASSERT(vn_count(rvp) == 1);
-
-	/*
-	 * Drop the reference count
-	 */
-	IRELE(rip);
-
-	/*
-	 * If we're forcing a shutdown, typically because of a media error,
-	 * we want to make sure we invalidate dirty pages that belong to
-	 * referenced vnodes as well.
-	 */
-	if (XFS_FORCED_SHUTDOWN(mp)) {
-		error = xfs_sync(mp, SYNC_WAIT | SYNC_CLOSE);
-		ASSERT(error != EFSCORRUPTED);
-	}
-	xfs_unmountfs_needed = 1;
-
-out:
-	/*	Send DMAPI event, if required.
-	 *	Then do xfs_unmountfs() if needed.
-	 *	Then return error (or zero).
-	 */
-	if (unmount_event_wanted) {
-		/* Note: mp structure must still exist for
-		 * XFS_SEND_UNMOUNT() call.
-		 */
-		XFS_SEND_UNMOUNT(mp, error == 0 ? rip : NULL,
-			DM_RIGHT_NULL, 0, error, unmount_event_flags);
-	}
-	if (xfs_unmountfs_needed) {
-		/*
-		 * Call common unmount function to flush to disk
-		 * and free the super block buffer & mount structures.
-		 */
-		xfs_unmountfs(mp, credp);
-		xfs_qmops_put(mp);
-		xfs_dmops_put(mp);
-		kmem_free(mp, sizeof(xfs_mount_t));
-	}
-
-	return XFS_ERROR(error);
-}
-
 STATIC void
 xfs_quiesce_fs(
 	xfs_mount_t		*mp)
Index: linux-2.6-xfs/fs/xfs/xfs_vfsops.h
===================================================================
--- linux-2.6-xfs.orig/fs/xfs/xfs_vfsops.h	2008-04-25 20:48:31.000000000 +0200
+++ linux-2.6-xfs/fs/xfs/xfs_vfsops.h	2008-04-25 20:48:32.000000000 +0200
@@ -10,7 +10,6 @@ struct xfs_mount_args;
 
 int xfs_mount(struct xfs_mount *mp, struct xfs_mount_args *args,
 		struct cred *credp);
-int xfs_unmount(struct xfs_mount *mp, int flags, struct cred *credp);
 int xfs_sync(struct xfs_mount *mp, int flags);
 void xfs_do_force_shutdown(struct xfs_mount *mp, int flags, char *fname,
 		int lnnum);

Re: [PATCH 1/10] merge xfs_unmount into xfs_fs_put_super / xfs_fs_fill_super

[Christoph Hellwig]

On Fri, May 02, 2008 at 12:00:48AM +0200, Christoph Hellwig wrote:
> xfs_unmount is small and already pretty Linux specific, so merge it into
> the callers.  The real unmount path is simplified a little by doing a
> WARN_ON on the xfs_unmount_flush retval directly instead of propagating
> the error back to the caller, and the mout failure case in simplified
> significantly by removing the forced shudown case and all the dmapi
> events that shouldn't be sent because the dmapi mount event hasn't been
> sent by that time either.

Respin ontop of the kmem_free signature change:


Index: linux-2.6-xfs/fs/xfs/linux-2.6/xfs_super.c
===================================================================
--- linux-2.6-xfs.orig/fs/xfs/linux-2.6/xfs_super.c	2008-05-09 08:18:05.000000000 +0200
+++ linux-2.6-xfs/fs/xfs/linux-2.6/xfs_super.c	2008-05-09 08:27:09.000000000 +0200
@@ -1088,14 +1088,61 @@ xfs_fs_put_super(
 	struct super_block	*sb)
 {
 	struct xfs_mount	*mp = XFS_M(sb);
+	struct xfs_inode	*rip = mp->m_rootip;
+	int			unmount_event_flags = 0;
 	int			error;
 
 	kthread_stop(mp->m_sync_task);
 
 	xfs_sync(mp, SYNC_ATTR | SYNC_DELWRI);
-	error = xfs_unmount(mp, 0, NULL);
-	if (error)
-		printk("XFS: unmount got error=%d\n", error);
+
+#ifdef HAVE_DMAPI
+	if (mp->m_flags & XFS_MOUNT_DMAPI) {
+		unmount_event_flags =
+			(mp->m_dmevmask & (1 << DM_EVENT_UNMOUNT)) ?
+				0 : DM_FLAGS_UNWANTED;
+		/*
+		 * Ignore error from dmapi here, first unmount is not allowed
+		 * to fail anyway, and second we wouldn't want to fail a
+		 * unmount because of dmapi.
+		 */
+		XFS_SEND_PREUNMOUNT(mp, rip, DM_RIGHT_NULL, rip, DM_RIGHT_NULL,
+				NULL, NULL, 0, 0, unmount_event_flags);
+	}
+#endif
+
+	/*
+	 * Blow away any referenced inode in the filestreams cache.
+	 * This can and will cause log traffic as inodes go inactive
+	 * here.
+	 */
+	xfs_filestream_unmount(mp);
+
+	XFS_bflush(mp->m_ddev_targp);
+	error = xfs_unmount_flush(mp, 0);
+	WARN_ON(error);
+
+	IRELE(rip);
+
+	/*
+	 * If we're forcing a shutdown, typically because of a media error,
+	 * we want to make sure we invalidate dirty pages that belong to
+	 * referenced vnodes as well.
+	 */
+	if (XFS_FORCED_SHUTDOWN(mp)) {
+		error = xfs_sync(mp, SYNC_WAIT | SYNC_CLOSE);
+		ASSERT(error != EFSCORRUPTED);
+	}
+
+	if (mp->m_flags & XFS_MOUNT_DMAPI) {
+		XFS_SEND_UNMOUNT(mp, rip, DM_RIGHT_NULL, 0, 0,
+				unmount_event_flags);
+	}
+
+	xfs_unmountfs(mp, NULL);
+	xfs_qmops_put(mp);
+	xfs_dmops_put(mp);
+	kmem_free(mp);
 }
 
 STATIC void
@@ -1402,7 +1449,23 @@ fail_vnrele:
 	}
 
 fail_unmount:
-	xfs_unmount(mp, 0, NULL);
+	/*
+	 * Blow away any referenced inode in the filestreams cache.
+	 * This can and will cause log traffic as inodes go inactive
+	 * here.
+	 */
+	xfs_filestream_unmount(mp);
+
+	XFS_bflush(mp->m_ddev_targp);
+	error = xfs_unmount_flush(mp, 0);
+	WARN_ON(error);
+
+	IRELE(mp->m_rootip);
+
+	xfs_unmountfs(mp, NULL);
+	xfs_qmops_put(mp);
+	xfs_dmops_put(mp);
+	kmem_free(mp);
 
 fail_vfsop:
 	kmem_free(args);
Index: linux-2.6-xfs/fs/xfs/xfs_vfsops.c
===================================================================
--- linux-2.6-xfs.orig/fs/xfs/xfs_vfsops.c	2008-05-09 08:16:55.000000000 +0200
+++ linux-2.6-xfs/fs/xfs/xfs_vfsops.c	2008-05-09 08:26:40.000000000 +0200
@@ -558,93 +558,6 @@ error0:
 	return error;
 }
 
-int
-xfs_unmount(
-	xfs_mount_t	*mp,
-	int		flags,
-	cred_t		*credp)
-{
-	xfs_inode_t	*rip;
-	bhv_vnode_t	*rvp;
-	int		unmount_event_wanted = 0;
-	int		unmount_event_flags = 0;
-	int		xfs_unmountfs_needed = 0;
-	int		error;
-
-	rip = mp->m_rootip;
-	rvp = XFS_ITOV(rip);
-
-#ifdef HAVE_DMAPI
-	if (mp->m_flags & XFS_MOUNT_DMAPI) {
-		error = XFS_SEND_PREUNMOUNT(mp,
-				rip, DM_RIGHT_NULL, rip, DM_RIGHT_NULL,
-				NULL, NULL, 0, 0,
-				(mp->m_dmevmask & (1<<DM_EVENT_PREUNMOUNT))?
-					0:DM_FLAGS_UNWANTED);
-			if (error)
-				return XFS_ERROR(error);
-		unmount_event_wanted = 1;
-		unmount_event_flags = (mp->m_dmevmask & (1<<DM_EVENT_UNMOUNT))?
-					0 : DM_FLAGS_UNWANTED;
-	}
-#endif
-
-	/*
-	 * Blow away any referenced inode in the filestreams cache.
-	 * This can and will cause log traffic as inodes go inactive
-	 * here.
-	 */
-	xfs_filestream_unmount(mp);
-
-	XFS_bflush(mp->m_ddev_targp);
-	error = xfs_unmount_flush(mp, 0);
-	if (error)
-		goto out;
-
-	ASSERT(vn_count(rvp) == 1);
-
-	/*
-	 * Drop the reference count
-	 */
-	IRELE(rip);
-
-	/*
-	 * If we're forcing a shutdown, typically because of a media error,
-	 * we want to make sure we invalidate dirty pages that belong to
-	 * referenced vnodes as well.
-	 */
-	if (XFS_FORCED_SHUTDOWN(mp)) {
-		error = xfs_sync(mp, SYNC_WAIT | SYNC_CLOSE);
-		ASSERT(error != EFSCORRUPTED);
-	}
-	xfs_unmountfs_needed = 1;
-
-out:
-	/*	Send DMAPI event, if required.
-	 *	Then do xfs_unmountfs() if needed.
-	 *	Then return error (or zero).
-	 */
-	if (unmount_event_wanted) {
-		/* Note: mp structure must still exist for
-		 * XFS_SEND_UNMOUNT() call.
-		 */
-		XFS_SEND_UNMOUNT(mp, error == 0 ? rip : NULL,
-			DM_RIGHT_NULL, 0, error, unmount_event_flags);
-	}
-	if (xfs_unmountfs_needed) {
-		/*
-		 * Call common unmount function to flush to disk
-		 * and free the super block buffer & mount structures.
-		 */
-		xfs_unmountfs(mp, credp);
-		xfs_qmops_put(mp);
-		xfs_dmops_put(mp);
-		kmem_free(mp);
-	}
-
-	return XFS_ERROR(error);
-}
-
 STATIC void
 xfs_quiesce_fs(
 	xfs_mount_t		*mp)
Index: linux-2.6-xfs/fs/xfs/xfs_vfsops.h
===================================================================
--- linux-2.6-xfs.orig/fs/xfs/xfs_vfsops.h	2008-05-09 08:16:55.000000000 +0200
+++ linux-2.6-xfs/fs/xfs/xfs_vfsops.h	2008-05-09 08:26:11.000000000 +0200
@@ -10,7 +10,6 @@ struct xfs_mount_args;
 
 int xfs_mount(struct xfs_mount *mp, struct xfs_mount_args *args,
 		struct cred *credp);
-int xfs_unmount(struct xfs_mount *mp, int flags, struct cred *credp);
 int xfs_sync(struct xfs_mount *mp, int flags);
 void xfs_do_force_shutdown(struct xfs_mount *mp, int flags, char *fname,
 		int lnnum);

Re: [PATCH 1/10] merge xfs_unmount into xfs_fs_put_super / xfs_fs_fill_super

[David Chinner]

On Fri, May 09, 2008 at 08:31:58AM +0200, Christoph Hellwig wrote:
> Index: linux-2.6-xfs/fs/xfs/xfs_vfsops.h
> ===================================================================
> --- linux-2.6-xfs.orig/fs/xfs/xfs_vfsops.h	2008-05-09 08:16:55.000000000 +0200
> +++ linux-2.6-xfs/fs/xfs/xfs_vfsops.h	2008-05-09 08:26:11.000000000 +0200
> @@ -10,7 +10,6 @@ struct xfs_mount_args;
>  
>  int xfs_mount(struct xfs_mount *mp, struct xfs_mount_args *args,
>  		struct cred *credp);
> -int xfs_unmount(struct xfs_mount *mp, int flags, struct cred *credp);
>  int xfs_sync(struct xfs_mount *mp, int flags);
>  void xfs_do_force_shutdown(struct xfs_mount *mp, int flags, char *fname,
>  		int lnnum);

This hunk does not apply - xfs_mntupdate() is defined in this file in
the current tree.

<sigh>

Oh, this is on top a bunch of other patches you sent a while back
that no-one has picked up. I'll go get those now....

Cheers,

Dave.
-- 
Dave Chinner
Principal Engineer
SGI Australian Software Group

Re: [PATCH 1/10] merge xfs_unmount into xfs_fs_put_super / xfs_fs_fill_super

[David Chinner]

On Fri, May 02, 2008 at 12:00:48AM +0200, Christoph Hellwig wrote:
> xfs_unmount is small and already pretty Linux specific, so merge it into
> the callers.  The real unmount path is simplified a little by doing a
> WARN_ON on the xfs_unmount_flush retval directly instead of propagating
> the error back to the caller, and the mout failure case in simplified
> significantly by removing the forced shudown case and all the dmapi
> events that shouldn't be sent because the dmapi mount event hasn't been
> sent by that time either.
> 
> 
> Signed-off-by: Christoph Hellwig <hch@lst.de>
> 
> Index: linux-2.6-xfs/fs/xfs/linux-2.6/xfs_super.c
> ===================================================================
> --- linux-2.6-xfs.orig/fs/xfs/linux-2.6/xfs_super.c	2008-04-25 20:48:31.000000000 +0200
> +++ linux-2.6-xfs/fs/xfs/linux-2.6/xfs_super.c	2008-04-25 20:50:16.000000000 +0200
> @@ -1087,14 +1087,61 @@ xfs_fs_put_super(
>  	struct super_block	*sb)
>  {
>  	struct xfs_mount	*mp = XFS_M(sb);
> +	struct xfs_inode	*rip = mp->m_rootip;
> +	int			unmount_event_flags = 0;
>  	int			error;
>  
>  	kthread_stop(mp->m_sync_task);
>  
>  	xfs_sync(mp, SYNC_ATTR | SYNC_DELWRI);
> -	error = xfs_unmount(mp, 0, NULL);
> -	if (error)
> -		printk("XFS: unmount got error=%d\n", error);
> +
> +#ifdef HAVE_DMAPI
> +	if (mp->m_flags & XFS_MOUNT_DMAPI) {
> +		unmount_event_flags =
> +			(mp->m_dmevmask & (1 << DM_EVENT_UNMOUNT)) ?
> +				0 : DM_FLAGS_UNWANTED;
> +		/*
> +		 * Ignore error from dmapi here, first unmount is not allowed
> +		 * to fail anyway, and second we wouldn't want to fail a
> +		 * unmount because of dmapi.
> +		 */
> +		XFS_SEND_PREUNMOUNT(mp, rip, DM_RIGHT_NULL, rip, DM_RIGHT_NULL,
> +				NULL, NULL, 0, 0, unmount_event_flags);
> +	}
> +#endif
> +
> +	/*
> +	 * Blow away any referenced inode in the filestreams cache.
> +	 * This can and will cause log traffic as inodes go inactive
> +	 * here.
> +	 */
> +	xfs_filestream_unmount(mp);
> +
> +	XFS_bflush(mp->m_ddev_targp);
> +	error = xfs_unmount_flush(mp, 0);
> +	WARN_ON(error);
> +
> +	IRELE(rip);
> +
> +	/*
> +	 * If we're forcing a shutdown, typically because of a media error,
> +	 * we want to make sure we invalidate dirty pages that belong to
> +	 * referenced vnodes as well.
> +	 */
> +	if (XFS_FORCED_SHUTDOWN(mp)) {
> +		error = xfs_sync(mp, SYNC_WAIT | SYNC_CLOSE);
> +		ASSERT(error != EFSCORRUPTED);
> +	}
> +
> +	if (mp->m_flags & XFS_MOUNT_DMAPI) {
> +		XFS_SEND_UNMOUNT(mp, rip, DM_RIGHT_NULL, 0, 0,
> +				unmount_event_flags);
> +	}

#ifdef HAVE_DMAPI around this chunk? I know the old code didn't,
but it would then match the pre-unmount event hunk above...

Cheers,

Dave.
-- 
Dave Chinner
Principal Engineer
SGI Australian Software Group

Re: [PATCH 1/10] merge xfs_unmount into xfs_fs_put_super / xfs_fs_fill_super

[Christoph Hellwig]

On Mon, May 12, 2008 at 12:21:38PM +1000, David Chinner wrote:
> > +	if (mp->m_flags & XFS_MOUNT_DMAPI) {
> > +		XFS_SEND_UNMOUNT(mp, rip, DM_RIGHT_NULL, 0, 0,
> > +				unmount_event_flags);
> > +	}
> 
> #ifdef HAVE_DMAPI around this chunk? I know the old code didn't,
> but it would then match the pre-unmount event hunk above...

We don't have #ifdefs around the other dmapi namespace bits.  In fact
I'd prefer to kill the one above aswell.  It's on my todo list because
I wan't to prove that using the normal event enabled macros are safe
in the unmount path first and just switch to those.