[PATCH 1/8] fs/namespace.c: introduce helper function path_unmounted()

[PATCH 1/8] fs/namespace.c: introduce helper function path_unmounted()

[Yan Hong]

In fs/namespace.c, we verify whether a path is a mountpoint by
comparing path->dentry and path->mnt->mnt_root. Introduce
path_unmounted() to make code more readable.

Signed-off-by: Yan Hong <clouds.yan@gmail.com>
---
 fs/namespace.c |   19 ++++++++++++-------
 1 file changed, 12 insertions(+), 7 deletions(-)

diff --git a/fs/namespace.c b/fs/namespace.c
index 4d31f73..3fdc239 100644
--- a/fs/namespace.c
+++ b/fs/namespace.c
@@ -1234,6 +1234,11 @@ static int do_umount(struct mount *mnt, int flags)
 	return retval;
 }
 
+static inline bool path_unmounted(struct path *path)
+{
+	return path->dentry != path->mnt->mnt_root;
+}
+
 /*
  * Now umount can handle mount points as well as block devices.
  * This is important for filesystems which use unnamed block devices.
@@ -1260,7 +1265,7 @@ SYSCALL_DEFINE2(umount, char __user *, name, int, flags)
 		goto out;
 	mnt = real_mount(path.mnt);
 	retval = -EINVAL;
-	if (path.dentry != path.mnt->mnt_root)
+	if (path_unmounted(&path))
 		goto dput_and_out;
 	if (!check_mnt(mnt))
 		goto dput_and_out;
@@ -1613,7 +1618,7 @@ static int do_change_type(struct path *path, int flag)
 	if (!capable(CAP_SYS_ADMIN))
 		return -EPERM;
 
-	if (path->dentry != path->mnt->mnt_root)
+	if (path_unmounted(path))
 		return -EINVAL;
 
 	type = flags_to_propagation_type(flag);
@@ -1727,7 +1732,7 @@ static int do_remount(struct path *path, int flags, int mnt_flags,
 	if (!check_mnt(mnt))
 		return -EINVAL;
 
-	if (path->dentry != path->mnt->mnt_root)
+	if (path_unmounted(path))
 		return -EINVAL;
 
 	err = security_sb_remount(sb, data);
@@ -1793,7 +1798,7 @@ static int do_move_mount(struct path *path, char *old_name)
 		goto out1;
 
 	err = -EINVAL;
-	if (old_path.dentry != old_path.mnt->mnt_root)
+	if (path_unmounted(&old_path))
 		goto out1;
 
 	if (!mnt_has_parent(old))
@@ -1892,7 +1897,7 @@ static int do_add_mount(struct mount *newmnt, struct path *path, int mnt_flags)
 	/* Refuse the same filesystem on the same mount point */
 	err = -EBUSY;
 	if (path->mnt->mnt_sb == newmnt->mnt.mnt_sb &&
-	    path->mnt->mnt_root == path->dentry)
+	    !path_unmounted(path))
 		goto unlock;
 
 	err = -EINVAL;
@@ -2534,11 +2539,11 @@ SYSCALL_DEFINE2(pivot_root, const char __user *, new_root,
 	    old.mnt == root.mnt)
 		goto out4; /* loop, on the same file system  */
 	error = -EINVAL;
-	if (root.mnt->mnt_root != root.dentry)
+	if (path_unmounted(&root))
 		goto out4; /* not a mountpoint */
 	if (!mnt_has_parent(root_mnt))
 		goto out4; /* not attached */
-	if (new.mnt->mnt_root != new.dentry)
+	if (path_unmounted(&new))
 		goto out4; /* not a mountpoint */
 	if (!mnt_has_parent(new_mnt))
 		goto out4; /* not attached */
-- 
1.7.9.5

[PATCH 2/8] fs/namespace.c: remove unused macro MNT_WRITER_UNDERFLOW_LIMIT

[Yan Hong]

Its users have gone.

Signed-off-by: Yan Hong <clouds.yan@gmail.com>
---
 fs/namespace.c |    1 -
 1 file changed, 1 deletion(-)

diff --git a/fs/namespace.c b/fs/namespace.c
index 3fdc239..dd969f8 100644
--- a/fs/namespace.c
+++ b/fs/namespace.c
@@ -59,7 +59,6 @@ static inline unsigned long hash(struct vfsmount *mnt, struct dentry *dentry)
 	return tmp & (HASH_SIZE - 1);
 }
 
-#define MNT_WRITER_UNDERFLOW_LIMIT -(1<<16)
 
 /*
  * allocation is serialized by namespace_sem, but we need the spinlock to
-- 
1.7.9.5

[PATCH 3/8] fs/namespace.c: trivial code clean

[Yan Hong]

Signed-off-by: Yan Hong <clouds.yan@gmail.com>
---
 fs/namespace.c |    4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/fs/namespace.c b/fs/namespace.c
index dd969f8..bbe9014 100644
--- a/fs/namespace.c
+++ b/fs/namespace.c
@@ -763,7 +763,7 @@ vfs_kern_mount(struct file_system_type *type, int flags, const char *name, void
 
 	mnt->mnt.mnt_root = root;
 	mnt->mnt.mnt_sb = root->d_sb;
-	mnt->mnt_mountpoint = mnt->mnt.mnt_root;
+	mnt->mnt_mountpoint = root;
 	mnt->mnt_parent = mnt;
 	br_write_lock(&vfsmount_lock);
 	list_add_tail(&mnt->mnt_instance, &root->d_sb->s_mounts);
@@ -798,7 +798,7 @@ static struct mount *clone_mnt(struct mount *old, struct dentry *root,
 	atomic_inc(&sb->s_active);
 	mnt->mnt.mnt_sb = sb;
 	mnt->mnt.mnt_root = dget(root);
-	mnt->mnt_mountpoint = mnt->mnt.mnt_root;
+	mnt->mnt_mountpoint = root;
 	mnt->mnt_parent = mnt;
 	br_write_lock(&vfsmount_lock);
 	list_add_tail(&mnt->mnt_instance, &sb->s_mounts);
-- 
1.7.9.5

[PATCH 4/8] fs/namespace.c: check permission early in sys_[u]mount

[Yan Hong]

We have several branches in sys_mount, each of them will check
CAP_SYS_ADMIN capability seperately. Do this check at the beginning
of sys_mount.

Also check permission as early as possible in sys_umount.

Signed-off-by: Yan Hong <clouds.yan@gmail.com>
---
 fs/namespace.c |   26 +++++++-------------------
 1 file changed, 7 insertions(+), 19 deletions(-)

diff --git a/fs/namespace.c b/fs/namespace.c
index bbe9014..ca2b6e9 100644
--- a/fs/namespace.c
+++ b/fs/namespace.c
@@ -1256,6 +1256,9 @@ SYSCALL_DEFINE2(umount, char __user *, name, int, flags)
 	if (flags & ~(MNT_FORCE | MNT_DETACH | MNT_EXPIRE | UMOUNT_NOFOLLOW))
 		return -EINVAL;
 
+	if (!capable(CAP_SYS_ADMIN))
+		return -EPERM;
+
 	if (!(flags & UMOUNT_NOFOLLOW))
 		lookup_flags |= LOOKUP_FOLLOW;
 
@@ -1269,10 +1272,6 @@ SYSCALL_DEFINE2(umount, char __user *, name, int, flags)
 	if (!check_mnt(mnt))
 		goto dput_and_out;
 
-	retval = -EPERM;
-	if (!capable(CAP_SYS_ADMIN))
-		goto dput_and_out;
-
 	retval = do_umount(mnt, flags);
 dput_and_out:
 	/* we mustn't call path_put() as that would clear mnt_expiry_mark */
@@ -1296,9 +1295,7 @@ SYSCALL_DEFINE1(oldumount, char __user *, name)
 
 static int mount_is_safe(struct path *path)
 {
-	if (capable(CAP_SYS_ADMIN))
-		return 0;
-	return -EPERM;
+	return 0;
 #ifdef notyet
 	if (S_ISLNK(path->dentry->d_inode->i_mode))
 		return -EPERM;
@@ -1614,9 +1611,6 @@ static int do_change_type(struct path *path, int flag)
 	int type;
 	int err = 0;
 
-	if (!capable(CAP_SYS_ADMIN))
-		return -EPERM;
-
 	if (path_unmounted(path))
 		return -EINVAL;
 
@@ -1725,9 +1719,6 @@ static int do_remount(struct path *path, int flags, int mnt_flags,
 	struct super_block *sb = path->mnt->mnt_sb;
 	struct mount *mnt = real_mount(path->mnt);
 
-	if (!capable(CAP_SYS_ADMIN))
-		return -EPERM;
-
 	if (!check_mnt(mnt))
 		return -EINVAL;
 
@@ -1774,8 +1765,6 @@ static int do_move_mount(struct path *path, char *old_name)
 	struct mount *p;
 	struct mount *old;
 	int err = 0;
-	if (!capable(CAP_SYS_ADMIN))
-		return -EPERM;
 	if (!old_name || !*old_name)
 		return -EINVAL;
 	err = kern_path(old_name, LOOKUP_FOLLOW, &old_path);
@@ -1924,10 +1913,6 @@ static int do_new_mount(struct path *path, char *type, int flags,
 	if (!type)
 		return -EINVAL;
 
-	/* we need capabilities... */
-	if (!capable(CAP_SYS_ADMIN))
-		return -EPERM;
-
 	mnt = do_kern_mount(type, flags, name, data);
 	if (IS_ERR(mnt))
 		return PTR_ERR(mnt);
@@ -2410,6 +2395,9 @@ SYSCALL_DEFINE5(mount, char __user *, dev_name, char __user *, dir_name,
 	char *kernel_dev;
 	unsigned long data_page;
 
+	if (!capable(CAP_SYS_ADMIN))
+		return -EPERM;
+
 	ret = copy_mount_string(type, &kernel_type);
 	if (ret < 0)
 		goto out_type;
-- 
1.7.9.5

[PATCH 5/8] fs/namei.c: introduce macro AT_FDINV

[Yan Hong]

File descriptor is irrelevent when LOOKUP_ROOT is set.
Introduce AT_FDINV to avoid using hard coded value or reusing
existing macro.

Signed-off-by: Yan Hong <clouds.yan@gmail.com>
---
 fs/namei.c            |   11 ++++++-----
 include/linux/fcntl.h |    2 ++
 2 files changed, 8 insertions(+), 5 deletions(-)

diff --git a/fs/namei.c b/fs/namei.c
index c5b85b3..5ffd97d 100644
--- a/fs/namei.c
+++ b/fs/namei.c
@@ -2034,8 +2034,8 @@ int vfs_path_lookup(struct dentry *dentry, struct vfsmount *mnt,
 	nd.root.dentry = dentry;
 	nd.root.mnt = mnt;
 	BUG_ON(flags & LOOKUP_PARENT);
-	/* the first argument of do_path_lookup() is ignored with LOOKUP_ROOT */
-	err = do_path_lookup(AT_FDCWD, name, flags | LOOKUP_ROOT, &nd);
+
+	err = do_path_lookup(AT_FDINV, name, flags | LOOKUP_ROOT, &nd);
 	if (!err)
 		*path = nd.path;
 	return err;
@@ -2969,11 +2969,12 @@ struct file *do_file_open_root(struct dentry *dentry, struct vfsmount *mnt,
 	if (dentry->d_inode->i_op->follow_link && op->intent & LOOKUP_OPEN)
 		return ERR_PTR(-ELOOP);
 
-	file = path_openat(-1, name, &nd, op, flags | LOOKUP_RCU);
+	file = path_openat(AT_FDINV, name, &nd, op, flags | LOOKUP_RCU);
 	if (unlikely(file == ERR_PTR(-ECHILD)))
-		file = path_openat(-1, name, &nd, op, flags);
+		file = path_openat(AT_FDINV, name, &nd, op, flags);
 	if (unlikely(file == ERR_PTR(-ESTALE)))
-		file = path_openat(-1, name, &nd, op, flags | LOOKUP_REVAL);
+		file = path_openat(AT_FDINV, name, &nd, op,
+						flags | LOOKUP_REVAL);
 	return file;
 }
 
diff --git a/include/linux/fcntl.h b/include/linux/fcntl.h
index f550f89..abd94fe 100644
--- a/include/linux/fcntl.h
+++ b/include/linux/fcntl.h
@@ -41,6 +41,8 @@
 #define AT_FDCWD		-100    /* Special value used to indicate
                                            openat should use the current
                                            working directory. */
+#define AT_FDINV		-200	/* Special value used when LOOKUP_ROOT
+					   is set. */
 #define AT_SYMLINK_NOFOLLOW	0x100   /* Do not follow symbolic links.  */
 #define AT_REMOVEDIR		0x200   /* Remove directory instead of
                                            unlinking file.  */
-- 
1.7.9.5

[PATCH 6/8] fs/inode.c: call alloc_inode() in new_inode() directly

[Yan Hong]

This saves us a list head initialization.

Signed-off-by: Yan Hong <clouds.yan@gmail.com>
---
 fs/inode.c |    6 ++++--
 1 file changed, 4 insertions(+), 2 deletions(-)

diff --git a/fs/inode.c b/fs/inode.c
index ac8d904..3a2cd41 100644
--- a/fs/inode.c
+++ b/fs/inode.c
@@ -921,9 +921,11 @@ struct inode *new_inode(struct super_block *sb)
 
 	spin_lock_prefetch(&inode_sb_list_lock);
 
-	inode = new_inode_pseudo(sb);
-	if (inode)
+	inode = alloc_inode(sb);
+	if (inode) {
+		inode->i_state = 0;
 		inode_sb_list_add(inode);
+	}
 	return inode;
 }
 EXPORT_SYMBOL(new_inode);
-- 
1.7.9.5

[PATCH 7/8] fs/inode.c: remove outstanding spin lock prefetch

[Yan Hong]

Do we have particular reason to do this here?

Signed-off-by: Yan Hong <clouds.yan@gmail.com>
---
 fs/inode.c |    3 ---
 1 file changed, 3 deletions(-)

diff --git a/fs/inode.c b/fs/inode.c
index 3a2cd41..e89d30c 100644
--- a/fs/inode.c
+++ b/fs/inode.c
@@ -14,7 +14,6 @@
 #include <linux/fsnotify.h>
 #include <linux/mount.h>
 #include <linux/posix_acl.h>
-#include <linux/prefetch.h>
 #include <linux/buffer_head.h> /* for inode_has_buffers */
 #include <linux/ratelimit.h>
 #include "internal.h"
@@ -919,8 +918,6 @@ struct inode *new_inode(struct super_block *sb)
 {
 	struct inode *inode;
 
-	spin_lock_prefetch(&inode_sb_list_lock);
-
 	inode = alloc_inode(sb);
 	if (inode) {
 		inode->i_state = 0;
-- 
1.7.9.5

[PATCH 8/8] vfs: misc comment clean

[Yan Hong]

Mostly fix comments which reference inexist locks or parameters.

Signed-off-by: Yan Hong <clouds.yan@gmail.com>
---
 fs/dcache.c    |   12 ++++--------
 fs/inode.c     |    4 ++--
 fs/namei.c     |    2 +-
 fs/namespace.c |    9 +++------
 4 files changed, 10 insertions(+), 17 deletions(-)

diff --git a/fs/dcache.c b/fs/dcache.c
index 8086636..2d97518 100644
--- a/fs/dcache.c
+++ b/fs/dcache.c
@@ -44,8 +44,6 @@
  * Usage:
  * dcache->d_inode->i_lock protects:
  *   - i_dentry, d_alias, d_inode of aliases
- * dcache_hash_bucket lock protects:
- *   - the dcache hash table
  * s_anon bl list spinlock protects:
  *   - the s_anon list (see __d_drop)
  * dcache_lru_lock protects:
@@ -64,7 +62,6 @@
  * dentry->d_inode->i_lock
  *   dentry->d_lock
  *     dcache_lru_lock
- *     dcache_hash_bucket lock
  *     s_anon lock
  *
  * If there is an ancestor relationship:
@@ -145,10 +142,9 @@ int proc_nr_dentry(ctl_table *table, int write, void __user *buffer,
 
 #include <asm/word-at-a-time.h>
 /*
- * NOTE! 'cs' and 'scount' come from a dentry, so it has a
- * aligned allocation for this particular component. We don't
- * strictly need the load_unaligned_zeropad() safety, but it
- * doesn't hurt either.
+ * NOTE! 'cs' comes from a dentry, so it has an aligned allocation
+ * for this particular component. We don't strictly need the
+ * load_unaligned_zeropad() safety, but it doesn't hurt either.
  *
  * In contrast, 'ct' and 'tcount' can be from a pathname, and do
  * need the careful unaligned handling.
@@ -304,7 +300,7 @@ static void dentry_unlink_inode(struct dentry * dentry)
 }
 
 /*
- * dentry_lru_(add|del|prune|move_tail) must be called with d_lock held.
+ * dentry_lru_(add|del|prune|move_list) must be called with d_lock held.
  */
 static void dentry_lru_add(struct dentry *dentry)
 {
diff --git a/fs/inode.c b/fs/inode.c
index e89d30c..46d4f16 100644
--- a/fs/inode.c
+++ b/fs/inode.c
@@ -676,7 +676,7 @@ static int can_unuse(struct inode *inode)
  * Walk the superblock inode LRU for freeable inodes and attempt to free them.
  * This is called from the superblock shrinker function with a number of inodes
  * to trim from the LRU. Inodes to be freed are moved to a temporary list and
- * then are freed outside inode_lock by dispose_list().
+ * then are freed outside s_inode_lru_lock by dispose_list().
  *
  * Any inodes which are pinned purely because of attached pagecache have their
  * pagecache removed.  If the inode has metadata buffers attached to
@@ -777,7 +777,7 @@ void prune_icache_sb(struct super_block *sb, int nr_to_scan)
 
 static void __wait_on_freeing_inode(struct inode *inode);
 /*
- * Called with the inode lock held.
+ * Called with the inode_hash_lock held.
  */
 static struct inode *find_inode(struct super_block *sb,
 				struct hlist_head *head,
diff --git a/fs/namei.c b/fs/namei.c
index 5ffd97d..7b1f714 100644
--- a/fs/namei.c
+++ b/fs/namei.c
@@ -421,7 +421,7 @@ EXPORT_SYMBOL(path_put);
  * Path walking has 2 modes, rcu-walk and ref-walk (see
  * Documentation/filesystems/path-lookup.txt).  In situations when we can't
  * continue in RCU mode, we attempt to drop out of rcu-walk mode and grab
- * normal reference counts on dentries and vfsmounts to transition to rcu-walk
+ * normal reference counts on dentries and vfsmounts to transition to ref-walk
  * mode.  Refcounts are grabbed at the last known good point before rcu-walk
  * got stuck, so ref-walk may continue from there. If this is not successful
  * (eg. a seqcount has changed), then failure is returned and it's up to caller
diff --git a/fs/namespace.c b/fs/namespace.c
index ca2b6e9..b7477ee 100644
--- a/fs/namespace.c
+++ b/fs/namespace.c
@@ -1239,9 +1239,6 @@ static inline bool path_unmounted(struct path *path)
 }
 
 /*
- * Now umount can handle mount points as well as block devices.
- * This is important for filesystems which use unnamed block devices.
- *
  * We now support a flag for forced unmount like the other 'big iron'
  * unixes. Our API is identical to OSF/1 to avoid making a mess of AMD
  */
@@ -1431,9 +1428,9 @@ static int invent_group_ids(struct mount *mnt, bool recurse)
 }
 
 /*
- *  @source_mnt : mount tree to be attached
- *  @nd         : place the mount tree @source_mnt is attached
- *  @parent_nd  : if non-null, detach the source_mnt from its parent and
+ *  @source_mnt  : mount tree to be attached
+ *  @path        : place the mount tree @source_mnt is attached
+ *  @parent_path : if non-null, detach the @source_mnt from its parent and
  *  		   store the parent mount and mountpoint dentry.
  *  		   (done when source_mnt is moved)
  *
-- 
1.7.9.5