[RFC PATCH -V2 0/17] Buffered write and writeable mmap support for 9P

[RFC PATCH -V2 0/17] Buffered write and writeable mmap support for 9P

[Aneesh Kumar K.V]

The patch series implement buffered write and writeable mmap for 9P

Changes from V1:
a) Add fsync and syncfs support

-aneesh

[RFC PATCH -V2 01/17] fs/9p: set the cached file_operations struct during inode init

[Aneesh Kumar K.V]

With the old code we were not setting the file->f_op
with cached file operations during creat.

Signed-off-by: Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
---
 fs/9p/v9fs_vfs.h  |    2 ++
 fs/9p/vfs_file.c  |   18 ++++--------------
 fs/9p/vfs_inode.c |   10 ++++++++--
 3 files changed, 14 insertions(+), 16 deletions(-)

diff --git a/fs/9p/v9fs_vfs.h b/fs/9p/v9fs_vfs.h
index b789f8e..e4d5540 100644
--- a/fs/9p/v9fs_vfs.h
+++ b/fs/9p/v9fs_vfs.h
@@ -45,6 +45,8 @@ extern const struct file_operations v9fs_dir_operations;
 extern const struct file_operations v9fs_dir_operations_dotl;
 extern const struct dentry_operations v9fs_dentry_operations;
 extern const struct dentry_operations v9fs_cached_dentry_operations;
+extern const struct file_operations v9fs_cached_file_operations;
+extern const struct file_operations v9fs_cached_file_operations_dotl;
 
 #ifdef CONFIG_9P_FSCACHE
 struct inode *v9fs_alloc_inode(struct super_block *sb);
diff --git a/fs/9p/vfs_file.c b/fs/9p/vfs_file.c
index 240c306..6a67100 100644
--- a/fs/9p/vfs_file.c
+++ b/fs/9p/vfs_file.c
@@ -44,9 +44,6 @@
 #include "fid.h"
 #include "cache.h"
 
-static const struct file_operations v9fs_cached_file_operations;
-static const struct file_operations v9fs_cached_file_operations_dotl;
-
 /**
  * v9fs_file_open - open a file (or directory)
  * @inode: inode to be opened
@@ -89,19 +86,12 @@ int v9fs_file_open(struct inode *inode, struct file *file)
 	}
 
 	file->private_data = fid;
+#ifdef CONFIG_9P_FSCACHE
 	if ((fid->qid.version) && (v9ses->cache)) {
 		P9_DPRINTK(P9_DEBUG_VFS, "cached");
-		/* enable cached file options */
-		if(file->f_op == &v9fs_file_operations)
-			file->f_op = &v9fs_cached_file_operations;
-		else if (file->f_op == &v9fs_file_operations_dotl)
-			file->f_op = &v9fs_cached_file_operations_dotl;
-
-#ifdef CONFIG_9P_FSCACHE
 		v9fs_cache_inode_set_cookie(inode, file);
-#endif
 	}
-
+#endif
 	return 0;
 }
 
@@ -505,7 +495,7 @@ int v9fs_file_fsync_dotl(struct file *filp, int datasync)
 	return retval;
 }
 
-static const struct file_operations v9fs_cached_file_operations = {
+const struct file_operations v9fs_cached_file_operations = {
 	.llseek = generic_file_llseek,
 	.read = do_sync_read,
 	.aio_read = generic_file_aio_read,
@@ -517,7 +507,7 @@ static const struct file_operations v9fs_cached_file_operations = {
 	.fsync = v9fs_file_fsync,
 };
 
-static const struct file_operations v9fs_cached_file_operations_dotl = {
+const struct file_operations v9fs_cached_file_operations_dotl = {
 	.llseek = generic_file_llseek,
 	.read = do_sync_read,
 	.aio_read = generic_file_aio_read,
diff --git a/fs/9p/vfs_inode.c b/fs/9p/vfs_inode.c
index b76a40b..45e7ea4 100644
--- a/fs/9p/vfs_inode.c
+++ b/fs/9p/vfs_inode.c
@@ -292,10 +292,16 @@ struct inode *v9fs_get_inode(struct super_block *sb, int mode)
 	case S_IFREG:
 		if (v9fs_proto_dotl(v9ses)) {
 			inode->i_op = &v9fs_file_inode_operations_dotl;
-			inode->i_fop = &v9fs_file_operations_dotl;
+			if (v9ses->cache)
+				inode->i_fop = &v9fs_cached_file_operations_dotl;
+			else
+				inode->i_fop = &v9fs_file_operations_dotl;
 		} else {
 			inode->i_op = &v9fs_file_inode_operations;
-			inode->i_fop = &v9fs_file_operations;
+			if (v9ses->cache)
+				inode->i_fop = &v9fs_cached_file_operations;
+			else
+				inode->i_fop = &v9fs_file_operations;
 		}
 
 		break;
-- 
1.7.1

Re: [V9fs-developer] [RFC PATCH -V2 01/17] fs/9p: set the cached file_operations struct during inode init

[Eric Van Hensbergen]

On Sat, Feb 5, 2011 at 11:46 AM, Aneesh Kumar K.V
<aneesh.kumar@linux.vnet.ibm.com> wrote:

>  /**
>  * v9fs_file_open - open a file (or directory)
>  * @inode: inode to be opened
> @@ -89,19 +86,12 @@ int v9fs_file_open(struct inode *inode, struct file *file)
>        }
>
>        file->private_data = fid;
> +#ifdef CONFIG_9P_FSCACHE
>        if ((fid->qid.version) && (v9ses->cache)) {
>                P9_DPRINTK(P9_DEBUG_VFS, "cached");
> -               /* enable cached file options */
> -               if(file->f_op == &v9fs_file_operations)
> -                       file->f_op = &v9fs_cached_file_operations;
> -               else if (file->f_op == &v9fs_file_operations_dotl)
> -                       file->f_op = &v9fs_cached_file_operations_dotl;
> -
> -#ifdef CONFIG_9P_FSCACHE
>                v9fs_cache_inode_set_cookie(inode, file);
> -#endif
>        }
> -
> +#endif
>        return 0;
>  }

Doesn't this conditionalize all cached operations on FSCACHE operation?
What if we want loose cache policy without FSCACHE?

     -eric
--
To unsubscribe from this list: send the line "unsubscribe linux-fsdevel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Re: [V9fs-developer] [RFC PATCH -V2 01/17] fs/9p: set the cached file_operations struct during inode init

[Aneesh Kumar K. V]

On Mon, 7 Feb 2011 09:02:11 -0600, Eric Van Hensbergen <ericvh@gmail.com> wrote:
> On Sat, Feb 5, 2011 at 11:46 AM, Aneesh Kumar K.V
> <aneesh.kumar@linux.vnet.ibm.com> wrote:
> 
> >  /**
> >  * v9fs_file_open - open a file (or directory)
> >  * @inode: inode to be opened
> > @@ -89,19 +86,12 @@ int v9fs_file_open(struct inode *inode, struct file *file)
> >        }
> >
> >        file->private_data = fid;
> > +#ifdef CONFIG_9P_FSCACHE
> >        if ((fid->qid.version) && (v9ses->cache)) {
> >                P9_DPRINTK(P9_DEBUG_VFS, "cached");
> > -               /* enable cached file options */
> > -               if(file->f_op == &v9fs_file_operations)
> > -                       file->f_op = &v9fs_cached_file_operations;
> > -               else if (file->f_op == &v9fs_file_operations_dotl)
> > -                       file->f_op = &v9fs_cached_file_operations_dotl;
> > -
> > -#ifdef CONFIG_9P_FSCACHE
> >                v9fs_cache_inode_set_cookie(inode, file);
> > -#endif
> >        }
> > -
> > +#endif
> >        return 0;
> >  }
> 
> Doesn't this conditionalize all cached operations on FSCACHE operation?
> What if we want loose cache policy without FSCACHE?
> 

we set file_operations is set in inode init now. The only remaining part 
is v9fs_cache_inode_set_cookie which is FSCACHE specific

-aneesh

[RFC PATCH -V2 02/17] fs/9p: set fs cache cookie in create path also

[Aneesh Kumar K.V]

We need to call v9fs_cache_inode_set_cookie in create
path also

Signed-off-by: Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
---
 fs/9p/vfs_file.c       |    4 +---
 fs/9p/vfs_inode.c      |    4 ++++
 fs/9p/vfs_inode_dotl.c |    4 ++++
 3 files changed, 9 insertions(+), 3 deletions(-)

diff --git a/fs/9p/vfs_file.c b/fs/9p/vfs_file.c
index 6a67100..ce1eae4 100644
--- a/fs/9p/vfs_file.c
+++ b/fs/9p/vfs_file.c
@@ -87,10 +87,8 @@ int v9fs_file_open(struct inode *inode, struct file *file)
 
 	file->private_data = fid;
 #ifdef CONFIG_9P_FSCACHE
-	if ((fid->qid.version) && (v9ses->cache)) {
-		P9_DPRINTK(P9_DEBUG_VFS, "cached");
+	if (v9ses->cache)
 		v9fs_cache_inode_set_cookie(inode, file);
-	}
 #endif
 	return 0;
 }
diff --git a/fs/9p/vfs_inode.c b/fs/9p/vfs_inode.c
index 45e7ea4..ce56571 100644
--- a/fs/9p/vfs_inode.c
+++ b/fs/9p/vfs_inode.c
@@ -607,6 +607,10 @@ v9fs_vfs_create(struct inode *dir, struct dentry *dentry, int mode,
 		}
 
 		filp->private_data = fid;
+#ifdef CONFIG_9P_FSCACHE
+		if (v9ses->cache)
+			v9fs_cache_inode_set_cookie(dentry->d_inode, filp);
+#endif
 	} else
 		p9_client_clunk(fid);
 
diff --git a/fs/9p/vfs_inode_dotl.c b/fs/9p/vfs_inode_dotl.c
index fe3ffa9..fbe9572 100644
--- a/fs/9p/vfs_inode_dotl.c
+++ b/fs/9p/vfs_inode_dotl.c
@@ -226,6 +226,10 @@ v9fs_vfs_create_dotl(struct inode *dir, struct dentry *dentry, int omode,
 		return PTR_ERR(filp);
 	}
 	filp->private_data = ofid;
+#ifdef CONFIG_9P_FSCACHE
+	if (v9ses->cache)
+		v9fs_cache_inode_set_cookie(inode, filp);
+#endif
 	return 0;
 
 error:
-- 
1.7.1

[RFC PATCH -V2 03/17] fs/9p: increment inode->i_count in cached mode.

[Aneesh Kumar K.V]

We need to ihold even in cached mode

Signed-off-by: Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
---
 fs/9p/vfs_inode_dotl.c |    7 +------
 1 files changed, 1 insertions(+), 6 deletions(-)

diff --git a/fs/9p/vfs_inode_dotl.c b/fs/9p/vfs_inode_dotl.c
index fbe9572..265f583 100644
--- a/fs/9p/vfs_inode_dotl.c
+++ b/fs/9p/vfs_inode_dotl.c
@@ -636,13 +636,8 @@ v9fs_vfs_link_dotl(struct dentry *old_dentry, struct inode *dir,
 		v9fs_stat2inode_dotl(st, old_dentry->d_inode);
 
 		kfree(st);
-	} else {
-		/* Caching disabled. No need to get upto date stat info.
-		 * This dentry will be released immediately. So, just hold the
-		 * inode
-		 */
-		ihold(old_dentry->d_inode);
 	}
+	ihold(old_dentry->d_inode);
 	d_instantiate(dentry, old_dentry->d_inode);
 
 	return err;
-- 
1.7.1

[RFC PATCH -V2 04/17] fs/9p: [fscache] wait for page write in cached mode

[Aneesh Kumar K.V]

We need to call fscache_wait_on_page_write in launder_page
for fscache

Signed-off-by: Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
---
 fs/9p/cache.c    |   11 +++++++++++
 fs/9p/cache.h    |   16 ++++++++++++++--
 fs/9p/vfs_addr.c |    2 ++
 3 files changed, 27 insertions(+), 2 deletions(-)

diff --git a/fs/9p/cache.c b/fs/9p/cache.c
index 0dbe0d1..610913d 100644
--- a/fs/9p/cache.c
+++ b/fs/9p/cache.c
@@ -461,3 +461,14 @@ void __v9fs_readpage_to_fscache(struct inode *inode, struct page *page)
 	if (ret != 0)
 		v9fs_uncache_page(inode, page);
 }
+
+/*
+ * wait for a page to complete writing to the cache
+ */
+void __v9fs_fscache_wait_on_page_write(struct inode *inode, struct page *page)
+{
+	const struct v9fs_cookie *vcookie = v9fs_inode2cookie(inode);
+	P9_DPRINTK(P9_DEBUG_FSC, "inode %p page %p", inode, page);
+	if (PageFsCache(page))
+		fscache_wait_on_page_write(vcookie->fscache, page);
+}
diff --git a/fs/9p/cache.h b/fs/9p/cache.h
index a94192b..ec16fcd 100644
--- a/fs/9p/cache.h
+++ b/fs/9p/cache.h
@@ -64,8 +64,8 @@ extern int __v9fs_readpages_from_fscache(struct inode *inode,
 					 struct list_head *pages,
 					 unsigned *nr_pages);
 extern void __v9fs_readpage_to_fscache(struct inode *inode, struct page *page);
-
-
+extern void __v9fs_fscache_wait_on_page_write(struct inode *inode,
+					      struct page *page);
 /**
  * v9fs_cache_register - Register v9fs file system with the cache
  */
@@ -131,6 +131,12 @@ static inline void v9fs_vcookie_set_qid(struct inode *inode,
 	spin_unlock(&vcookie->lock);
 }
 
+static inline void v9fs_fscache_wait_on_page_write(struct inode *inode,
+						   struct page *page)
+{
+	return __v9fs_fscache_wait_on_page_write(inode, page);
+}
+
 #else /* CONFIG_9P_FSCACHE */
 
 static inline int v9fs_cache_register(void)
@@ -172,5 +178,11 @@ static inline void v9fs_vcookie_set_qid(struct inode *inode,
 					struct p9_qid *qid)
 {}
 
+static inline void v9fs_fscache_wait_on_page_write(struct inode *inode,
+						   struct page *page)
+{
+	return;
+}
+
 #endif /* CONFIG_9P_FSCACHE */
 #endif /* _9P_CACHE_H */
diff --git a/fs/9p/vfs_addr.c b/fs/9p/vfs_addr.c
index b7f2a8e..637bd70 100644
--- a/fs/9p/vfs_addr.c
+++ b/fs/9p/vfs_addr.c
@@ -151,6 +151,8 @@ static void v9fs_invalidate_page(struct page *page, unsigned long offset)
 
 static int v9fs_launder_page(struct page *page)
 {
+	struct inode *inode = page->mapping->host;
+	v9fs_fscache_wait_on_page_write(inode, page);
 	return 0;
 }
 
-- 
1.7.1

[RFC PATCH -V2 05/17] fs/9p: Add read write helper function

[Aneesh Kumar K.V]

We add read write helper function here which will
be used later by the mmap patch

Signed-off-by: Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
---
 fs/9p/v9fs_vfs.h |    4 ++-
 fs/9p/vfs_file.c |  100 ++++++++++++++++++++++++++++++++---------------------
 2 files changed, 63 insertions(+), 41 deletions(-)

diff --git a/fs/9p/v9fs_vfs.h b/fs/9p/v9fs_vfs.h
index e4d5540..276ad44 100644
--- a/fs/9p/v9fs_vfs.h
+++ b/fs/9p/v9fs_vfs.h
@@ -64,8 +64,10 @@ void v9fs_inode2stat(struct inode *inode, struct p9_wstat *stat);
 int v9fs_uflags2omode(int uflags, int extended);
 
 ssize_t v9fs_file_readn(struct file *, char *, char __user *, u32, u64);
+ssize_t v9fs_fid_readn(struct p9_fid *, char *, char __user *, u32, u64);
 void v9fs_blank_wstat(struct p9_wstat *wstat);
 int v9fs_vfs_setattr_dotl(struct dentry *, struct iattr *);
 int v9fs_file_fsync_dotl(struct file *filp, int datasync);
-
+ssize_t v9fs_file_write_internal(struct inode *, struct p9_fid *, char *,
+				 size_t, loff_t *, int);
 #define P9_LOCK_TIMEOUT (30*HZ)
diff --git a/fs/9p/vfs_file.c b/fs/9p/vfs_file.c
index ce1eae4..ecdd512 100644
--- a/fs/9p/vfs_file.c
+++ b/fs/9p/vfs_file.c
@@ -323,25 +323,22 @@ out_err:
 }
 
 /**
- * v9fs_file_readn - read from a file
- * @filp: file pointer to read
+ * v9fs_fid_readn - read from a fid
+ * @fid: fid to read
  * @data: data buffer to read data into
  * @udata: user data buffer to read data into
  * @count: size of buffer
  * @offset: offset at which to read data
  *
  */
-
 ssize_t
-v9fs_file_readn(struct file *filp, char *data, char __user *udata, u32 count,
+v9fs_fid_readn(struct p9_fid *fid, char *data, char __user *udata, u32 count,
 	       u64 offset)
 {
 	int n, total, size;
-	struct p9_fid *fid = filp->private_data;
 
 	P9_DPRINTK(P9_DEBUG_VFS, "fid %d offset %llu count %d\n", fid->fid,
-					(long long unsigned) offset, count);
-
+		   (long long unsigned) offset, count);
 	n = 0;
 	total = 0;
 	size = fid->iounit ? fid->iounit : fid->clnt->msize - P9_IOHDRSZ;
@@ -367,6 +364,22 @@ v9fs_file_readn(struct file *filp, char *data, char __user *udata, u32 count,
 }
 
 /**
+ * v9fs_file_readn - read from a file
+ * @filp: file pointer to read
+ * @data: data buffer to read data into
+ * @udata: user data buffer to read data into
+ * @count: size of buffer
+ * @offset: offset at which to read data
+ *
+ */
+ssize_t
+v9fs_file_readn(struct file *filp, char *data, char __user *udata, u32 count,
+	       u64 offset)
+{
+	return v9fs_fid_readn(filp->private_data, data, udata, count, offset);
+}
+
+/**
  * v9fs_file_read - read from a file
  * @filp: file pointer to read
  * @udata: user data buffer to read data into
@@ -398,45 +411,20 @@ v9fs_file_read(struct file *filp, char __user *udata, size_t count,
 	return ret;
 }
 
-/**
- * v9fs_file_write - write to a file
- * @filp: file pointer to write
- * @data: data buffer to write data from
- * @count: size of buffer
- * @offset: offset at which to write data
- *
- */
-
-static ssize_t
-v9fs_file_write(struct file *filp, const char __user * data,
-		size_t count, loff_t * offset)
+ssize_t
+v9fs_file_write_internal(struct inode *inode, struct p9_fid *fid, char *data,
+			 size_t count, loff_t *offset, int invalidate)
 {
-	ssize_t retval;
-	size_t total = 0;
 	int n;
-	struct p9_fid *fid;
+	size_t total = 0;
 	struct p9_client *clnt;
-	struct inode *inode = filp->f_path.dentry->d_inode;
 	loff_t origin = *offset;
 	unsigned long pg_start, pg_end;
 
 	P9_DPRINTK(P9_DEBUG_VFS, "data %p count %d offset %x\n", data,
 		(int)count, (int)*offset);
 
-	fid = filp->private_data;
 	clnt = fid->clnt;
-
-	retval = generic_write_checks(filp, &origin, &count, 0);
-	if (retval)
-		goto out;
-
-	retval = -EINVAL;
-	if ((ssize_t) count < 0)
-		goto out;
-	retval = 0;
-	if (!count)
-		goto out;
-
 	do {
 		n = p9_client_write(fid, NULL, data+total, origin+total, count);
 		if (n <= 0)
@@ -445,7 +433,7 @@ v9fs_file_write(struct file *filp, const char __user * data,
 		total += n;
 	} while (count > 0);
 
-	if (total > 0) {
+	if (invalidate && total > 0) {
 		pg_start = origin >> PAGE_CACHE_SHIFT;
 		pg_end = (origin + total - 1) >> PAGE_CACHE_SHIFT;
 		if (inode->i_mapping && inode->i_mapping->nrpages)
@@ -457,9 +445,41 @@ v9fs_file_write(struct file *filp, const char __user * data,
 	}
 
 	if (n < 0)
-		retval = n;
-	else
-		retval = total;
+		return n;
+
+	return total;
+}
+
+/**
+ * v9fs_file_write - write to a file
+ * @filp: file pointer to write
+ * @data: data buffer to write data from
+ * @count: size of buffer
+ * @offset: offset at which to write data
+ *
+ */
+static ssize_t
+v9fs_file_write(struct file *filp, const char __user * data,
+		size_t count, loff_t *offset)
+{
+	ssize_t retval = 0;
+	loff_t origin = *offset;
+
+
+	retval = generic_write_checks(filp, &origin, &count, 0);
+	if (retval)
+		goto out;
+
+	retval = -EINVAL;
+	if ((ssize_t) count < 0)
+		goto out;
+	retval = 0;
+	if (!count)
+		goto out;
+
+	return v9fs_file_write_internal(filp->f_path.dentry->d_inode,
+					filp->private_data,
+					(char *)data, count, offset, 1);
 out:
 	return retval;
 }
-- 
1.7.1

[RFC PATCH -V2 06/17] fs/9p: Add fid to inode in cached mode

[Aneesh Kumar K.V]

The fid attached to inode will be opened O_RDWR mode and is used
for dirty page writeback only.

FIXME!!: Should we make the fid owned by uid = 0

Signed-off-by: Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
---
 fs/9p/vfs_file.c       |   31 +++++++++++++++++++++++++++++++
 fs/9p/vfs_inode.c      |   34 +++++++++++++++++++++++++++++++++-
 fs/9p/vfs_inode_dotl.c |   30 ++++++++++++++++++++++++++++--
 3 files changed, 92 insertions(+), 3 deletions(-)

diff --git a/fs/9p/vfs_file.c b/fs/9p/vfs_file.c
index ecdd512..0f62b26 100644
--- a/fs/9p/vfs_file.c
+++ b/fs/9p/vfs_file.c
@@ -86,11 +86,42 @@ int v9fs_file_open(struct inode *inode, struct file *file)
 	}
 
 	file->private_data = fid;
+	/*
+	 * In cached mode we need to attach a fid to inode. The fid
+	 * attached to inode will only be used to write back the
+	 * dirty pages. We always request for the open fid in read-write
+	 * mode so that a partial page write which result in page
+	 * read can work.
+	 */
+	if (v9ses->cache && !inode->i_private) {
+		/*
+		 * clone a fid and add it to inode->i_private
+		 * we do it during open time instead of
+		 * page dirty time via write_begin/page_mkwrite
+		 * because we want write after unlink usecase
+		 * to work.
+		 */
+		fid = v9fs_fid_clone(file->f_path.dentry);
+		if (IS_ERR(fid)) {
+			err = PTR_ERR(fid);
+			goto out_error;
+		}
+		err = p9_client_open(fid, O_RDWR);
+		if (err < 0) {
+			p9_client_clunk(fid);
+			goto out_error;
+		}
+		inode->i_private = (void *) fid;
+	}
 #ifdef CONFIG_9P_FSCACHE
 	if (v9ses->cache)
 		v9fs_cache_inode_set_cookie(inode, file);
 #endif
 	return 0;
+out_error:
+	p9_client_clunk(file->private_data);
+	file->private_data = 0;
+	return err;
 }
 
 /**
diff --git a/fs/9p/vfs_inode.c b/fs/9p/vfs_inode.c
index ce56571..ecd2c6f 100644
--- a/fs/9p/vfs_inode.c
+++ b/fs/9p/vfs_inode.c
@@ -416,6 +416,11 @@ void v9fs_evict_inode(struct inode *inode)
 #ifdef CONFIG_9P_FSCACHE
 	v9fs_cache_inode_put_cookie(inode);
 #endif
+	/* clunk the fid stashed in inode->i_private */
+	if (inode->i_private) {
+		p9_client_clunk((struct p9_fid *)inode->i_private);
+		inode->i_private = 0;
+	}
 }
 
 struct inode *
@@ -577,7 +582,7 @@ v9fs_vfs_create(struct inode *dir, struct dentry *dentry, int mode,
 	u32 perm;
 	int flags;
 	struct v9fs_session_info *v9ses;
-	struct p9_fid *fid;
+	struct p9_fid *fid, *inode_fid;
 	struct file *filp;
 
 	err = 0;
@@ -600,6 +605,33 @@ v9fs_vfs_create(struct inode *dir, struct dentry *dentry, int mode,
 
 	/* if we are opening a file, assign the open fid to the file */
 	if (nd && nd->flags & LOOKUP_OPEN) {
+		/*
+		 * In cached mode we need to attach a fid to inode. The fid
+		 * attached to inode will only be used to write back the
+		 * dirty pages. We always request for the open fid in read-write
+		 * mode so that a partial page write which result in page
+		 * read can work.
+		 */
+		if (v9ses->cache && !dentry->d_inode->i_private) {
+			/*
+			 * clone a fid and add it to inode->i_private
+			 * we do it during open time instead of
+			 * page dirty time via write_begin/page_mkwrite
+			 * because we want write after unlink usecase
+			 * to work.
+			 */
+			inode_fid = v9fs_fid_clone(dentry);
+			if (IS_ERR(inode_fid)) {
+				err = PTR_ERR(inode_fid);
+				goto error;
+			}
+			err = p9_client_open(inode_fid, O_RDWR);
+			if (err < 0) {
+				p9_client_clunk(inode_fid);
+				goto error;
+			}
+			dentry->d_inode->i_private = (void *) inode_fid;
+		}
 		filp = lookup_instantiate_filp(nd, dentry, generic_file_open);
 		if (IS_ERR(filp)) {
 			err = PTR_ERR(filp);
diff --git a/fs/9p/vfs_inode_dotl.c b/fs/9p/vfs_inode_dotl.c
index 265f583..d5aa986 100644
--- a/fs/9p/vfs_inode_dotl.c
+++ b/fs/9p/vfs_inode_dotl.c
@@ -142,7 +142,7 @@ v9fs_vfs_create_dotl(struct inode *dir, struct dentry *dentry, int omode,
 	mode_t mode;
 	struct v9fs_session_info *v9ses;
 	struct p9_fid *fid = NULL;
-	struct p9_fid *dfid, *ofid;
+	struct p9_fid *dfid, *ofid, *inode_fid;
 	struct file *filp;
 	struct p9_qid qid;
 	struct inode *inode;
@@ -218,7 +218,33 @@ v9fs_vfs_create_dotl(struct inode *dir, struct dentry *dentry, int omode,
 
 	/* Now set the ACL based on the default value */
 	v9fs_set_create_acl(dentry, dacl, pacl);
-
+	/*
+	 * In cached mode we need to attach a fid to inode. The fid
+	 * attached to inode will only be used to write back the
+	 * dirty pages. We always request for the open fid in read-write
+	 * mode so that a partial page write which result in page
+	 * read can work.
+	 */
+	if (v9ses->cache && !inode->i_private) {
+		/*
+		 * clone a fid and add it to inode->i_private
+		 * we do it during open time instead of
+		 * page dirty time via write_begin/page_mkwrite
+		 * because we want write after unlink usecase
+		 * to work.
+		 */
+		inode_fid = v9fs_fid_clone(dentry);
+		if (IS_ERR(inode_fid)) {
+			err = PTR_ERR(inode_fid);
+			goto error;
+		}
+		err = p9_client_open(inode_fid, O_RDWR);
+		if (err < 0) {
+			p9_client_clunk(inode_fid);
+			goto error;
+		}
+		inode->i_private = (void *) inode_fid;
+	}
 	/* Since we are opening a file, assign the open fid to the file */
 	filp = lookup_instantiate_filp(nd, dentry, generic_file_open);
 	if (IS_ERR(filp)) {
-- 
1.7.1

Re: [V9fs-developer] [RFC PATCH -V2 06/17] fs/9p: Add fid to inode in cached mode

[Eric Van Hensbergen]

On Sat, Feb 5, 2011 at 11:46 AM, Aneesh Kumar K.V
<aneesh.kumar@linux.vnet.ibm.com> wrote:
> The fid attached to inode will be opened O_RDWR mode and is used
> for dirty page writeback only.
>
> FIXME!!: Should we make the fid owned by uid = 0
>

Good question, the idea was for the kernel to own the fid while the
cache needed it.  Although
its unlikely to ever be matched to anything else since its an "open"
fid (so it won't be a candidate
for a walk).  It might be more secure to have it owned by some special
constant (a NOUID which could be ~0)
and marks the fid as being owned by the client and no one else without
any sort of implication of authority/ownership.
We could then put a special check in the find fid code to ignore any
fids with NOUID.

      -eric

[RFC PATCH -V2 07/17] fs/9p: Add buffered write support for v9fs.

[Aneesh Kumar K.V]

We can now support writeable mmaps.
Based on the original patch from Badari Pulavarty <pbadari@us.ibm.com>

TODO: O_DIRECT

Signed-off-by: Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
---
 fs/9p/vfs_addr.c |  181 +++++++++++++++++++++++++++++++++++++++++++++++++-----
 fs/9p/vfs_file.c |   54 +++++++++++++++-
 2 files changed, 216 insertions(+), 19 deletions(-)

diff --git a/fs/9p/vfs_addr.c b/fs/9p/vfs_addr.c
index 637bd70..792c68a 100644
--- a/fs/9p/vfs_addr.c
+++ b/fs/9p/vfs_addr.c
@@ -39,16 +39,16 @@
 #include "v9fs.h"
 #include "v9fs_vfs.h"
 #include "cache.h"
+#include "fid.h"
 
 /**
- * v9fs_vfs_readpage - read an entire page in from 9P
+ * v9fs_fid_readpage - read an entire page in from 9P
  *
- * @filp: file being read
+ * @fid: fid being read
  * @page: structure to page
  *
  */
-
-static int v9fs_vfs_readpage(struct file *filp, struct page *page)
+static int v9fs_fid_readpage(struct p9_fid *fid, struct page *page)
 {
 	int retval;
 	loff_t offset;
@@ -67,7 +67,7 @@ static int v9fs_vfs_readpage(struct file *filp, struct page *page)
 	buffer = kmap(page);
 	offset = page_offset(page);
 
-	retval = v9fs_file_readn(filp, buffer, NULL, PAGE_CACHE_SIZE, offset);
+	retval = v9fs_fid_readn(fid, buffer, NULL, PAGE_CACHE_SIZE, offset);
 	if (retval < 0) {
 		v9fs_uncache_page(inode, page);
 		goto done;
@@ -87,6 +87,19 @@ done:
 }
 
 /**
+ * v9fs_vfs_readpage - read an entire page in from 9P
+ *
+ * @filp: file being read
+ * @page: structure to page
+ *
+ */
+
+static int v9fs_vfs_readpage(struct file *filp, struct page *page)
+{
+	return v9fs_fid_readpage(filp->private_data, page);
+}
+
+/**
  * v9fs_vfs_readpages - read a set of pages from 9P
  *
  * @filp: file being read
@@ -124,7 +137,6 @@ static int v9fs_release_page(struct page *page, gfp_t gfp)
 {
 	if (PagePrivate(page))
 		return 0;
-
 	return v9fs_fscache_release_page(page, gfp);
 }
 
@@ -137,22 +149,86 @@ static int v9fs_release_page(struct page *page, gfp_t gfp)
 
 static void v9fs_invalidate_page(struct page *page, unsigned long offset)
 {
+	/*
+	 * If called with zero offset, we should release
+	 * the private state assocated with the page
+	 */
 	if (offset == 0)
 		v9fs_fscache_invalidate_page(page);
 }
 
+static int v9fs_vfs_writepage_locked(struct page *page)
+{
+	char *buffer;
+	int retval, len;
+	loff_t offset, size;
+	mm_segment_t old_fs;
+	struct inode *inode = page->mapping->host;
+
+	size = i_size_read(inode);
+	if (page->index == size >> PAGE_CACHE_SHIFT)
+		len = size & ~PAGE_CACHE_MASK;
+	else
+		len = PAGE_CACHE_SIZE;
+
+	set_page_writeback(page);
+
+	buffer = kmap(page);
+	offset = page_offset(page);
+
+	old_fs = get_fs();
+	set_fs(get_ds());
+	/* We should have i_private always set */
+	BUG_ON(!inode->i_private);
+
+	retval = v9fs_file_write_internal(inode,
+					  (struct p9_fid *)inode->i_private,
+					  buffer, len, &offset, 0);
+	if (retval > 0)
+		retval = 0;
+
+	set_fs(old_fs);
+	kunmap(page);
+	end_page_writeback(page);
+	return retval;
+}
+
+static int v9fs_vfs_writepage(struct page *page, struct writeback_control *wbc)
+{
+	int retval;
+
+	retval = v9fs_vfs_writepage_locked(page);
+	if (retval < 0) {
+		if (retval == -EAGAIN) {
+			redirty_page_for_writepage(wbc, page);
+			retval = 0;
+		} else {
+			SetPageError(page);
+			mapping_set_error(page->mapping, retval);
+		}
+	} else
+		retval = 0;
+
+	unlock_page(page);
+	return retval;
+}
+
 /**
  * v9fs_launder_page - Writeback a dirty page
- * Since the writes go directly to the server, we simply return a 0
- * here to indicate success.
- *
  * Returns 0 on success.
  */
 
 static int v9fs_launder_page(struct page *page)
 {
+	int retval;
 	struct inode *inode = page->mapping->host;
+
 	v9fs_fscache_wait_on_page_write(inode, page);
+	if (clear_page_dirty_for_io(page)) {
+		retval = v9fs_vfs_writepage_locked(page);
+		if (retval)
+			return retval;
+	}
 	return 0;
 }
 
@@ -178,6 +254,11 @@ static int v9fs_launder_page(struct page *page)
 ssize_t v9fs_direct_IO(int rw, struct kiocb *iocb, const struct iovec *iov,
 		loff_t pos, unsigned long nr_segs)
 {
+	/*
+	 * FIXME
+	 * Now that we do caching with cache mode enabled, We need
+	 * to support direct IO
+	 */
 	P9_DPRINTK(P9_DEBUG_VFS, "v9fs_direct_IO: v9fs_direct_IO (%s) "
 			"off/no(%lld/%lu) EINVAL\n",
 			iocb->ki_filp->f_path.dentry->d_name.name,
@@ -185,11 +266,81 @@ ssize_t v9fs_direct_IO(int rw, struct kiocb *iocb, const struct iovec *iov,
 
 	return -EINVAL;
 }
+
+static int v9fs_write_begin(struct file *filp, struct address_space *mapping,
+			    loff_t pos, unsigned len, unsigned flags,
+			    struct page **pagep, void **fsdata)
+{
+	int retval = 0;
+	struct page *page;
+	pgoff_t index = pos >> PAGE_CACHE_SHIFT;
+	struct inode *inode = mapping->host;
+
+start:
+	page = grab_cache_page_write_begin(mapping, index, flags);
+	if (!page) {
+		retval = -ENOMEM;
+		goto out;
+	}
+	BUG_ON(!inode->i_private);
+	if (PageUptodate(page))
+		goto out;
+
+	if (len == PAGE_CACHE_SIZE)
+		goto out;
+
+	retval = v9fs_fid_readpage(inode->i_private, page);
+	page_cache_release(page);
+	if (!retval)
+		goto start;
+out:
+	*pagep = page;
+	return retval;
+}
+
+static int v9fs_write_end(struct file *filp, struct address_space *mapping,
+			  loff_t pos, unsigned len, unsigned copied,
+			  struct page *page, void *fsdata)
+{
+	loff_t last_pos = pos + copied;
+	struct inode *inode = page->mapping->host;
+
+	if (unlikely(copied < len)) {
+		/*
+		 * zero out the rest of the area
+		 */
+		unsigned from = pos & (PAGE_CACHE_SIZE - 1);
+
+		zero_user(page, from + copied, len - copied);
+		flush_dcache_page(page);
+	}
+
+	if (!PageUptodate(page))
+		SetPageUptodate(page);
+	/*
+	 * No need to use i_size_read() here, the i_size
+	 * cannot change under us because we hold the i_mutex.
+	 */
+	if (last_pos > inode->i_size)
+		i_size_write(inode, last_pos);
+
+	set_page_dirty(page);
+	unlock_page(page);
+	page_cache_release(page);
+
+	return copied;
+}
+
+
 const struct address_space_operations v9fs_addr_operations = {
-      .readpage = v9fs_vfs_readpage,
-      .readpages = v9fs_vfs_readpages,
-      .releasepage = v9fs_release_page,
-      .invalidatepage = v9fs_invalidate_page,
-      .launder_page = v9fs_launder_page,
-      .direct_IO = v9fs_direct_IO,
+	.readpage = v9fs_vfs_readpage,
+	.readpages = v9fs_vfs_readpages,
+	.set_page_dirty = __set_page_dirty_nobuffers,
+	.writepage = v9fs_vfs_writepage,
+	.write_begin = v9fs_write_begin,
+	.write_end = v9fs_write_end,
+	.releasepage = v9fs_release_page,
+	.invalidatepage = v9fs_invalidate_page,
+	.launder_page = v9fs_launder_page,
+	.direct_IO = v9fs_direct_IO,
 };
diff --git a/fs/9p/vfs_file.c b/fs/9p/vfs_file.c
index 0f62b26..6e38947 100644
--- a/fs/9p/vfs_file.c
+++ b/fs/9p/vfs_file.c
@@ -44,6 +44,8 @@
 #include "fid.h"
 #include "cache.h"
 
+static const struct vm_operations_struct v9fs_file_vm_ops;
+
 /**
  * v9fs_file_open - open a file (or directory)
  * @inode: inode to be opened
@@ -515,6 +517,7 @@ out:
 	return retval;
 }
 
+
 static int v9fs_file_fsync(struct file *filp, int datasync)
 {
 	struct p9_fid *fid;
@@ -544,28 +547,71 @@ int v9fs_file_fsync_dotl(struct file *filp, int datasync)
 	return retval;
 }
 
+static int
+v9fs_file_mmap(struct file *file, struct vm_area_struct *vma)
+{
+	int retval;
+
+	retval = generic_file_mmap(file, vma);
+	if (!retval)
+		vma->vm_ops = &v9fs_file_vm_ops;
+
+	return retval;
+}
+
+static int
+v9fs_vm_page_mkwrite(struct vm_area_struct *vma, struct vm_fault *vmf)
+{
+	struct page *page = vmf->page;
+	struct file *filp = vma->vm_file;
+	struct inode *inode = filp->f_path.dentry->d_inode;
+
+
+	P9_DPRINTK(P9_DEBUG_VFS, "page %p fid %lx\n",
+		   page, (unsigned long)filp->private_data);
+
+	/* make sure the cache has finished storing the page */
+	v9fs_fscache_wait_on_page_write(inode, page);
+	BUG_ON(!inode->i_private);
+	lock_page(page);
+	if (page->mapping != inode->i_mapping)
+		goto out_unlock;
+
+	return VM_FAULT_LOCKED;
+out_unlock:
+	unlock_page(page);
+	return VM_FAULT_NOPAGE;
+}
+
+static const struct vm_operations_struct v9fs_file_vm_ops = {
+	.fault = filemap_fault,
+	.page_mkwrite = v9fs_vm_page_mkwrite,
+};
+
 const struct file_operations v9fs_cached_file_operations = {
 	.llseek = generic_file_llseek,
 	.read = do_sync_read,
+	.write = do_sync_write,
 	.aio_read = generic_file_aio_read,
-	.write = v9fs_file_write,
+	.aio_write = generic_file_aio_write,
 	.open = v9fs_file_open,
 	.release = v9fs_dir_release,
 	.lock = v9fs_file_lock,
-	.mmap = generic_file_readonly_mmap,
+	.mmap = v9fs_file_mmap,
 	.fsync = v9fs_file_fsync,
 };
 
 const struct file_operations v9fs_cached_file_operations_dotl = {
 	.llseek = generic_file_llseek,
 	.read = do_sync_read,
+	.write = do_sync_write,
 	.aio_read = generic_file_aio_read,
-	.write = v9fs_file_write,
+	.aio_write = generic_file_aio_write,
 	.open = v9fs_file_open,
 	.release = v9fs_dir_release,
 	.lock = v9fs_file_lock_dotl,
 	.flock = v9fs_file_flock_dotl,
-	.mmap = generic_file_readonly_mmap,
+	.mmap = v9fs_file_mmap,
 	.fsync = v9fs_file_fsync_dotl,
 };
 
-- 
1.7.1

[RFC PATCH -V2 08/17] fs/9p: Clarify cached dentry delete operation

[Aneesh Kumar K.V]

Update the comment to indicate that we don't want to cache
negative dentries.

Signed-off-by: Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
---
 fs/9p/vfs_dentry.c |   13 ++++---------
 1 files changed, 4 insertions(+), 9 deletions(-)

diff --git a/fs/9p/vfs_dentry.c b/fs/9p/vfs_dentry.c
index 233b7d4..a4ae4be 100644
--- a/fs/9p/vfs_dentry.c
+++ b/fs/9p/vfs_dentry.c
@@ -63,20 +63,15 @@ static int v9fs_dentry_delete(const struct dentry *dentry)
  * v9fs_cached_dentry_delete - called when dentry refcount equals 0
  * @dentry:  dentry in question
  *
- * Only return 1 if our inode is invalid.  Only non-synthetic files
- * (ones without mtime == 0) should be calling this function.
- *
  */
-
 static int v9fs_cached_dentry_delete(const struct dentry *dentry)
 {
-	struct inode *inode = dentry->d_inode;
-	P9_DPRINTK(P9_DEBUG_VFS, " dentry: %s (%p)\n", dentry->d_name.name,
-									dentry);
+	P9_DPRINTK(P9_DEBUG_VFS, " dentry: %s (%p)\n",
+		   dentry->d_name.name, dentry);
 
-	if(!inode)
+	/* Don't cache negative dentries */
+	if (!dentry->d_inode)
 		return 1;
-
 	return 0;
 }
 
-- 
1.7.1

[RFC PATCH -V2 09/17] fs/9p: Mark file system with MS_SYNCHRONOUS only if it is not cached mode

[Aneesh Kumar K.V]

We should not mark file system synchronous if mounted cache=* option

Signed-off-by: Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
---
 fs/9p/vfs_super.c |    5 +++--
 1 files changed, 3 insertions(+), 2 deletions(-)

diff --git a/fs/9p/vfs_super.c b/fs/9p/vfs_super.c
index dbaabe3..76e11e8 100644
--- a/fs/9p/vfs_super.c
+++ b/fs/9p/vfs_super.c
@@ -87,8 +87,9 @@ v9fs_fill_super(struct super_block *sb, struct v9fs_session_info *v9ses,
 		sb->s_op = &v9fs_super_ops;
 	sb->s_bdi = &v9ses->bdi;
 
-	sb->s_flags = flags | MS_ACTIVE | MS_SYNCHRONOUS | MS_DIRSYNC |
-	    MS_NOATIME;
+	sb->s_flags = flags | MS_ACTIVE | MS_DIRSYNC | MS_NOATIME;
+	if (!v9ses->cache)
+		sb->s_flags |= MS_SYNCHRONOUS;
 
 #ifdef CONFIG_9P_FS_POSIX_ACL
 	if ((v9ses->flags & V9FS_ACCESS_MASK) == V9FS_ACCESS_CLIENT)
-- 
1.7.1

[RFC PATCH -V2 10/17] net/9p: Implement syncfs 9P operation

[Aneesh Kumar K.V]

Signed-off-by: Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
---
 include/net/9p/9p.h     |    2 ++
 include/net/9p/client.h |    1 +
 net/9p/client.c         |   21 +++++++++++++++++++++
 3 files changed, 24 insertions(+), 0 deletions(-)

diff --git a/include/net/9p/9p.h b/include/net/9p/9p.h
index 071fd7a..73dc432 100644
--- a/include/net/9p/9p.h
+++ b/include/net/9p/9p.h
@@ -139,6 +139,8 @@ do { \
  */
 
 enum p9_msg_t {
+	P9_TSYNCFS = 0,
+	P9_RSYNCFS,
 	P9_TLERROR = 6,
 	P9_RLERROR,
 	P9_TSTATFS = 8,
diff --git a/include/net/9p/client.h b/include/net/9p/client.h
index 83ba6a4..0a30977 100644
--- a/include/net/9p/client.h
+++ b/include/net/9p/client.h
@@ -230,6 +230,7 @@ int p9_client_create_dotl(struct p9_fid *ofid, char *name, u32 flags, u32 mode,
 		gid_t gid, struct p9_qid *qid);
 int p9_client_clunk(struct p9_fid *fid);
 int p9_client_fsync(struct p9_fid *fid, int datasync);
+int p9_client_sync_fs(struct p9_fid *fid);
 int p9_client_remove(struct p9_fid *fid);
 int p9_client_read(struct p9_fid *fid, char *data, char __user *udata,
 							u64 offset, u32 count);
diff --git a/net/9p/client.c b/net/9p/client.c
index a848bca..a05673b 100644
--- a/net/9p/client.c
+++ b/net/9p/client.c
@@ -1191,6 +1191,27 @@ error:
 }
 EXPORT_SYMBOL(p9_client_fsync);
 
+int p9_client_sync_fs(struct p9_fid *fid)
+{
+	int err = 0;
+	struct p9_req_t *req;
+	struct p9_client *clnt;
+
+	P9_DPRINTK(P9_DEBUG_9P, ">>> TSYNC_FS fid %d\n", fid->fid);
+
+	clnt = fid->clnt;
+	req = p9_client_rpc(clnt, P9_TSYNCFS, "d", fid->fid);
+	if (IS_ERR(req)) {
+		err = PTR_ERR(req);
+		goto error;
+	}
+	P9_DPRINTK(P9_DEBUG_9P, "<<< RSYNCFS fid %d\n", fid->fid);
+	p9_free_req(clnt, req);
+error:
+	return err;
+}
+EXPORT_SYMBOL(p9_client_sync_fs);
+
 int p9_client_clunk(struct p9_fid *fid)
 {
 	int err;
-- 
1.7.1

[RFC PATCH -V2 11/17] fs/9p: Implement syncfs call back for 9Pfs

[Aneesh Kumar K.V]

FIXME!! what about dotu ?

Signed-off-by: Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
---
 fs/9p/v9fs.h      |    1 +
 fs/9p/vfs_super.c |   19 +++++++++++++++++--
 2 files changed, 18 insertions(+), 2 deletions(-)

diff --git a/fs/9p/v9fs.h b/fs/9p/v9fs.h
index c4b5d88..41f29b9 100644
--- a/fs/9p/v9fs.h
+++ b/fs/9p/v9fs.h
@@ -109,6 +109,7 @@ struct v9fs_session_info {
 	struct list_head slist; /* list of sessions registered with v9fs */
 	struct backing_dev_info bdi;
 	struct rw_semaphore rename_sem;
+	struct p9_fid *root_fid; /* Used for file system sync */
 };
 
 struct p9_fid *v9fs_session_init(struct v9fs_session_info *, const char *,
diff --git a/fs/9p/vfs_super.c b/fs/9p/vfs_super.c
index 76e11e8..fde5fe2 100644
--- a/fs/9p/vfs_super.c
+++ b/fs/9p/vfs_super.c
@@ -152,7 +152,6 @@ static struct dentry *v9fs_mount(struct file_system_type *fs_type, int flags,
 		retval = PTR_ERR(inode);
 		goto release_sb;
 	}
-
 	root = d_alloc_root(inode);
 	if (!root) {
 		iput(inode);
@@ -188,6 +187,13 @@ static struct dentry *v9fs_mount(struct file_system_type *fs_type, int flags,
 	if (retval)
 		goto release_sb;
 	v9fs_fid_add(root, fid);
+	/*
+	 * Add the root fid to session info. This is used
+	 * for file system sync. We want a cloned fid here
+	 * so that we can do a sync_filesystem after a
+	 * shrink_dcache_for_umount
+	 */
+	v9ses->root_fid = v9fs_fid_clone(root);
 
 	P9_DPRINTK(P9_DEBUG_VFS, " simple set mount, return 0\n");
 	return dget(sb->s_root);
@@ -224,7 +230,7 @@ static void v9fs_kill_super(struct super_block *s)
 	P9_DPRINTK(P9_DEBUG_VFS, " %p\n", s);
 
 	kill_anon_super(s);
-
+	p9_client_clunk(v9ses->root_fid);
 	v9fs_session_cancel(v9ses);
 	v9fs_session_close(v9ses);
 	kfree(v9ses);
@@ -277,6 +283,14 @@ done:
 	return res;
 }
 
+static int v9fs_sync_fs(struct super_block *sb, int wait)
+{
+	struct v9fs_session_info *v9ses = sb->s_fs_info;
+
+	P9_DPRINTK(P9_DEBUG_VFS, "v9fs_sync_fs: super_block %p\n", sb);
+	return p9_client_sync_fs(v9ses->root_fid);
+}
+
 static const struct super_operations v9fs_super_ops = {
 #ifdef CONFIG_9P_FSCACHE
 	.alloc_inode = v9fs_alloc_inode,
@@ -293,6 +307,7 @@ static const struct super_operations v9fs_super_ops_dotl = {
 	.alloc_inode = v9fs_alloc_inode,
 	.destroy_inode = v9fs_destroy_inode,
 #endif
+	.sync_fs = v9fs_sync_fs,
 	.statfs = v9fs_statfs,
 	.evict_inode = v9fs_evict_inode,
 	.show_options = generic_show_options,
-- 
1.7.1

[RFC PATCH -V2 12/17] fs/9p: We need not writeback dirty pages during close

[Aneesh Kumar K.V]

Signed-off-by: Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
---
 fs/9p/vfs_dir.c |    1 -
 1 files changed, 0 insertions(+), 1 deletions(-)

diff --git a/fs/9p/vfs_dir.c b/fs/9p/vfs_dir.c
index b84ebe8..9c2bdda 100644
--- a/fs/9p/vfs_dir.c
+++ b/fs/9p/vfs_dir.c
@@ -295,7 +295,6 @@ int v9fs_dir_release(struct inode *inode, struct file *filp)
 	P9_DPRINTK(P9_DEBUG_VFS,
 			"v9fs_dir_release: inode: %p filp: %p fid: %d\n",
 			inode, filp, fid ? fid->fid : -1);
-	filemap_write_and_wait(inode->i_mapping);
 	if (fid)
 		p9_client_clunk(fid);
 	return 0;
-- 
1.7.1

[RFC PATCH -V2 13/17] fs/9p: Add inode hashing

[Aneesh Kumar K.V]

We didn't add the inode to inode hash in 9p. We need to do that
to get sync to work, otherwise __mark_inode_dirty will not
add the inode to super block's dirty list.

Signed-off-by: Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
---
 fs/9p/acl.c            |    2 +-
 fs/9p/v9fs_vfs.h       |    2 +
 fs/9p/vfs_inode.c      |  121 +++++++++++++++++++++++++++++------------------
 fs/9p/vfs_inode_dotl.c |   69 ++++++++++++++++++---------
 4 files changed, 124 insertions(+), 70 deletions(-)

diff --git a/fs/9p/acl.c b/fs/9p/acl.c
index 02a2cf6..5b314c7 100644
--- a/fs/9p/acl.c
+++ b/fs/9p/acl.c
@@ -21,8 +21,8 @@
 #include <linux/posix_acl_xattr.h>
 #include "xattr.h"
 #include "acl.h"
-#include "v9fs_vfs.h"
 #include "v9fs.h"
+#include "v9fs_vfs.h"
 
 static struct posix_acl *__v9fs_get_acl(struct p9_fid *fid, char *name)
 {
diff --git a/fs/9p/v9fs_vfs.h b/fs/9p/v9fs_vfs.h
index 276ad44..ce9ba20 100644
--- a/fs/9p/v9fs_vfs.h
+++ b/fs/9p/v9fs_vfs.h
@@ -54,6 +54,8 @@ void v9fs_destroy_inode(struct inode *inode);
 #endif
 
 struct inode *v9fs_get_inode(struct super_block *sb, int mode);
+int v9fs_init_inode(struct v9fs_session_info *v9ses,
+		    struct inode *inode, int mode);
 void v9fs_evict_inode(struct inode *inode);
 ino_t v9fs_qid2ino(struct p9_qid *qid);
 void v9fs_stat2inode(struct p9_wstat *, struct inode *, struct super_block *);
diff --git a/fs/9p/vfs_inode.c b/fs/9p/vfs_inode.c
index ecd2c6f..472d447 100644
--- a/fs/9p/vfs_inode.c
+++ b/fs/9p/vfs_inode.c
@@ -243,26 +243,10 @@ void v9fs_destroy_inode(struct inode *inode)
 }
 #endif
 
-/**
- * v9fs_get_inode - helper function to setup an inode
- * @sb: superblock
- * @mode: mode to setup inode with
- *
- */
-
-struct inode *v9fs_get_inode(struct super_block *sb, int mode)
+int v9fs_init_inode(struct v9fs_session_info *v9ses,
+		    struct inode *inode, int mode)
 {
-	int err;
-	struct inode *inode;
-	struct v9fs_session_info *v9ses = sb->s_fs_info;
-
-	P9_DPRINTK(P9_DEBUG_VFS, "super block: %p mode: %o\n", sb, mode);
-
-	inode = new_inode(sb);
-	if (!inode) {
-		P9_EPRINTK(KERN_WARNING, "Problem allocating inode\n");
-		return ERR_PTR(-ENOMEM);
-	}
+	int err = 0;
 
 	inode_init_owner(inode, NULL, mode);
 	inode->i_blocks = 0;
@@ -305,7 +289,6 @@ struct inode *v9fs_get_inode(struct super_block *sb, int mode)
 		}
 
 		break;
-
 	case S_IFLNK:
 		if (!v9fs_proto_dotu(v9ses) && !v9fs_proto_dotl(v9ses)) {
 			P9_DPRINTK(P9_DEBUG_ERROR, "extended modes used with "
@@ -341,12 +324,37 @@ struct inode *v9fs_get_inode(struct super_block *sb, int mode)
 		err = -EINVAL;
 		goto error;
 	}
+error:
+	return err;
 
-	return inode;
+}
 
-error:
-	iput(inode);
-	return ERR_PTR(err);
+/**
+ * v9fs_get_inode - helper function to setup an inode
+ * @sb: superblock
+ * @mode: mode to setup inode with
+ *
+ */
+
+struct inode *v9fs_get_inode(struct super_block *sb, int mode)
+{
+	int err;
+	struct inode *inode;
+	struct v9fs_session_info *v9ses = sb->s_fs_info;
+
+	P9_DPRINTK(P9_DEBUG_VFS, "super block: %p mode: %o\n", sb, mode);
+
+	inode = new_inode(sb);
+	if (!inode) {
+		P9_EPRINTK(KERN_WARNING, "Problem allocating inode\n");
+		return ERR_PTR(-ENOMEM);
+	}
+	err = v9fs_init_inode(v9ses, inode, mode);
+	if (err) {
+		iput(inode);
+		return ERR_PTR(err);
+	}
+	return inode;
 }
 
 /*
@@ -423,39 +431,60 @@ void v9fs_evict_inode(struct inode *inode)
 	}
 }
 
-struct inode *
-v9fs_inode(struct v9fs_session_info *v9ses, struct p9_fid *fid,
-	struct super_block *sb)
+static struct inode *v9fs_qid_iget(struct super_block *sb,
+				   struct p9_qid *qid,
+				   struct p9_wstat *st)
 {
-	int err, umode;
-	struct inode *ret = NULL;
-	struct p9_wstat *st;
-
-	st = p9_client_stat(fid);
-	if (IS_ERR(st))
-		return ERR_CAST(st);
+	int retval, umode;
+	unsigned long i_ino;
+	struct inode *inode;
+	struct v9fs_session_info *v9ses = sb->s_fs_info;
 
+	i_ino = v9fs_qid2ino(qid);
+	inode = iget_locked(sb, i_ino);
+	if (!inode)
+		return ERR_PTR(-ENOMEM);
+	if (!(inode->i_state & I_NEW))
+		return inode;
+	/*
+	 * initialize the inode with the stat info
+	 * FIXME!! we may need support for stale inodes
+	 * later.
+	 */
 	umode = p9mode2unixmode(v9ses, st->mode);
-	ret = v9fs_get_inode(sb, umode);
-	if (IS_ERR(ret)) {
-		err = PTR_ERR(ret);
+	retval = v9fs_init_inode(v9ses, inode, umode);
+	if (retval)
 		goto error;
-	}
-
-	v9fs_stat2inode(st, ret, sb);
-	ret->i_ino = v9fs_qid2ino(&st->qid);
 
+	v9fs_stat2inode(st, inode, sb);
 #ifdef CONFIG_9P_FSCACHE
 	v9fs_vcookie_set_qid(ret, &st->qid);
-	v9fs_cache_inode_get_cookie(ret);
+	v9fs_cache_inode_get_cookie(inode);
 #endif
-	p9stat_free(st);
-	kfree(st);
-	return ret;
+	unlock_new_inode(inode);
+	return inode;
 error:
+	unlock_new_inode(inode);
+	iput(inode);
+	return ERR_PTR(retval);
+
+}
+
+struct inode *
+v9fs_inode(struct v9fs_session_info *v9ses, struct p9_fid *fid,
+	   struct super_block *sb)
+{
+	struct p9_wstat *st;
+	struct inode *inode = NULL;
+
+	st = p9_client_stat(fid);
+	if (IS_ERR(st))
+		return ERR_CAST(st);
+
+	inode = v9fs_qid_iget(sb, &st->qid, st);
 	p9stat_free(st);
 	kfree(st);
-	return ERR_PTR(err);
+	return inode;
 }
 
 /**
diff --git a/fs/9p/vfs_inode_dotl.c b/fs/9p/vfs_inode_dotl.c
index d5aa986..a11e04a 100644
--- a/fs/9p/vfs_inode_dotl.c
+++ b/fs/9p/vfs_inode_dotl.c
@@ -86,40 +86,63 @@ static struct dentry *v9fs_dentry_from_dir_inode(struct inode *inode)
 	return dentry;
 }
 
+static struct inode *v9fs_qid_iget_dotl(struct super_block *sb,
+					struct p9_qid *qid,
+					struct p9_fid *fid,
+					struct p9_stat_dotl *st)
+{
+	int retval;
+	unsigned long i_ino;
+	struct inode *inode;
+	struct v9fs_session_info *v9ses = sb->s_fs_info;
+
+	i_ino = v9fs_qid2ino(qid);
+	inode = iget_locked(sb, i_ino);
+	if (!inode)
+		return ERR_PTR(-ENOMEM);
+	if (!(inode->i_state & I_NEW))
+		return inode;
+	/*
+	 * initialize the inode with the stat info
+	 * FIXME!! we may need support for stale inodes
+	 * later.
+	 */
+	retval = v9fs_init_inode(v9ses, inode, st->st_mode);
+	if (retval)
+		goto error;
+
+	v9fs_stat2inode_dotl(st, inode);
+#ifdef CONFIG_9P_FSCACHE
+	v9fs_vcookie_set_qid(inode, &st->qid);
+	v9fs_cache_inode_get_cookie(inode);
+#endif
+	retval = v9fs_get_acl(inode, fid);
+	if (retval)
+		goto error;
+
+	unlock_new_inode(inode);
+	return inode;
+error:
+	unlock_new_inode(inode);
+	iput(inode);
+	return ERR_PTR(retval);
+
+}
+
 struct inode *
 v9fs_inode_dotl(struct v9fs_session_info *v9ses, struct p9_fid *fid,
 	struct super_block *sb)
 {
-	struct inode *ret = NULL;
-	int err;
 	struct p9_stat_dotl *st;
+	struct inode *inode = NULL;
 
 	st = p9_client_getattr_dotl(fid, P9_STATS_BASIC);
 	if (IS_ERR(st))
 		return ERR_CAST(st);
 
-	ret = v9fs_get_inode(sb, st->st_mode);
-	if (IS_ERR(ret)) {
-		err = PTR_ERR(ret);
-		goto error;
-	}
-
-	v9fs_stat2inode_dotl(st, ret);
-	ret->i_ino = v9fs_qid2ino(&st->qid);
-#ifdef CONFIG_9P_FSCACHE
-	v9fs_vcookie_set_qid(ret, &st->qid);
-	v9fs_cache_inode_get_cookie(ret);
-#endif
-	err = v9fs_get_acl(ret, fid);
-	if (err) {
-		iput(ret);
-		goto error;
-	}
-	kfree(st);
-	return ret;
-error:
+	inode = v9fs_qid_iget_dotl(sb, &st->qid, fid, st);
 	kfree(st);
-	return ERR_PTR(err);
+	return inode;
 }
 
 /**
-- 
1.7.1

[RFC PATCH -V2 14/17] fs/9p: Don't set stat.st_blocks based on nrpages

[Aneesh Kumar K.V]

simple_getattr does set stat.st_blocks to a value
derived from nrpages. That is not correct with 9p

Signed-off-by: Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
---
 fs/9p/vfs_inode.c      |    7 ++++---
 fs/9p/vfs_inode_dotl.c |    7 ++++---
 2 files changed, 8 insertions(+), 6 deletions(-)

diff --git a/fs/9p/vfs_inode.c b/fs/9p/vfs_inode.c
index 472d447..8cf88eb 100644
--- a/fs/9p/vfs_inode.c
+++ b/fs/9p/vfs_inode.c
@@ -902,9 +902,10 @@ v9fs_vfs_getattr(struct vfsmount *mnt, struct dentry *dentry,
 	P9_DPRINTK(P9_DEBUG_VFS, "dentry: %p\n", dentry);
 	err = -EPERM;
 	v9ses = v9fs_inode2v9ses(dentry->d_inode);
-	if (v9ses->cache == CACHE_LOOSE || v9ses->cache == CACHE_FSCACHE)
-		return simple_getattr(mnt, dentry, stat);
-
+	if (v9ses->cache == CACHE_LOOSE || v9ses->cache == CACHE_FSCACHE) {
+		generic_fillattr(dentry->d_inode, stat);
+		return 0;
+	}
 	fid = v9fs_fid_lookup(dentry);
 	if (IS_ERR(fid))
 		return PTR_ERR(fid);
diff --git a/fs/9p/vfs_inode_dotl.c b/fs/9p/vfs_inode_dotl.c
index a11e04a..4f27fee 100644
--- a/fs/9p/vfs_inode_dotl.c
+++ b/fs/9p/vfs_inode_dotl.c
@@ -399,9 +399,10 @@ v9fs_vfs_getattr_dotl(struct vfsmount *mnt, struct dentry *dentry,
 	P9_DPRINTK(P9_DEBUG_VFS, "dentry: %p\n", dentry);
 	err = -EPERM;
 	v9ses = v9fs_inode2v9ses(dentry->d_inode);
-	if (v9ses->cache == CACHE_LOOSE || v9ses->cache == CACHE_FSCACHE)
-		return simple_getattr(mnt, dentry, stat);
-
+	if (v9ses->cache == CACHE_LOOSE || v9ses->cache == CACHE_FSCACHE) {
+		generic_fillattr(dentry->d_inode, stat);
+		return 0;
+	}
 	fid = v9fs_fid_lookup(dentry);
 	if (IS_ERR(fid))
 		return PTR_ERR(fid);
-- 
1.7.1

[RFC PATCH -V2 15/17] fs/9p: Add v9fs_inode

[Aneesh Kumar K.V]

Switch to the fscache code to v9fs_inode. We will later use
v9fs_inode in cache=loose mode to track the inode cache
validity timeout. Ie if we find an inode in cache older
that a specific jiffie range we will consider it stale

Signed-off-by: Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
---
 fs/9p/cache.c          |  197 +++++++++++++++++-------------------------------
 fs/9p/cache.h          |   54 ++------------
 fs/9p/v9fs.c           |   58 ++++++++++++++
 fs/9p/v9fs.h           |   37 ++++++---
 fs/9p/v9fs_vfs.h       |    4 +-
 fs/9p/vfs_inode.c      |   36 ++++-----
 fs/9p/vfs_inode_dotl.c |   14 ++--
 fs/9p/vfs_super.c      |    4 -
 8 files changed, 184 insertions(+), 220 deletions(-)

diff --git a/fs/9p/cache.c b/fs/9p/cache.c
index 610913d..5b335c5 100644
--- a/fs/9p/cache.c
+++ b/fs/9p/cache.c
@@ -33,67 +33,11 @@
 
 #define CACHETAG_LEN  11
 
-struct kmem_cache *vcookie_cache;
-
 struct fscache_netfs v9fs_cache_netfs = {
 	.name 		= "9p",
 	.version 	= 0,
 };
 
-static void init_once(void *foo)
-{
-	struct v9fs_cookie *vcookie = (struct v9fs_cookie *) foo;
-	vcookie->fscache = NULL;
-	vcookie->qid = NULL;
-	inode_init_once(&vcookie->inode);
-}
-
-/**
- * v9fs_init_vcookiecache - initialize a cache for vcookies to maintain
- *			    vcookie to inode mapping
- *
- * Returns 0 on success.
- */
-
-static int v9fs_init_vcookiecache(void)
-{
-	vcookie_cache = kmem_cache_create("vcookie_cache",
-					  sizeof(struct v9fs_cookie),
-					  0, (SLAB_RECLAIM_ACCOUNT|
-					      SLAB_MEM_SPREAD),
-					  init_once);
-	if (!vcookie_cache)
-		return -ENOMEM;
-
-	return 0;
-}
-
-/**
- * v9fs_destroy_vcookiecache - destroy the cache of vcookies
- *
- */
-
-static void v9fs_destroy_vcookiecache(void)
-{
-	kmem_cache_destroy(vcookie_cache);
-}
-
-int __v9fs_cache_register(void)
-{
-	int ret;
-	ret = v9fs_init_vcookiecache();
-	if (ret < 0)
-		return ret;
-
-	return fscache_register_netfs(&v9fs_cache_netfs);
-}
-
-void __v9fs_cache_unregister(void)
-{
-	v9fs_destroy_vcookiecache();
-	fscache_unregister_netfs(&v9fs_cache_netfs);
-}
-
 /**
  * v9fs_random_cachetag - Generate a random tag to be associated
  *			  with a new cache session.
@@ -133,9 +77,9 @@ static uint16_t v9fs_cache_session_get_key(const void *cookie_netfs_data,
 }
 
 const struct fscache_cookie_def v9fs_cache_session_index_def = {
-	.name 		= "9P.session",
-	.type 		= FSCACHE_COOKIE_TYPE_INDEX,
-	.get_key 	= v9fs_cache_session_get_key,
+	.name		= "9P.session",
+	.type		= FSCACHE_COOKIE_TYPE_INDEX,
+	.get_key	= v9fs_cache_session_get_key,
 };
 
 void v9fs_cache_session_get_cookie(struct v9fs_session_info *v9ses)
@@ -163,33 +107,33 @@ void v9fs_cache_session_put_cookie(struct v9fs_session_info *v9ses)
 static uint16_t v9fs_cache_inode_get_key(const void *cookie_netfs_data,
 					 void *buffer, uint16_t bufmax)
 {
-	const struct v9fs_cookie *vcookie = cookie_netfs_data;
-	memcpy(buffer, &vcookie->qid->path, sizeof(vcookie->qid->path));
-
-	P9_DPRINTK(P9_DEBUG_FSC, "inode %p get key %llu", &vcookie->inode,
-		   vcookie->qid->path);
-	return sizeof(vcookie->qid->path);
+	const struct v9fs_inode *v9inode = cookie_netfs_data;
+	memcpy(buffer, &v9inode->fscache_key->path,
+	       sizeof(v9inode->fscache_key->path));
+	P9_DPRINTK(P9_DEBUG_FSC, "inode %p get key %llu", &v9inode->vfs_inode,
+		   v9inode->fscache_key->path);
+	return sizeof(v9inode->fscache_key->path);
 }
 
 static void v9fs_cache_inode_get_attr(const void *cookie_netfs_data,
 				      uint64_t *size)
 {
-	const struct v9fs_cookie *vcookie = cookie_netfs_data;
-	*size = i_size_read(&vcookie->inode);
+	const struct v9fs_inode *v9inode = cookie_netfs_data;
+	*size = i_size_read(&v9inode->vfs_inode);
 
-	P9_DPRINTK(P9_DEBUG_FSC, "inode %p get attr %llu", &vcookie->inode,
+	P9_DPRINTK(P9_DEBUG_FSC, "inode %p get attr %llu", &v9inode->vfs_inode,
 		   *size);
 }
 
 static uint16_t v9fs_cache_inode_get_aux(const void *cookie_netfs_data,
 					 void *buffer, uint16_t buflen)
 {
-	const struct v9fs_cookie *vcookie = cookie_netfs_data;
-	memcpy(buffer, &vcookie->qid->version, sizeof(vcookie->qid->version));
-
-	P9_DPRINTK(P9_DEBUG_FSC, "inode %p get aux %u", &vcookie->inode,
-		   vcookie->qid->version);
-	return sizeof(vcookie->qid->version);
+	const struct v9fs_inode *v9inode = cookie_netfs_data;
+	memcpy(buffer, &v9inode->fscache_key->version,
+	       sizeof(v9inode->fscache_key->version));
+	P9_DPRINTK(P9_DEBUG_FSC, "inode %p get aux %u", &v9inode->vfs_inode,
+		   v9inode->fscache_key->version);
+	return sizeof(v9inode->fscache_key->version);
 }
 
 static enum
@@ -197,13 +141,13 @@ fscache_checkaux v9fs_cache_inode_check_aux(void *cookie_netfs_data,
 					    const void *buffer,
 					    uint16_t buflen)
 {
-	const struct v9fs_cookie *vcookie = cookie_netfs_data;
+	const struct v9fs_inode *v9inode = cookie_netfs_data;
 
-	if (buflen != sizeof(vcookie->qid->version))
+	if (buflen != sizeof(v9inode->fscache_key->version))
 		return FSCACHE_CHECKAUX_OBSOLETE;
 
-	if (memcmp(buffer, &vcookie->qid->version,
-		   sizeof(vcookie->qid->version)))
+	if (memcmp(buffer, &v9inode->fscache_key->version,
+		   sizeof(v9inode->fscache_key->version)))
 		return FSCACHE_CHECKAUX_OBSOLETE;
 
 	return FSCACHE_CHECKAUX_OKAY;
@@ -211,7 +155,7 @@ fscache_checkaux v9fs_cache_inode_check_aux(void *cookie_netfs_data,
 
 static void v9fs_cache_inode_now_uncached(void *cookie_netfs_data)
 {
-	struct v9fs_cookie *vcookie = cookie_netfs_data;
+	struct v9fs_inode *v9inode = cookie_netfs_data;
 	struct pagevec pvec;
 	pgoff_t first;
 	int loop, nr_pages;
@@ -220,7 +164,7 @@ static void v9fs_cache_inode_now_uncached(void *cookie_netfs_data)
 	first = 0;
 
 	for (;;) {
-		nr_pages = pagevec_lookup(&pvec, vcookie->inode.i_mapping,
+		nr_pages = pagevec_lookup(&pvec, v9inode->vfs_inode.i_mapping,
 					  first,
 					  PAGEVEC_SIZE - pagevec_count(&pvec));
 		if (!nr_pages)
@@ -249,115 +193,114 @@ const struct fscache_cookie_def v9fs_cache_inode_index_def = {
 
 void v9fs_cache_inode_get_cookie(struct inode *inode)
 {
-	struct v9fs_cookie *vcookie;
+	struct v9fs_inode *v9inode;
 	struct v9fs_session_info *v9ses;
 
 	if (!S_ISREG(inode->i_mode))
 		return;
 
-	vcookie = v9fs_inode2cookie(inode);
-	if (vcookie->fscache)
+	v9inode = V9FS_I(inode);
+	if (v9inode->fscache)
 		return;
 
 	v9ses = v9fs_inode2v9ses(inode);
-	vcookie->fscache = fscache_acquire_cookie(v9ses->fscache,
+	v9inode->fscache = fscache_acquire_cookie(v9ses->fscache,
 						  &v9fs_cache_inode_index_def,
-						  vcookie);
+						  v9inode);
 
 	P9_DPRINTK(P9_DEBUG_FSC, "inode %p get cookie %p", inode,
-		   vcookie->fscache);
+		   v9inode->fscache);
 }
 
 void v9fs_cache_inode_put_cookie(struct inode *inode)
 {
-	struct v9fs_cookie *vcookie = v9fs_inode2cookie(inode);
+	struct v9fs_inode *v9inode = V9FS_I(inode);
 
-	if (!vcookie->fscache)
+	if (!v9inode->fscache)
 		return;
 	P9_DPRINTK(P9_DEBUG_FSC, "inode %p put cookie %p", inode,
-		   vcookie->fscache);
+		   v9inode->fscache);
 
-	fscache_relinquish_cookie(vcookie->fscache, 0);
-	vcookie->fscache = NULL;
+	fscache_relinquish_cookie(v9inode->fscache, 0);
+	v9inode->fscache = NULL;
 }
 
 void v9fs_cache_inode_flush_cookie(struct inode *inode)
 {
-	struct v9fs_cookie *vcookie = v9fs_inode2cookie(inode);
+	struct v9fs_inode *v9inode = V9FS_I(inode);
 
-	if (!vcookie->fscache)
+	if (!v9inode->fscache)
 		return;
 	P9_DPRINTK(P9_DEBUG_FSC, "inode %p flush cookie %p", inode,
-		   vcookie->fscache);
+		   v9inode->fscache);
 
-	fscache_relinquish_cookie(vcookie->fscache, 1);
-	vcookie->fscache = NULL;
+	fscache_relinquish_cookie(v9inode->fscache, 1);
+	v9inode->fscache = NULL;
 }
 
 void v9fs_cache_inode_set_cookie(struct inode *inode, struct file *filp)
 {
-	struct v9fs_cookie *vcookie = v9fs_inode2cookie(inode);
+	struct v9fs_inode *v9inode = V9FS_I(inode);
 	struct p9_fid *fid;
 
-	if (!vcookie->fscache)
+	if (!v9inode->fscache)
 		return;
 
-	spin_lock(&vcookie->lock);
+	spin_lock(&v9inode->fscache_lock);
 	fid = filp->private_data;
 	if ((filp->f_flags & O_ACCMODE) != O_RDONLY)
 		v9fs_cache_inode_flush_cookie(inode);
 	else
 		v9fs_cache_inode_get_cookie(inode);
 
-	spin_unlock(&vcookie->lock);
+	spin_unlock(&v9inode->fscache_lock);
 }
 
 void v9fs_cache_inode_reset_cookie(struct inode *inode)
 {
-	struct v9fs_cookie *vcookie = v9fs_inode2cookie(inode);
+	struct v9fs_inode *v9inode = V9FS_I(inode);
 	struct v9fs_session_info *v9ses;
 	struct fscache_cookie *old;
 
-	if (!vcookie->fscache)
+	if (!v9inode->fscache)
 		return;
 
-	old = vcookie->fscache;
+	old = v9inode->fscache;
 
-	spin_lock(&vcookie->lock);
-	fscache_relinquish_cookie(vcookie->fscache, 1);
+	spin_lock(&v9inode->fscache_lock);
+	fscache_relinquish_cookie(v9inode->fscache, 1);
 
 	v9ses = v9fs_inode2v9ses(inode);
-	vcookie->fscache = fscache_acquire_cookie(v9ses->fscache,
+	v9inode->fscache = fscache_acquire_cookie(v9ses->fscache,
 						  &v9fs_cache_inode_index_def,
-						  vcookie);
-
+						  v9inode);
 	P9_DPRINTK(P9_DEBUG_FSC, "inode %p revalidating cookie old %p new %p",
-		   inode, old, vcookie->fscache);
+		   inode, old, v9inode->fscache);
 
-	spin_unlock(&vcookie->lock);
+	spin_unlock(&v9inode->fscache_lock);
 }
 
 int __v9fs_fscache_release_page(struct page *page, gfp_t gfp)
 {
 	struct inode *inode = page->mapping->host;
-	struct v9fs_cookie *vcookie = v9fs_inode2cookie(inode);
+	struct v9fs_inode *v9inode = V9FS_I(inode);
 
-	BUG_ON(!vcookie->fscache);
+	BUG_ON(!v9inode->fscache);
 
-	return fscache_maybe_release_page(vcookie->fscache, page, gfp);
+	return fscache_maybe_release_page(v9inode->fscache, page, gfp);
 }
 
 void __v9fs_fscache_invalidate_page(struct page *page)
 {
 	struct inode *inode = page->mapping->host;
-	struct v9fs_cookie *vcookie = v9fs_inode2cookie(inode);
+	struct v9fs_inode *v9inode = V9FS_I(inode);
 
-	BUG_ON(!vcookie->fscache);
+	BUG_ON(!v9inode->fscache);
 
 	if (PageFsCache(page)) {
-		fscache_wait_on_page_write(vcookie->fscache, page);
+		fscache_wait_on_page_write(v9inode->fscache, page);
 		BUG_ON(!PageLocked(page));
-		fscache_uncache_page(vcookie->fscache, page);
+		fscache_uncache_page(v9inode->fscache, page);
 	}
 }
 
@@ -380,13 +323,13 @@ static void v9fs_vfs_readpage_complete(struct page *page, void *data,
 int __v9fs_readpage_from_fscache(struct inode *inode, struct page *page)
 {
 	int ret;
-	const struct v9fs_cookie *vcookie = v9fs_inode2cookie(inode);
+	const struct v9fs_inode *v9inode = V9FS_I(inode);
 
 	P9_DPRINTK(P9_DEBUG_FSC, "inode %p page %p", inode, page);
-	if (!vcookie->fscache)
+	if (!v9inode->fscache)
 		return -ENOBUFS;
 
-	ret = fscache_read_or_alloc_page(vcookie->fscache,
+	ret = fscache_read_or_alloc_page(v9inode->fscache,
 					 page,
 					 v9fs_vfs_readpage_complete,
 					 NULL,
@@ -418,13 +361,13 @@ int __v9fs_readpages_from_fscache(struct inode *inode,
 				  unsigned *nr_pages)
 {
 	int ret;
-	const struct v9fs_cookie *vcookie = v9fs_inode2cookie(inode);
+	const struct v9fs_inode *v9inode = V9FS_I(inode);
 
 	P9_DPRINTK(P9_DEBUG_FSC, "inode %p pages %u", inode, *nr_pages);
-	if (!vcookie->fscache)
+	if (!v9inode->fscache)
 		return -ENOBUFS;
 
-	ret = fscache_read_or_alloc_pages(vcookie->fscache,
+	ret = fscache_read_or_alloc_pages(v9inode->fscache,
 					  mapping, pages, nr_pages,
 					  v9fs_vfs_readpage_complete,
 					  NULL,
@@ -453,10 +396,10 @@ int __v9fs_readpages_from_fscache(struct inode *inode,
 void __v9fs_readpage_to_fscache(struct inode *inode, struct page *page)
 {
 	int ret;
-	const struct v9fs_cookie *vcookie = v9fs_inode2cookie(inode);
+	const struct v9fs_inode *v9inode = V9FS_I(inode);
 
 	P9_DPRINTK(P9_DEBUG_FSC, "inode %p page %p", inode, page);
-	ret = fscache_write_page(vcookie->fscache, page, GFP_KERNEL);
+	ret = fscache_write_page(v9inode->fscache, page, GFP_KERNEL);
 	P9_DPRINTK(P9_DEBUG_FSC, "ret =  %d", ret);
 	if (ret != 0)
 		v9fs_uncache_page(inode, page);
@@ -467,8 +410,8 @@ void __v9fs_readpage_to_fscache(struct inode *inode, struct page *page)
  */
 void __v9fs_fscache_wait_on_page_write(struct inode *inode, struct page *page)
 {
-	const struct v9fs_cookie *vcookie = v9fs_inode2cookie(inode);
+	const struct v9fs_inode *v9inode = V9FS_I(inode);
 	P9_DPRINTK(P9_DEBUG_FSC, "inode %p page %p", inode, page);
 	if (PageFsCache(page))
-		fscache_wait_on_page_write(vcookie->fscache, page);
+		fscache_wait_on_page_write(v9inode->fscache, page);
 }
diff --git a/fs/9p/cache.h b/fs/9p/cache.h
index ec16fcd..049507a 100644
--- a/fs/9p/cache.h
+++ b/fs/9p/cache.h
@@ -25,20 +25,6 @@
 #include <linux/fscache.h>
 #include <linux/spinlock.h>
 
-extern struct kmem_cache *vcookie_cache;
-
-struct v9fs_cookie {
-	spinlock_t lock;
-	struct inode inode;
-	struct fscache_cookie *fscache;
-	struct p9_qid *qid;
-};
-
-static inline struct v9fs_cookie *v9fs_inode2cookie(const struct inode *inode)
-{
-	return container_of(inode, struct v9fs_cookie, inode);
-}
-
 extern struct fscache_netfs v9fs_cache_netfs;
 extern const struct fscache_cookie_def v9fs_cache_session_index_def;
 extern const struct fscache_cookie_def v9fs_cache_inode_index_def;
@@ -66,21 +52,6 @@ extern int __v9fs_readpages_from_fscache(struct inode *inode,
 extern void __v9fs_readpage_to_fscache(struct inode *inode, struct page *page);
 extern void __v9fs_fscache_wait_on_page_write(struct inode *inode,
 					      struct page *page);
-/**
- * v9fs_cache_register - Register v9fs file system with the cache
- */
-static inline int v9fs_cache_register(void)
-{
-	return __v9fs_cache_register();
-}
-
-/**
- * v9fs_cache_unregister - Unregister v9fs from the cache
- */
-static inline void v9fs_cache_unregister(void)
-{
-	__v9fs_cache_unregister();
-}
 
 static inline int v9fs_fscache_release_page(struct page *page,
 					    gfp_t gfp)
@@ -117,18 +88,18 @@ static inline void v9fs_readpage_to_fscache(struct inode *inode,
 
 static inline void v9fs_uncache_page(struct inode *inode, struct page *page)
 {
-	struct v9fs_cookie *vcookie = v9fs_inode2cookie(inode);
-	fscache_uncache_page(vcookie->fscache, page);
+	struct v9fs_inode *v9inode = V9FS_I(inode);
+	fscache_uncache_page(v9inode->fscache, page);
 	BUG_ON(PageFsCache(page));
 }
 
-static inline void v9fs_vcookie_set_qid(struct inode *inode,
+static inline void v9fs_fscache_set_key(struct inode *inode,
 					struct p9_qid *qid)
 {
-	struct v9fs_cookie *vcookie = v9fs_inode2cookie(inode);
-	spin_lock(&vcookie->lock);
-	vcookie->qid = qid;
-	spin_unlock(&vcookie->lock);
+	struct v9fs_inode *v9inode = V9FS_I(inode);
+	spin_lock(&v9inode->fscache_lock);
+	v9inode->fscache_key = qid;
+	spin_unlock(&v9inode->fscache_lock);
 }
 
 static inline void v9fs_fscache_wait_on_page_write(struct inode *inode,
@@ -139,13 +110,6 @@ static inline void v9fs_fscache_wait_on_page_write(struct inode *inode,
 
 #else /* CONFIG_9P_FSCACHE */
 
-static inline int v9fs_cache_register(void)
-{
-	return 1;
-}
-
-static inline void v9fs_cache_unregister(void) {}
-
 static inline int v9fs_fscache_release_page(struct page *page,
 					    gfp_t gfp) {
 	return 1;
@@ -174,10 +138,6 @@ static inline void v9fs_readpage_to_fscache(struct inode *inode,
 static inline void v9fs_uncache_page(struct inode *inode, struct page *page)
 {}
 
-static inline void v9fs_vcookie_set_qid(struct inode *inode,
-					struct p9_qid *qid)
-{}
-
 static inline void v9fs_fscache_wait_on_page_write(struct inode *inode,
 						   struct page *page)
 {
diff --git a/fs/9p/v9fs.c b/fs/9p/v9fs.c
index 2f77cd3..814b77b 100644
--- a/fs/9p/v9fs.c
+++ b/fs/9p/v9fs.c
@@ -39,6 +39,7 @@
 
 static DEFINE_SPINLOCK(v9fs_sessionlist_lock);
 static LIST_HEAD(v9fs_sessionlist);
+struct kmem_cache *v9fs_inode_cache;
 
 /*
  * Option Parsing (code inspired by NFS code)
@@ -467,6 +468,63 @@ static void v9fs_sysfs_cleanup(void)
 	kobject_put(v9fs_kobj);
 }
 
+static void v9fs_inode_init_once(void *foo)
+{
+	struct v9fs_inode *v9inode = (struct v9fs_inode *)foo;
+#ifdef CONFIG_9P_FSCACHE
+	v9inode->fscache = NULL;
+	v9inode->fscache_key = NULL;
+#endif
+	inode_init_once(&v9inode->vfs_inode);
+}
+
+/**
+ * v9fs_init_inode_cache - initialize a cache for 9P
+ * Returns 0 on success.
+ */
+static int v9fs_init_inode_cache(void)
+{
+	v9fs_inode_cache = kmem_cache_create("v9fs_inode_cache",
+					  sizeof(struct v9fs_inode),
+					  0, (SLAB_RECLAIM_ACCOUNT|
+					      SLAB_MEM_SPREAD),
+					  v9fs_inode_init_once);
+	if (!v9fs_inode_cache)
+		return -ENOMEM;
+
+	return 0;
+}
+
+/**
+ * v9fs_destroy_inode_cache - destroy the cache of 9P inode
+ *
+ */
+static void v9fs_destroy_inode_cache(void)
+{
+	kmem_cache_destroy(v9fs_inode_cache);
+}
+
+static int v9fs_cache_register(void)
+{
+	int ret;
+	ret = v9fs_init_inode_cache();
+	if (ret < 0)
+		return ret;
+#ifdef CONFIG_9P_FSCACHE
+	return fscache_register_netfs(&v9fs_cache_netfs);
+#else
+	return ret;
+#endif
+}
+
+static void v9fs_cache_unregister(void)
+{
+	v9fs_destroy_inode_cache();
+#ifdef CONFIG_9P_FSCACHE
+	fscache_unregister_netfs(&v9fs_cache_netfs);
+#endif
+}
+
 /**
  * init_v9fs - Initialize module
  *
diff --git a/fs/9p/v9fs.h b/fs/9p/v9fs.h
index 41f29b9..e852b20 100644
--- a/fs/9p/v9fs.h
+++ b/fs/9p/v9fs.h
@@ -112,6 +112,20 @@ struct v9fs_session_info {
 	struct p9_fid *root_fid; /* Used for file system sync */
 };
 
+struct v9fs_inode {
+#ifdef CONFIG_9P_FSCACHE
+	spinlock_t fscache_lock;
+	struct fscache_cookie *fscache;
+	struct p9_qid *fscache_key;
+#endif
+	struct inode vfs_inode;
+};
+
+static inline struct v9fs_inode *V9FS_I(const struct inode *inode)
+{
+	return container_of(inode, struct v9fs_inode, vfs_inode);
+}
+
 struct p9_fid *v9fs_session_init(struct v9fs_session_info *, const char *,
 									char *);
 extern void v9fs_session_close(struct v9fs_session_info *v9ses);
@@ -125,16 +139,15 @@ extern int v9fs_vfs_rename(struct inode *old_dir, struct dentry *old_dentry,
 			struct inode *new_dir, struct dentry *new_dentry);
 extern void v9fs_vfs_put_link(struct dentry *dentry, struct nameidata *nd,
 			void *p);
-extern struct inode *v9fs_inode(struct v9fs_session_info *v9ses,
-			struct p9_fid *fid,
-			struct super_block *sb);
-
+extern struct inode *v9fs_inode_from_fid(struct v9fs_session_info *v9ses,
+					 struct p9_fid *fid,
+					 struct super_block *sb);
 extern const struct inode_operations v9fs_dir_inode_operations_dotl;
 extern const struct inode_operations v9fs_file_inode_operations_dotl;
 extern const struct inode_operations v9fs_symlink_inode_operations_dotl;
-extern struct inode *v9fs_inode_dotl(struct v9fs_session_info *v9ses,
-			struct p9_fid *fid,
-			struct super_block *sb);
+extern struct inode *v9fs_inode_from_fid_dotl(struct v9fs_session_info *v9ses,
+					      struct p9_fid *fid,
+					      struct super_block *sb);
 
 /* other default globals */
 #define V9FS_PORT	564
@@ -159,7 +172,7 @@ static inline int v9fs_proto_dotl(struct v9fs_session_info *v9ses)
 }
 
 /**
- * v9fs_inode_from_fid - Helper routine to populate an inode by
+ * v9fs_get_inode_from_fid - Helper routine to populate an inode by
  * issuing a attribute request
  * @v9ses: session information
  * @fid: fid to issue attribute request for
@@ -167,11 +180,11 @@ static inline int v9fs_proto_dotl(struct v9fs_session_info *v9ses)
  *
  */
 static inline struct inode *
-v9fs_inode_from_fid(struct v9fs_session_info *v9ses, struct p9_fid *fid,
-				struct super_block *sb)
+v9fs_get_inode_from_fid(struct v9fs_session_info *v9ses, struct p9_fid *fid,
+			struct super_block *sb)
 {
 	if (v9fs_proto_dotl(v9ses))
-		return v9fs_inode_dotl(v9ses, fid, sb);
+		return v9fs_inode_from_fid_dotl(v9ses, fid, sb);
 	else
-		return v9fs_inode(v9ses, fid, sb);
+		return v9fs_inode_from_fid(v9ses, fid, sb);
 }
diff --git a/fs/9p/v9fs_vfs.h b/fs/9p/v9fs_vfs.h
index ce9ba20..ce22080 100644
--- a/fs/9p/v9fs_vfs.h
+++ b/fs/9p/v9fs_vfs.h
@@ -47,12 +47,10 @@ extern const struct dentry_operations v9fs_dentry_operations;
 extern const struct dentry_operations v9fs_cached_dentry_operations;
 extern const struct file_operations v9fs_cached_file_operations;
 extern const struct file_operations v9fs_cached_file_operations_dotl;
+extern struct kmem_cache *v9fs_inode_cache;
 
-#ifdef CONFIG_9P_FSCACHE
 struct inode *v9fs_alloc_inode(struct super_block *sb);
 void v9fs_destroy_inode(struct inode *inode);
-#endif
-
 struct inode *v9fs_get_inode(struct super_block *sb, int mode);
 int v9fs_init_inode(struct v9fs_session_info *v9ses,
 		    struct inode *inode, int mode);
diff --git a/fs/9p/vfs_inode.c b/fs/9p/vfs_inode.c
index 8cf88eb..2de196b 100644
--- a/fs/9p/vfs_inode.c
+++ b/fs/9p/vfs_inode.c
@@ -203,26 +203,23 @@ v9fs_blank_wstat(struct p9_wstat *wstat)
 	wstat->extension = NULL;
 }
 
-#ifdef CONFIG_9P_FSCACHE
 /**
  * v9fs_alloc_inode - helper function to allocate an inode
- * This callback is executed before setting up the inode so that we
- * can associate a vcookie with each inode.
  *
  */
-
 struct inode *v9fs_alloc_inode(struct super_block *sb)
 {
-	struct v9fs_cookie *vcookie;
-	vcookie = (struct v9fs_cookie *)kmem_cache_alloc(vcookie_cache,
-							 GFP_KERNEL);
-	if (!vcookie)
+	struct v9fs_inode *v9inode;
+	v9inode = (struct v9fs_inode *)kmem_cache_alloc(v9fs_inode_cache,
+							GFP_KERNEL);
+	if (!v9inode)
 		return NULL;
-
-	vcookie->fscache = NULL;
-	vcookie->qid = NULL;
-	spin_lock_init(&vcookie->lock);
-	return &vcookie->inode;
+#ifdef CONFIG_9P_FSCACHE
+	v9inode->fscache = NULL;
+	v9inode->fscache_key = NULL;
+	spin_lock_init(&v9inode->fscache_lock);
+#endif
+	return &v9inode->vfs_inode;
 }
 
 /**
@@ -234,14 +231,13 @@ static void v9fs_i_callback(struct rcu_head *head)
 {
 	struct inode *inode = container_of(head, struct inode, i_rcu);
 	INIT_LIST_HEAD(&inode->i_dentry);
-	kmem_cache_free(vcookie_cache, v9fs_inode2cookie(inode));
+	kmem_cache_free(v9fs_inode_cache, V9FS_I(inode));
 }
 
 void v9fs_destroy_inode(struct inode *inode)
 {
 	call_rcu(&inode->i_rcu, v9fs_i_callback);
 }
-#endif
 
 int v9fs_init_inode(struct v9fs_session_info *v9ses,
 		    struct inode *inode, int mode)
@@ -458,7 +454,7 @@ static struct inode *v9fs_qid_iget(struct super_block *sb,
 
 	v9fs_stat2inode(st, inode, sb);
 #ifdef CONFIG_9P_FSCACHE
-	v9fs_vcookie_set_qid(ret, &st->qid);
+	v9fs_fscache_set_key(inode, &st->qid);
 	v9fs_cache_inode_get_cookie(inode);
 #endif
 	unlock_new_inode(inode);
@@ -471,8 +467,8 @@ error:
 }
 
 struct inode *
-v9fs_inode(struct v9fs_session_info *v9ses, struct p9_fid *fid,
-	   struct super_block *sb)
+v9fs_inode_from_fid(struct v9fs_session_info *v9ses, struct p9_fid *fid,
+		    struct super_block *sb)
 {
 	struct p9_wstat *st;
 	struct inode *inode = NULL;
@@ -571,7 +567,7 @@ v9fs_create(struct v9fs_session_info *v9ses, struct inode *dir,
 	}
 
 	/* instantiate inode and assign the unopened fid to the dentry */
-	inode = v9fs_inode_from_fid(v9ses, fid, dir->i_sb);
+	inode = v9fs_get_inode_from_fid(v9ses, fid, dir->i_sb);
 	if (IS_ERR(inode)) {
 		err = PTR_ERR(inode);
 		P9_DPRINTK(P9_DEBUG_VFS, "inode creation failed %d\n", err);
@@ -758,7 +754,7 @@ struct dentry *v9fs_vfs_lookup(struct inode *dir, struct dentry *dentry,
 		return ERR_PTR(result);
 	}
 
-	inode = v9fs_inode_from_fid(v9ses, fid, dir->i_sb);
+	inode = v9fs_get_inode_from_fid(v9ses, fid, dir->i_sb);
 	if (IS_ERR(inode)) {
 		result = PTR_ERR(inode);
 		inode = NULL;
diff --git a/fs/9p/vfs_inode_dotl.c b/fs/9p/vfs_inode_dotl.c
index 4f27fee..95da317 100644
--- a/fs/9p/vfs_inode_dotl.c
+++ b/fs/9p/vfs_inode_dotl.c
@@ -113,7 +113,7 @@ static struct inode *v9fs_qid_iget_dotl(struct super_block *sb,
 
 	v9fs_stat2inode_dotl(st, inode);
 #ifdef CONFIG_9P_FSCACHE
-	v9fs_vcookie_set_qid(inode, &st->qid);
+	v9fs_fscache_set_key(inode, &st->qid);
 	v9fs_cache_inode_get_cookie(inode);
 #endif
 	retval = v9fs_get_acl(inode, fid);
@@ -130,8 +130,8 @@ error:
 }
 
 struct inode *
-v9fs_inode_dotl(struct v9fs_session_info *v9ses, struct p9_fid *fid,
-	struct super_block *sb)
+v9fs_inode_from_fid_dotl(struct v9fs_session_info *v9ses, struct p9_fid *fid,
+			 struct super_block *sb)
 {
 	struct p9_stat_dotl *st;
 	struct inode *inode = NULL;
@@ -228,7 +228,7 @@ v9fs_vfs_create_dotl(struct inode *dir, struct dentry *dentry, int omode,
 		fid = NULL;
 		goto error;
 	}
-	inode = v9fs_inode_from_fid(v9ses, fid, dir->i_sb);
+	inode = v9fs_get_inode_from_fid(v9ses, fid, dir->i_sb);
 	if (IS_ERR(inode)) {
 		err = PTR_ERR(inode);
 		P9_DPRINTK(P9_DEBUG_VFS, "inode creation failed %d\n", err);
@@ -353,7 +353,7 @@ static int v9fs_vfs_mkdir_dotl(struct inode *dir,
 			goto error;
 		}
 
-		inode = v9fs_inode_from_fid(v9ses, fid, dir->i_sb);
+		inode = v9fs_get_inode_from_fid(v9ses, fid, dir->i_sb);
 		if (IS_ERR(inode)) {
 			err = PTR_ERR(inode);
 			P9_DPRINTK(P9_DEBUG_VFS, "inode creation failed %d\n",
@@ -600,7 +600,7 @@ v9fs_vfs_symlink_dotl(struct inode *dir, struct dentry *dentry,
 		}
 
 		/* instantiate inode and assign the unopened fid to dentry */
-		inode = v9fs_inode_from_fid(v9ses, fid, dir->i_sb);
+		inode = v9fs_get_inode_from_fid(v9ses, fid, dir->i_sb);
 		if (IS_ERR(inode)) {
 			err = PTR_ERR(inode);
 			P9_DPRINTK(P9_DEBUG_VFS, "inode creation failed %d\n",
@@ -759,7 +759,7 @@ v9fs_vfs_mknod_dotl(struct inode *dir, struct dentry *dentry, int omode,
 			goto error;
 		}
 
-		inode = v9fs_inode_from_fid(v9ses, fid, dir->i_sb);
+		inode = v9fs_get_inode_from_fid(v9ses, fid, dir->i_sb);
 		if (IS_ERR(inode)) {
 			err = PTR_ERR(inode);
 			P9_DPRINTK(P9_DEBUG_VFS, "inode creation failed %d\n",
diff --git a/fs/9p/vfs_super.c b/fs/9p/vfs_super.c
index fde5fe2..4838a56 100644
--- a/fs/9p/vfs_super.c
+++ b/fs/9p/vfs_super.c
@@ -292,10 +292,8 @@ static int v9fs_sync_fs(struct super_block *sb, int wait)
 }
 
 static const struct super_operations v9fs_super_ops = {
-#ifdef CONFIG_9P_FSCACHE
 	.alloc_inode = v9fs_alloc_inode,
 	.destroy_inode = v9fs_destroy_inode,
-#endif
 	.statfs = simple_statfs,
 	.evict_inode = v9fs_evict_inode,
 	.show_options = generic_show_options,
@@ -303,10 +301,8 @@ static const struct super_operations v9fs_super_ops = {
 };
 
 static const struct super_operations v9fs_super_ops_dotl = {
-#ifdef CONFIG_9P_FSCACHE
 	.alloc_inode = v9fs_alloc_inode,
 	.destroy_inode = v9fs_destroy_inode,
-#endif
 	.sync_fs = v9fs_sync_fs,
 	.statfs = v9fs_statfs,
 	.evict_inode = v9fs_evict_inode,
-- 
1.7.1

[RFC PATCH -V2 16/17] fs/9p: Move writeback fid to v9fs_inode

[Aneesh Kumar K.V]

Signed-off-by: Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
---
 fs/9p/v9fs.h           |    1 +
 fs/9p/vfs_addr.c       |   14 +++++++++-----
 fs/9p/vfs_file.c       |   12 ++++++++----
 fs/9p/vfs_inode.c      |   20 ++++++++++++--------
 fs/9p/vfs_inode_dotl.c |   16 +++++++++-------
 5 files changed, 39 insertions(+), 24 deletions(-)

diff --git a/fs/9p/v9fs.h b/fs/9p/v9fs.h
index e852b20..aed2a25 100644
--- a/fs/9p/v9fs.h
+++ b/fs/9p/v9fs.h
@@ -118,6 +118,7 @@ struct v9fs_inode {
 	struct fscache_cookie *fscache;
 	struct p9_qid *fscache_key;
 #endif
+	struct p9_fid *writeback_fid;
 	struct inode vfs_inode;
 };
 
diff --git a/fs/9p/vfs_addr.c b/fs/9p/vfs_addr.c
index 792c68a..8eecfa3 100644
--- a/fs/9p/vfs_addr.c
+++ b/fs/9p/vfs_addr.c
@@ -163,8 +163,10 @@ static int v9fs_vfs_writepage_locked(struct page *page)
 	int retval, len;
 	loff_t offset, size;
 	mm_segment_t old_fs;
+	struct v9fs_inode *v9inode;
 	struct inode *inode = page->mapping->host;
 
+	v9inode = V9FS_I(inode);
 	size = i_size_read(inode);
 	if (page->index == size >> PAGE_CACHE_SHIFT)
 		len = size & ~PAGE_CACHE_MASK;
@@ -178,11 +180,11 @@ static int v9fs_vfs_writepage_locked(struct page *page)
 
 	old_fs = get_fs();
 	set_fs(get_ds());
-	/* We should have i_private always set */
-	BUG_ON(!inode->i_private);
+	/* We should have writeback_fid always set */
+	BUG_ON(!v9inode->writeback_fid);
 
 	retval = v9fs_file_write_internal(inode,
-					  (struct p9_fid *)inode->i_private,
+					  v9inode->writeback_fid,
 					  buffer, len, &offset, 0);
 	if (retval > 0)
 		retval = 0;
@@ -273,23 +275,25 @@ static int v9fs_write_begin(struct file *filp, struct address_space *mapping,
 {
 	int retval = 0;
 	struct page *page;
+	struct v9fs_inode *v9inode;
 	pgoff_t index = pos >> PAGE_CACHE_SHIFT;
 	struct inode *inode = mapping->host;
 
+	v9inode = V9FS_I(inode);
 start:
 	page = grab_cache_page_write_begin(mapping, index, flags);
 	if (!page) {
 		retval = -ENOMEM;
 		goto out;
 	}
-	BUG_ON(!inode->i_private);
+	BUG_ON(!v9inode->writeback_fid);
 	if (PageUptodate(page))
 		goto out;
 
 	if (len == PAGE_CACHE_SIZE)
 		goto out;
 
-	retval = v9fs_fid_readpage(inode->i_private, page);
+	retval = v9fs_fid_readpage(v9inode->writeback_fid, page);
 	page_cache_release(page);
 	if (!retval)
 		goto start;
diff --git a/fs/9p/vfs_file.c b/fs/9p/vfs_file.c
index 6e38947..67f01bb 100644
--- a/fs/9p/vfs_file.c
+++ b/fs/9p/vfs_file.c
@@ -56,11 +56,13 @@ static const struct vm_operations_struct v9fs_file_vm_ops;
 int v9fs_file_open(struct inode *inode, struct file *file)
 {
 	int err;
+	struct v9fs_inode *v9inode;
 	struct v9fs_session_info *v9ses;
 	struct p9_fid *fid;
 	int omode;
 
 	P9_DPRINTK(P9_DEBUG_VFS, "inode: %p file: %p\n", inode, file);
+	v9inode = V9FS_I(inode);
 	v9ses = v9fs_inode2v9ses(inode);
 	if (v9fs_proto_dotl(v9ses))
 		omode = file->f_flags;
@@ -95,9 +97,9 @@ int v9fs_file_open(struct inode *inode, struct file *file)
 	 * mode so that a partial page write which result in page
 	 * read can work.
 	 */
-	if (v9ses->cache && !inode->i_private) {
+	if (v9ses->cache && !v9inode->writeback_fid) {
 		/*
-		 * clone a fid and add it to inode->i_private
+		 * clone a fid and add it to writeback_fid
 		 * we do it during open time instead of
 		 * page dirty time via write_begin/page_mkwrite
 		 * because we want write after unlink usecase
@@ -113,7 +115,7 @@ int v9fs_file_open(struct inode *inode, struct file *file)
 			p9_client_clunk(fid);
 			goto out_error;
 		}
-		inode->i_private = (void *) fid;
+		v9inode->writeback_fid = (void *) fid;
 	}
 #ifdef CONFIG_9P_FSCACHE
 	if (v9ses->cache)
@@ -562,6 +564,7 @@ v9fs_file_mmap(struct file *file, struct vm_area_struct *vma)
 static int
 v9fs_vm_page_mkwrite(struct vm_area_struct *vma, struct vm_fault *vmf)
 {
+	struct v9fs_inode *v9inode;
 	struct page *page = vmf->page;
 	struct file *filp = vma->vm_file;
 	struct inode *inode = filp->f_path.dentry->d_inode;
@@ -570,9 +573,10 @@ v9fs_vm_page_mkwrite(struct vm_area_struct *vma, struct vm_fault *vmf)
 	P9_DPRINTK(P9_DEBUG_VFS, "page %p fid %lx\n",
 		   page, (unsigned long)filp->private_data);
 
+	v9inode = V9FS_I(inode);
 	/* make sure the cache has finished storing the page */
 	v9fs_fscache_wait_on_page_write(inode, page);
-	BUG_ON(!inode->i_private);
+	BUG_ON(!v9inode->writeback_fid);
 	lock_page(page);
 	if (page->mapping != inode->i_mapping)
 		goto out_unlock;
diff --git a/fs/9p/vfs_inode.c b/fs/9p/vfs_inode.c
index 2de196b..b68f116 100644
--- a/fs/9p/vfs_inode.c
+++ b/fs/9p/vfs_inode.c
@@ -413,6 +413,8 @@ error:
  */
 void v9fs_evict_inode(struct inode *inode)
 {
+	struct v9fs_inode *v9inode = V9FS_I(inode);
+
 	truncate_inode_pages(inode->i_mapping, 0);
 	end_writeback(inode);
 	filemap_fdatawrite(inode->i_mapping);
@@ -420,10 +422,10 @@ void v9fs_evict_inode(struct inode *inode)
 #ifdef CONFIG_9P_FSCACHE
 	v9fs_cache_inode_put_cookie(inode);
 #endif
-	/* clunk the fid stashed in inode->i_private */
-	if (inode->i_private) {
-		p9_client_clunk((struct p9_fid *)inode->i_private);
-		inode->i_private = 0;
+	/* clunk the fid stashed in writeback_fid */
+	if (v9inode->writeback_fid) {
+		p9_client_clunk(v9inode->writeback_fid);
+		v9inode->writeback_fid = 0;
 	}
 }
 
@@ -606,9 +608,10 @@ v9fs_vfs_create(struct inode *dir, struct dentry *dentry, int mode,
 	int err;
 	u32 perm;
 	int flags;
+	struct file *filp;
+	struct v9fs_inode *v9inode;
 	struct v9fs_session_info *v9ses;
 	struct p9_fid *fid, *inode_fid;
-	struct file *filp;
 
 	err = 0;
 	fid = NULL;
@@ -630,6 +633,7 @@ v9fs_vfs_create(struct inode *dir, struct dentry *dentry, int mode,
 
 	/* if we are opening a file, assign the open fid to the file */
 	if (nd && nd->flags & LOOKUP_OPEN) {
+		v9inode = V9FS_I(dentry->d_inode);
 		/*
 		 * In cached mode we need to attach a fid to inode. The fid
 		 * attached to inode will only be used to write back the
@@ -637,9 +641,9 @@ v9fs_vfs_create(struct inode *dir, struct dentry *dentry, int mode,
 		 * mode so that a partial page write which result in page
 		 * read can work.
 		 */
-		if (v9ses->cache && !dentry->d_inode->i_private) {
+		if (v9ses->cache && !v9inode->writeback_fid) {
 			/*
-			 * clone a fid and add it to inode->i_private
+			 * clone a fid and add it to writeback_fid
 			 * we do it during open time instead of
 			 * page dirty time via write_begin/page_mkwrite
 			 * because we want write after unlink usecase
@@ -655,7 +659,7 @@ v9fs_vfs_create(struct inode *dir, struct dentry *dentry, int mode,
 				p9_client_clunk(inode_fid);
 				goto error;
 			}
-			dentry->d_inode->i_private = (void *) inode_fid;
+			v9inode->writeback_fid = (void *) inode_fid;
 		}
 		filp = lookup_instantiate_filp(nd, dentry, generic_file_open);
 		if (IS_ERR(filp)) {
diff --git a/fs/9p/vfs_inode_dotl.c b/fs/9p/vfs_inode_dotl.c
index 95da317..aebe532 100644
--- a/fs/9p/vfs_inode_dotl.c
+++ b/fs/9p/vfs_inode_dotl.c
@@ -159,16 +159,17 @@ v9fs_vfs_create_dotl(struct inode *dir, struct dentry *dentry, int omode,
 		struct nameidata *nd)
 {
 	int err = 0;
-	char *name = NULL;
 	gid_t gid;
 	int flags;
 	mode_t mode;
-	struct v9fs_session_info *v9ses;
-	struct p9_fid *fid = NULL;
-	struct p9_fid *dfid, *ofid, *inode_fid;
+	char *name = NULL;
 	struct file *filp;
 	struct p9_qid qid;
 	struct inode *inode;
+	struct p9_fid *fid = NULL;
+	struct v9fs_inode *v9inode;
+	struct p9_fid *dfid, *ofid, *inode_fid;
+	struct v9fs_session_info *v9ses;
 	struct posix_acl *pacl = NULL, *dacl = NULL;
 
 	v9ses = v9fs_inode2v9ses(dir);
@@ -248,9 +249,10 @@ v9fs_vfs_create_dotl(struct inode *dir, struct dentry *dentry, int omode,
 	 * mode so that a partial page write which result in page
 	 * read can work.
 	 */
-	if (v9ses->cache && !inode->i_private) {
+	v9inode = V9FS_I(inode);
+	if (v9ses->cache && !v9inode->writeback_fid) {
 		/*
-		 * clone a fid and add it to inode->i_private
+		 * clone a fid and add it to writeback_fid
 		 * we do it during open time instead of
 		 * page dirty time via write_begin/page_mkwrite
 		 * because we want write after unlink usecase
@@ -266,7 +268,7 @@ v9fs_vfs_create_dotl(struct inode *dir, struct dentry *dentry, int omode,
 			p9_client_clunk(inode_fid);
 			goto error;
 		}
-		inode->i_private = (void *) inode_fid;
+		v9inode->writeback_fid = (void *) inode_fid;
 	}
 	/* Since we are opening a file, assign the open fid to the file */
 	filp = lookup_instantiate_filp(nd, dentry, generic_file_open);
-- 
1.7.1

[RFC PATCH -V2 17/17] fs/9p: set default readahead pages in cached mode

[Aneesh Kumar K.V]

We want to enable readahead in cached mode

Signed-off-by: Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
---
 fs/9p/vfs_super.c |    2 ++
 1 files changed, 2 insertions(+), 0 deletions(-)

diff --git a/fs/9p/vfs_super.c b/fs/9p/vfs_super.c
index 4838a56..cfb8e97 100644
--- a/fs/9p/vfs_super.c
+++ b/fs/9p/vfs_super.c
@@ -86,6 +86,8 @@ v9fs_fill_super(struct super_block *sb, struct v9fs_session_info *v9ses,
 	} else
 		sb->s_op = &v9fs_super_ops;
 	sb->s_bdi = &v9ses->bdi;
+	if (v9ses->cache)
+		sb->s_bdi->ra_pages = (VM_MAX_READAHEAD * 1024)/PAGE_CACHE_SIZE;
 
 	sb->s_flags = flags | MS_ACTIVE | MS_DIRSYNC | MS_NOATIME;
 	if (!v9ses->cache)
-- 
1.7.1

Re: [RFC PATCH -V2 0/17] Buffered write and writeable mmap support for 9P

[Aneesh Kumar K. V]

On Sat,  5 Feb 2011 23:16:28 +0530, "Aneesh Kumar K.V" <aneesh.kumar@linux.vnet.ibm.com> wrote:
> The patch series implement buffered write and writeable mmap for 9P
> 
> Changes from V1:
> a) Add fsync and syncfs support
> 

I haven't yet addressed the previous review feedback from
Miklos. http://article.gmane.org/gmane.linux.kernel/1093869
Running 9p server and client on the same host even though possible
is not a strong use case at this point. I will get to that later.

The patch series also don't address cache consistency issue that
can arise due to direct modification in the server. I am not
sure we need to complicate the client code for that considering
the cache model that we follow is loose. 

In short the set of patches help the virtualization pass-through use case
(http://wiki.qemu.org/Documentation/9psetup)

-aneesh

Re: [V9fs-developer] [RFC PATCH -V2 0/17] Buffered write and writeable mmap support for 9P

[Eric Van Hensbergen]

On Sat, Feb 5, 2011 at 12:03 PM, Aneesh Kumar K. V
<aneesh.kumar@linux.vnet.ibm.com> wrote:
> On Sat,  5 Feb 2011 23:16:28 +0530, "Aneesh Kumar K.V" <aneesh.kumar@linux.vnet.ibm.com> wrote:
>> The patch series implement buffered write and writeable mmap for 9P
>>
>> Changes from V1:
>> a) Add fsync and syncfs support
>>
>
> I haven't yet addressed the previous review feedback from
> Miklos. http://article.gmane.org/gmane.linux.kernel/1093869
> Running 9p server and client on the same host even though possible
> is not a strong use case at this point. I will get to that later.
>

It is for a lot of the legacy cases including using 9p to access
user-space file servers.  Of course, there are some long standing
issues with this for anything which re-exports the name space.  Also -
most user-space file servers on the same host wouldn't want cache
support anyways.

>
> The patch series also don't address cache consistency issue that
> can arise due to direct modification in the server. I am not
> sure we need to complicate the client code for that considering
> the cache model that we follow is loose.
>

Agreed, the loose cache model doesn't respect any consistency
guarantees.  Once we implement any of the more conventional cache
policies we'll have to worry about it.

       -eric
--
To unsubscribe from this list: send the line "unsubscribe linux-fsdevel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html