Re: [RFC][1/6] Sysfs backing store release 0.1

From: Maneesh Soni
Date: Wed Feb 18 2004 - 08:04:32 EST




=> changes in version 0.1
o Corrected sysfs_umount_begin(), it doesnot need lock_super() and also the
s_root check is not required. The reason being, that sysfs filesystem is
always mounted internally during init and there is no chance of sysfs super
block going away.

o corrected comments for umount_begin()

============================================================================
o The following patch contains the sysfs_dirent structure definition.
sysfs_dirent can represent kobject, attribute group, text attribute or
binary attribute for kobjects registered with sysfs. sysfs_dirent is
allocated with a ref count (s_count) of 1. Ref count is incremented when
a dentry is associated with the sysfs_dirent and it is decremented when
the corresponding dentry is freed.

o sysfs_dirent's corresponding to the attribute files of a kobject or attribute
group are linked together with s_sibling and are anchored at s_children of
the corresponding kobject's or attribute-group's sysfs_dirent.

o The patch also contains the mount related changes for sysfs backing store.
Because we mount sysfs once while init(), plain umount of sysfs doesnot
free all the un-used dentries (present in LRU list). To use force umount
flag, umount_begin() routine is provided which does a shrink_dcache_parent()
to release all the unused dentries.



fs/sysfs/mount.c | 27 +++++++++++++++++++++++++--
include/linux/sysfs.h | 19 +++++++++++++++++++
2 files changed, 44 insertions(+), 2 deletions(-)

diff -puN fs/sysfs/mount.c~sysfs-leaves-mount fs/sysfs/mount.c
--- linux-2.6.3-mm1/fs/sysfs/mount.c~sysfs-leaves-mount 2004-02-18 14:47:52.000000000 +0530
+++ linux-2.6.3-mm1-maneesh/fs/sysfs/mount.c 2004-02-18 15:18:47.000000000 +0530
@@ -20,6 +20,14 @@ struct super_block * sysfs_sb = NULL;
static struct super_operations sysfs_ops = {
.statfs = simple_statfs,
.drop_inode = generic_delete_inode,
+ .umount_begin = sysfs_umount_begin,
+};
+
+struct sysfs_dirent sysfs_root = {
+ .s_sibling = LIST_HEAD_INIT(sysfs_root.s_sibling),
+ .s_children = LIST_HEAD_INIT(sysfs_root.s_children),
+ .s_element = NULL,
+ .s_type = SYSFS_ROOT,
};

static int sysfs_fill_super(struct super_block *sb, void *data, int silent)
@@ -35,8 +43,8 @@ static int sysfs_fill_super(struct super

inode = sysfs_new_inode(S_IFDIR | S_IRWXU | S_IRUGO | S_IXUGO);
if (inode) {
- inode->i_op = &simple_dir_inode_operations;
- inode->i_fop = &simple_dir_operations;
+ inode->i_op = &sysfs_dir_inode_operations;
+ inode->i_fop = &sysfs_dir_operations;
/* directory inodes start off with i_nlink == 2 (for "." entry) */
inode->i_nlink++;
} else {
@@ -50,6 +58,7 @@ static int sysfs_fill_super(struct super
iput(inode);
return -ENOMEM;
}
+ root->d_fsdata = &sysfs_root;
sb->s_root = root;
return 0;
}
@@ -60,6 +69,20 @@ static struct super_block *sysfs_get_sb(
return get_sb_single(fs_type, flags, data, sysfs_fill_super);
}

+/* For freeing zero refenced dentries / inodes while force unmounting
+ *
+ * sysfs is mounted once within kernel during init(), and this keeps the super
+ * block always active. So in case of user just doing normal umount,
+ * ->kill_sb is never called. So, in order to immediately free the memory used
+ * by un-used dentries and inodes, sysfs should to be umounted with force
+ * option. In anycase there will be normal pruning of unused dentries/inodes
+ * as usual due to memory pressure.
+ */
+void sysfs_umount_begin(struct super_block * sb)
+{
+ shrink_dcache_parent(sb->s_root);
+}
+
static struct file_system_type sysfs_fs_type = {
.name = "sysfs",
.get_sb = sysfs_get_sb,
diff -puN include/linux/sysfs.h~sysfs-leaves-mount include/linux/sysfs.h
--- linux-2.6.3-mm1/include/linux/sysfs.h~sysfs-leaves-mount 2004-02-18 14:40:32.000000000 +0530
+++ linux-2.6.3-mm1-maneesh/include/linux/sysfs.h 2004-02-18 14:48:03.000000000 +0530
@@ -9,6 +9,8 @@
#ifndef _SYSFS_H_
#define _SYSFS_H_

+#include <asm/atomic.h>
+
struct kobject;
struct module;

@@ -47,6 +49,23 @@ sysfs_remove_dir(struct kobject *);
extern void
sysfs_rename_dir(struct kobject *, const char *new_name);

+struct sysfs_dirent {
+ atomic_t s_count;
+ struct list_head s_sibling;
+ struct list_head s_children;
+ void * s_element;
+ int s_type;
+ struct dentry * s_dentry;
+};
+
+#define SYSFS_ROOT 0x0001
+#define SYSFS_KOBJECT 0x0002
+#define SYSFS_KOBJ_ATTR 0x0004
+#define SYSFS_KOBJ_BIN_ATTR 0x0008
+#define SYSFS_KOBJ_ATTR_GROUP 0x0010
+#define SYSFS_KOBJ_LINK 0x0020
+#define SYSFS_NOT_PINNED (SYSFS_KOBJ_ATTR | SYSFS_KOBJ_BIN_ATTR | SYSFS_KOBJ_LINK)
+
extern int
sysfs_create_file(struct kobject *, const struct attribute *);


_
--
Maneesh Soni
Linux Technology Center,
IBM Software Lab, Bangalore, India
email: maneesh@xxxxxxxxxx
Phone: 91-80-25044999 Fax: 91-80-5268553
T/L : 9243696
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/