2007-10-05 15:31:32

by Cordenner jean noel

[permalink] [raw]
Subject: [PATCH 1/2] i_version update - vfs part

Hi,

This is an update of the i_version patch.
The i_version field is a 64bit counter that is set on every inode
creation and that is incremented every time the inode data is modified
(similarly to the "ctime" time-stamp).
The aim is to fulfill a NFSv4 requirement for rfc3530:
"5.5. Mandatory Attributes - Definitions
Name # DataType Access Description
___________________________________________________________________
change 3 uint64 READ A value created by the
server that the client can use to determine if file
data, directory contents or attributes of the object
have been modified. The servermay return the object's
time_metadata attribute for this attribute's value but
only if the filesystem object can not be updated more
frequently than the resolution of time_metadata.
"

This first part deals with adding a flag in the super block and incrementing the i_version in the vfs.

Signed-off-by: Jean Noel Cordenner <[email protected]>
---
fs/inode.c | 23 +++++++++++++++++++++++
fs/libfs.c | 12 ++++++++++++
include/linux/fs.h | 3 +++
3 files changed, 38 insertions(+)

Index: linux-2.6.23-rc8-ext4-i_version/fs/inode.c
===================================================================
--- linux-2.6.23-rc8-ext4-i_version.orig/fs/inode.c 2007-09-26 14:41:41.000000000 +0200
+++ linux-2.6.23-rc8-ext4-i_version/fs/inode.c 2007-10-05 16:14:41.000000000 +0200
@@ -1216,6 +1216,24 @@
EXPORT_SYMBOL(touch_atime);

/**
+ * inode_inc_iversion - increments i_version
+ * @inode: inode that need to be updated
+ *
+ * Every time the inode is modified, the i_version field
+ * will be incremented.
+ * The filesystem has to be mounted with i_version flag
+ *
+ */
+
+void inode_inc_iversion(struct inode *inode)
+{
+ spin_lock(&inode->i_lock);
+ inode->i_version++;
+ spin_unlock(&inode->i_lock);
+}
+EXPORT_SYMBOL(inode_inc_iversion);
+
+/**
* file_update_time - update mtime and ctime time
* @file: file accessed
*
@@ -1249,6 +1267,11 @@
sync_it = 1;
}

+ if (IS_I_VERSION(inode)) {
+ inode_inc_iversion(inode);
+ sync_it = 1;
+ }
+
if (sync_it)
mark_inode_dirty_sync(inode);
}
Index: linux-2.6.23-rc8-ext4-i_version/fs/libfs.c
===================================================================
--- linux-2.6.23-rc8-ext4-i_version.orig/fs/libfs.c 2007-07-09 01:32:17.000000000 +0200
+++ linux-2.6.23-rc8-ext4-i_version/fs/libfs.c 2007-09-26 14:51:08.000000000 +0200
@@ -255,6 +255,10 @@
struct inode *inode = old_dentry->d_inode;

inode->i_ctime = dir->i_ctime = dir->i_mtime = CURRENT_TIME;
+ if (IS_I_VERSION(inode)) {
+ inode_inc_iversion(inode);
+ inode_inc_iversion(dir);
+ }
inc_nlink(inode);
atomic_inc(&inode->i_count);
dget(dentry);
@@ -287,6 +291,10 @@
struct inode *inode = dentry->d_inode;

inode->i_ctime = dir->i_ctime = dir->i_mtime = CURRENT_TIME;
+ if (IS_I_VERSION(inode)) {
+ inode_inc_iversion(inode);
+ inode_inc_iversion(dir);
+ }
drop_nlink(inode);
dput(dentry);
return 0;
@@ -323,6 +331,10 @@

old_dir->i_ctime = old_dir->i_mtime = new_dir->i_ctime =
new_dir->i_mtime = inode->i_ctime = CURRENT_TIME;
+ if (IS_I_VERSION(old_dir)) {
+ inode_inc_iversion(old_dir);
+ inode_inc_iversion(new_dir);
+ }

return 0;
}
Index: linux-2.6.23-rc8-ext4-i_version/include/linux/fs.h
===================================================================
--- linux-2.6.23-rc8-ext4-i_version.orig/include/linux/fs.h 2007-09-26 14:46:15.000000000 +0200
+++ linux-2.6.23-rc8-ext4-i_version/include/linux/fs.h 2007-09-26 14:51:08.000000000 +0200
@@ -123,6 +123,7 @@
#define MS_SLAVE (1<<19) /* change to slave */
#define MS_SHARED (1<<20) /* change to shared */
#define MS_RELATIME (1<<21) /* Update atime relative to mtime/ctime. */
+#define MS_I_VERSION (1<<22) /* Update inode i_version field */
#define MS_ACTIVE (1<<30)
#define MS_NOUSER (1<<31)

@@ -172,6 +173,7 @@
((inode)->i_flags & (S_SYNC|S_DIRSYNC)))
#define IS_MANDLOCK(inode) __IS_FLG(inode, MS_MANDLOCK)
#define IS_NOATIME(inode) __IS_FLG(inode, MS_RDONLY|MS_NOATIME)
+#define IS_I_VERSION(inode) __IS_FLG(inode, MS_I_VERSION)

#define IS_NOQUOTA(inode) ((inode)->i_flags & S_NOQUOTA)
#define IS_APPEND(inode) ((inode)->i_flags & S_APPEND)
@@ -1284,6 +1286,7 @@
mark_inode_dirty(inode);
}

+extern void inode_inc_iversion(struct inode *inode);
extern void touch_atime(struct vfsmount *mnt, struct dentry *dentry);
static inline void file_accessed(struct file *file)
{



2007-10-06 00:59:30

by Mingming Cao

[permalink] [raw]
Subject: Re: [PATCH 1/2] i_version update - vfs part

On Fri, 2007-10-05 at 17:28 +0200, Cordenner jean noel wrote:
> Hi,
>
Hi Jean Noel,

> This is an update of the i_version patch.

Just to make sure, is this vfs patch and next ext4 patch together going
to replace the 4 inode-version related patches currently in
ext4-patch-queue (and git tree)?

> The i_version field is a 64bit counter that is set on every inode
> creation and that is incremented every time the inode data is modified
> (similarly to the "ctime" time-stamp).
> The aim is to fulfill a NFSv4 requirement for rfc3530:
> "5.5. Mandatory Attributes - Definitions
> Name # DataType Access Description
> ___________________________________________________________________
> change 3 uint64 READ A value created by the
> server that the client can use to determine if file
> data, directory contents or attributes of the object
> have been modified. The servermay return the object's
> time_metadata attribute for this attribute's value but
> only if the filesystem object can not be updated more
> frequently than the resolution of time_metadata.
> "
>
> This first part deals with adding a flag in the super block and incrementing the i_version in the vfs.
>
> Signed-off-by: Jean Noel Cordenner <[email protected]>
> ---
> fs/inode.c | 23 +++++++++++++++++++++++
> fs/libfs.c | 12 ++++++++++++
> include/linux/fs.h | 3 +++
> 3 files changed, 38 insertions(+)
>
> Index: linux-2.6.23-rc8-ext4-i_version/fs/inode.c
> ===================================================================
> --- linux-2.6.23-rc8-ext4-i_version.orig/fs/inode.c 2007-09-26 14:41:41.000000000 +0200
> +++ linux-2.6.23-rc8-ext4-i_version/fs/inode.c 2007-10-05 16:14:41.000000000 +0200
> @@ -1216,6 +1216,24 @@
> EXPORT_SYMBOL(touch_atime);
>
> /**
> + * inode_inc_iversion - increments i_version
> + * @inode: inode that need to be updated
> + *
> + * Every time the inode is modified, the i_version field
> + * will be incremented.
> + * The filesystem has to be mounted with i_version flag
> + *
> + */
> +
> +void inode_inc_iversion(struct inode *inode)
> +{
> + spin_lock(&inode->i_lock);
> + inode->i_version++;
> + spin_unlock(&inode->i_lock);
> +}

I suspect we need a lock here, the places where need to update the
inode->i_version are already doing update for inode, mostly protected by
i_mutex.

You could remove the above function and update the counter directly at
the places it need to.

> +EXPORT_SYMBOL(inode_inc_iversion);
> +

Seems unnecessary.

> +/**
> * file_update_time - update mtime and ctime time
> * @file: file accessed
> *
> @@ -1249,6 +1267,11 @@
> sync_it = 1;
> }
>
> + if (IS_I_VERSION(inode)) {
> + inode_inc_iversion(inode);
> + sync_it = 1;
> + }
> +
> if (sync_it)
> mark_inode_dirty_sync(inode);
> }
> Index: linux-2.6.23-rc8-ext4-i_version/fs/libfs.c
> ===================================================================
> --- linux-2.6.23-rc8-ext4-i_version.orig/fs/libfs.c 2007-07-09 01:32:17.000000000 +0200
> +++ linux-2.6.23-rc8-ext4-i_version/fs/libfs.c 2007-09-26 14:51:08.000000000 +0200
> @@ -255,6 +255,10 @@
> struct inode *inode = old_dentry->d_inode;
>
> inode->i_ctime = dir->i_ctime = dir->i_mtime = CURRENT_TIME;
> + if (IS_I_VERSION(inode)) {
> + inode_inc_iversion(inode);
> + inode_inc_iversion(dir);
> + }
> inc_nlink(inode);
> atomic_inc(&inode->i_count);
> dget(dentry);
> @@ -287,6 +291,10 @@
> struct inode *inode = dentry->d_inode;
>
> inode->i_ctime = dir->i_ctime = dir->i_mtime = CURRENT_TIME;
> + if (IS_I_VERSION(inode)) {
> + inode_inc_iversion(inode);
> + inode_inc_iversion(dir);
> + }
> drop_nlink(inode);
> dput(dentry);
> return 0;
> @@ -323,6 +331,10 @@
>
> old_dir->i_ctime = old_dir->i_mtime = new_dir->i_ctime =
> new_dir->i_mtime = inode->i_ctime = CURRENT_TIME;
> + if (IS_I_VERSION(old_dir)) {
> + inode_inc_iversion(old_dir);
> + inode_inc_iversion(new_dir);
> + }
>
> return 0;
> }

Need to update the counter in libfs.c?

> Index: linux-2.6.23-rc8-ext4-i_version/include/linux/fs.h
> ===================================================================
> --- linux-2.6.23-rc8-ext4-i_version.orig/include/linux/fs.h 2007-09-26 14:46:15.000000000 +0200
> +++ linux-2.6.23-rc8-ext4-i_version/include/linux/fs.h 2007-09-26 14:51:08.000000000 +0200
> @@ -123,6 +123,7 @@
> #define MS_SLAVE (1<<19) /* change to slave */
> #define MS_SHARED (1<<20) /* change to shared */
> #define MS_RELATIME (1<<21) /* Update atime relative to mtime/ctime. */
> +#define MS_I_VERSION (1<<22) /* Update inode i_version field */
> #define MS_ACTIVE (1<<30)
> #define MS_NOUSER (1<<31)
>
> @@ -172,6 +173,7 @@
> ((inode)->i_flags & (S_SYNC|S_DIRSYNC)))
> #define IS_MANDLOCK(inode) __IS_FLG(inode, MS_MANDLOCK)
> #define IS_NOATIME(inode) __IS_FLG(inode, MS_RDONLY|MS_NOATIME)
> +#define IS_I_VERSION(inode) __IS_FLG(inode, MS_I_VERSION)
>
> #define IS_NOQUOTA(inode) ((inode)->i_flags & S_NOQUOTA)
> #define IS_APPEND(inode) ((inode)->i_flags & S_APPEND)
> @@ -1284,6 +1286,7 @@
> mark_inode_dirty(inode);
> }
>
> +extern void inode_inc_iversion(struct inode *inode);
> extern void touch_atime(struct vfsmount *mnt, struct dentry *dentry);
> static inline void file_accessed(struct file *file)
> {
>
>
> -
> To unsubscribe from this list: send the line "unsubscribe linux-ext4" in
> the body of a message to [email protected]
> More majordomo info at http://vger.kernel.org/majordomo-info.html

2007-10-25 17:09:26

by Cordenner jean noel

[permalink] [raw]
Subject: Re: [PATCH 1/2] i_version update - vfs part

Hi,

This is an update of the previous patches on the ext4 git tree, the 2
coming patches applies at the end of the current ext4-patch-queue, and
replaces the inode-version related patches:
64-bit-i_version.patch
i_version_hi.patch
ext4_i_version_hi_2.patch
i_version_update_ext4.patch

The first part deals with the vfs part.
The i_version field of the inode is changed to be a 64-bit counter that
is set on every inode creation and that is incremented every time the
inode data is modified (similarly to the "ctime" time-stamp).
The aim is to fulfill a NFSv4 requirement for rfc3530.
This first part concerns the vfs, it converts the 32-bit i_version in
the generic inode to a 64-bit, a flag is added in the super block in
order to check if the feature is enabled and the i_version is
incremented in the vfs.

Signed-off-by: Mingming Cao <[email protected]>
Signed-off-by: Jean Noel Cordenner <[email protected]>
Signed-off-by: Kalpak Shah <[email protected]>
---
fs/inode.c | 22 ++++++++++++++++++++++
include/linux/fs.h | 5 ++++-
2 files changed, 26 insertions(+), 1 deletion(-)

Index: linux-2.6.23-ext4-1/include/linux/fs.h
===================================================================
--- linux-2.6.23-ext4-1.orig/include/linux/fs.h 2007-10-25
16:25:23.000000000 +0200
+++ linux-2.6.23-ext4-1/include/linux/fs.h 2007-10-25 16:25:53.000000000
+0200
@@ -123,6 +123,7 @@
#define MS_SLAVE (1<<19) /* change to slave */
#define MS_SHARED (1<<20) /* change to shared */
#define MS_RELATIME (1<<21) /* Update atime relative to mtime/ctime. */
+#define MS_I_VERSION (1<<22) /* Update inode I_version field */
#define MS_ACTIVE (1<<30)
#define MS_NOUSER (1<<31)

@@ -172,6 +173,7 @@
((inode)->i_flags & (S_SYNC|S_DIRSYNC)))
#define IS_MANDLOCK(inode) __IS_FLG(inode, MS_MANDLOCK)
#define IS_NOATIME(inode) __IS_FLG(inode, MS_RDONLY|MS_NOATIME)
+#define IS_I_VERSION(inode) __IS_FLG(inode, MS_I_VERSION)

#define IS_NOQUOTA(inode) ((inode)->i_flags & S_NOQUOTA)
#define IS_APPEND(inode) ((inode)->i_flags & S_APPEND)
@@ -541,7 +543,7 @@
uid_t i_uid;
gid_t i_gid;
dev_t i_rdev;
- unsigned long i_version;
+ u64 i_version;
loff_t i_size;
#ifdef __NEED_I_SIZE_ORDERED
seqcount_t i_size_seqcount;
@@ -1284,6 +1286,7 @@
mark_inode_dirty(inode);
}

+extern void inode_inc_iversion(struct inode *inode);
extern void touch_atime(struct vfsmount *mnt, struct dentry *dentry);
static inline void file_accessed(struct file *file)
{
Index: linux-2.6.23-ext4-1/fs/inode.c
===================================================================
--- linux-2.6.23-ext4-1.orig/fs/inode.c 2007-10-25 16:15:52.000000000
+0200
+++ linux-2.6.23-ext4-1/fs/inode.c 2007-10-25 16:25:53.000000000 +0200
@@ -1216,6 +1216,24 @@
EXPORT_SYMBOL(touch_atime);

/**
+ * inode_inc_iversion - increments i_version
+ * @inode: inode that need to be updated
+ *
+ * Every time the inode is modified, the i_version field
+ * will be incremented.
+ * The filesystem has to be mounted with i_version flag
+ *
+ */
+
+void inode_inc_iversion(struct inode *inode)
+{
+ spin_lock(&inode->i_lock);
+ inode->i_version++;
+ spin_unlock(&inode->i_lock);
+}
+
+/**
* file_update_time - update mtime and ctime time
* @file: file accessed
*
@@ -1249,6 +1267,11 @@
sync_it = 1;
}

+ if (IS_I_VERSION(inode)) {
+ inode_inc_iversion(inode);
+ sync_it = 1;
+ }
+
if (sync_it)
mark_inode_dirty_sync(inode);
}


2007-10-25 21:05:59

by Mingming Cao

[permalink] [raw]
Subject: Re: [PATCH 1/2] i_version update - vfs part

On Thu, 2007-10-25 at 19:04 +0200, Cordenner jean noel wrote:
> Hi,
>
> This is an update of the previous patches on the ext4 git tree, the 2
> coming patches applies at the end of the current ext4-patch-queue, and
> replaces the inode-version related patches:
> 64-bit-i_version.patch
> i_version_hi.patch
> ext4_i_version_hi_2.patch
> i_version_update_ext4.patch
>
> The first part deals with the vfs part.
> The i_version field of the inode is changed to be a 64-bit counter that
> is set on every inode creation and that is incremented every time the
> inode data is modified (similarly to the "ctime" time-stamp).
> The aim is to fulfill a NFSv4 requirement for rfc3530.
> This first part concerns the vfs, it converts the 32-bit i_version in
> the generic inode to a 64-bit, a flag is added in the super block in
> order to check if the feature is enabled and the i_version is
> incremented in the vfs.
>
Thanks for reposting it.

> Index: linux-2.6.23-ext4-1/fs/inode.c
> ===================================================================
> --- linux-2.6.23-ext4-1.orig/fs/inode.c 2007-10-25 16:15:52.000000000
> +0200
> +++ linux-2.6.23-ext4-1/fs/inode.c 2007-10-25 16:25:53.000000000 +0200
> @@ -1216,6 +1216,24 @@
> EXPORT_SYMBOL(touch_atime);
>
> /**
> + * inode_inc_iversion - increments i_version
> + * @inode: inode that need to be updated
> + *
> + * Every time the inode is modified, the i_version field
> + * will be incremented.
> + * The filesystem has to be mounted with i_version flag
> + *
> + */
> +
> +void inode_inc_iversion(struct inode *inode)
> +{
> + spin_lock(&inode->i_lock);
> + inode->i_version++;
> + spin_unlock(&inode->i_lock);
> +}

I wonder do we really need i_lock here for inode versioning update?

Understand this is a 64 bit counter, but file_update_time() and
ext4_mark_inode_dirty() (where the inode version is updated) is called
on the file write path so i_mutex should be hold all the time. As long
as the read patch holding i_mutex everything should be fine, isn't it?

Have you get a chance to check the performance impact to ext4?

> +
> +/**
> * file_update_time - update mtime and ctime time
> * @file: file accessed
> *
> @@ -1249,6 +1267,11 @@
> sync_it = 1;
> }
>
> + if (IS_I_VERSION(inode)) {
> + inode_inc_iversion(inode);
> + sync_it = 1;
> + }
> +
> if (sync_it)
> mark_inode_dirty_sync(inode);
> }
>
>
> -
> To unsubscribe from this list: send the line "unsubscribe linux-ext4" in
> the body of a message to [email protected]
> More majordomo info at http://vger.kernel.org/majordomo-info.html

2007-11-20 16:43:40

by Cordenner jean noel

[permalink] [raw]
Subject: Re: [PATCH 1/2] i_version update - vfs part

Hi,

Here you can find some performance comparison in terms of CPU
utilization and transactions per second (using FFSB) on ext4 filesystem
with and without i_version option.

http://bullopensource.org/ext4/20071116/ffsb-write.html

regards,
Jean noel