This patch makes dentries and inodes for sysfs directories reclaimable. * sysfs_notify() is modified to walk sysfs_dirent tree instead of dentry tree. * sysfs_update_file() and sysfs_chmod_file() use sysfs_get_dentry() to grab the victim dentry. * sysfs_rename_dir() and sysfs_move_dir() grab all dentries using sysfs_get_dentry() on startup. * Dentries for all shadowed directories are pinned in memory to serve as lookup start point. * parent mtime/ctime update on inode instantiation is dropped. I' not sure what to do about it yet. Signed-off-by: Tejun Heo <htejun@xxxxxxxxx> --- fs/sysfs/dir.c | 238 ++++++++++++++++++++++++++++--------------------- fs/sysfs/file.c | 134 ++++++++++++--------------- fs/sysfs/inode.c | 15 +-- fs/sysfs/mount.c | 2 +- fs/sysfs/sysfs.h | 1 + include/linux/sysfs.h | 1 - 6 files changed, 202 insertions(+), 189 deletions(-) diff --git a/fs/sysfs/dir.c b/fs/sysfs/dir.c index 0242cbd..ae1f379 100644 --- a/fs/sysfs/dir.c +++ b/fs/sysfs/dir.c @@ -482,73 +482,46 @@ struct sysfs_dirent *sysfs_get_dirent(st static int create_dir(struct kobject *kobj, struct sysfs_dirent *parent_sd, const char *name, struct sysfs_dirent **p_sd) { - struct dentry *parent = parent_sd->s_dentry; - int error; umode_t mode = S_IFDIR| S_IRWXU | S_IRUGO | S_IXUGO; - struct dentry *dentry; - struct inode *inode; + struct dentry *parent_dentry = NULL; struct sysfs_dirent *sd; - - mutex_lock(&parent->d_inode->i_mutex); + int error; /* allocate */ - dentry = lookup_one_len(name, parent, strlen(name)); - if (IS_ERR(dentry)) { - error = PTR_ERR(dentry); - goto out_unlock; - } - - error = -EEXIST; - if (dentry->d_inode) - goto out_dput; - - error = -ENOMEM; sd = sysfs_new_dirent(name, mode, SYSFS_DIR); if (!sd) - goto out_drop; + return -ENOMEM; sd->s_elem.dir.kobj = kobj; - inode = sysfs_get_inode(sd); - if (!inode) - goto out_sput; - - if (inode->i_state & I_NEW) { - inode->i_op = &sysfs_dir_inode_operations; - inode->i_fop = &sysfs_dir_operations; - /* directory inodes start off with i_nlink == 2 (for ".") */ - inc_nlink(inode); - } - /* link in */ + error = -EEXIST; spin_lock(&sysfs_lock); + if (!sysfs_find_dirent(parent_sd, name)) { + sysfs_attach_dirent(sd, parent_sd, NULL); + error = 0; - error = -EEXIST; - if (sysfs_find_dirent(parent_sd, name)) { - spin_unlock(&sysfs_lock); - goto out_iput; + spin_lock(&dcache_lock); + if (parent_sd->s_dentry && parent_sd->s_dentry->d_inode) + parent_dentry = dget_locked(parent_sd->s_dentry); + spin_unlock(&dcache_lock); } + spin_unlock(&sysfs_lock); - sysfs_instantiate(dentry, inode); - inc_nlink(parent->d_inode); - sysfs_attach_dirent(sd, parent_sd, dentry); + if (error) { + sysfs_put(sd); + return error; + } - spin_unlock(&sysfs_lock); + /* adjust nlink of parent */ + if (parent_dentry) { + mutex_lock(&parent_dentry->d_inode->i_mutex); + inc_nlink(parent_dentry->d_inode); + mutex_unlock(&parent_dentry->d_inode->i_mutex); + dput(parent_dentry); + } *p_sd = sd; - error = 0; - goto out_unlock; /* pin directory dentry in core */ - - out_iput: - iput(inode); - out_sput: - sysfs_put(sd); - out_drop: - d_drop(dentry); - out_dput: - dput(dentry); - out_unlock: - mutex_unlock(&parent->d_inode->i_mutex); - return error; + return 0; } int sysfs_create_subdir(struct kobject *kobj, const char *name, @@ -582,9 +555,21 @@ int sysfs_create_dir(struct kobject * ko error = create_dir(kobj, parent_sd, kobject_name(kobj), &sd); if (!error) kobj->sd = sd; + return error; } +static int sysfs_nr_subdirs(struct sysfs_dirent *sd) +{ + struct sysfs_dirent *child; + int nr = 0; + + for (child = sd->s_children; child; child = child->s_sibling) + if (sysfs_type(child) & SYSFS_DIR) + nr++; + return nr; +} + static struct dentry * sysfs_lookup(struct inode *dir, struct dentry *dentry, struct nameidata *nd) { @@ -595,7 +580,7 @@ static struct dentry * sysfs_lookup(stru int found = 0; for (sd = parent_sd->s_children; sd; sd = sd->s_sibling) { - if ((sysfs_type(sd) & SYSFS_NOT_PINNED) && + if (sysfs_type(sd) && !strcmp(sd->s_name, dentry->d_name.name)) { found = 1; break; @@ -616,6 +601,11 @@ static struct dentry * sysfs_lookup(stru if (inode->i_state & I_NEW) { /* initialize inode according to type */ switch (sysfs_type(sd)) { + case SYSFS_DIR: + inode->i_op = &sysfs_dir_inode_operations; + inode->i_fop = &sysfs_dir_operations; + inode->i_nlink = 2 + sysfs_nr_subdirs(sd); + break; case SYSFS_KOBJ_ATTR: inode->i_size = PAGE_SIZE; inode->i_fop = &sysfs_file_operations; @@ -680,7 +670,7 @@ static void __sysfs_remove_dir(struct sy while (*pos) { struct sysfs_dirent *sd = *pos; - if (sysfs_type(sd) && (sysfs_type(sd) & SYSFS_NOT_PINNED)) { + if (sysfs_type(sd) && !(sysfs_type(sd) & SYSFS_DIR)) { sd->s_flags |= SYSFS_FLAG_REMOVED; *pos = sd->s_sibling; sd->s_sibling = removed; @@ -728,14 +718,25 @@ int sysfs_rename_dir(struct kobject *kob const char *new_name) { struct sysfs_dirent *sd = kobj->sd; - struct dentry *new_parent = new_parent_sd->s_dentry; - struct dentry *new_dentry; - char *dup_name; + struct dentry *new_parent = NULL; + struct dentry *old_dentry = NULL, *new_dentry = NULL; + const char *dup_name = NULL; int error; - if (!new_parent_sd) - return -EFAULT; + /* get dentries */ + old_dentry = sysfs_get_dentry(sd); + if (IS_ERR(old_dentry)) { + error = PTR_ERR(old_dentry); + goto out_dput; + } + + new_parent = sysfs_get_dentry(new_parent_sd); + if (IS_ERR(new_parent)) { + error = PTR_ERR(new_parent); + goto out_dput; + } + /* lock new_parent and get dentry for new name */ mutex_lock(&new_parent->d_inode->i_mutex); new_dentry = lookup_one_len(new_name, new_parent, strlen(new_name)); @@ -749,14 +750,14 @@ int sysfs_rename_dir(struct kobject *kob * shadows of the same directory */ error = -EINVAL; - if (sd->s_parent->s_dentry->d_inode != new_parent->d_inode || + if (old_dentry->d_parent->d_inode != new_parent->d_inode || new_dentry->d_parent->d_inode != new_parent->d_inode || - new_dentry == sd->s_dentry) - goto out_dput; + old_dentry == new_dentry) + goto out_unlock; error = -EEXIST; if (new_dentry->d_inode) - goto out_dput; + goto out_unlock; /* rename kobject and sysfs_dirent */ error = -ENOMEM; @@ -766,9 +767,9 @@ int sysfs_rename_dir(struct kobject *kob error = kobject_set_name(kobj, "%s", new_name); if (error) - goto out_free; + goto out_drop; - kfree(sd->s_name); + dup_name = sd->s_name; sd->s_name = new_name; /* move under the new parent */ @@ -788,45 +789,58 @@ int sysfs_rename_dir(struct kobject *kob error = 0; goto out_unlock; - out_free: - kfree(dup_name); out_drop: d_drop(new_dentry); - out_dput: - dput(new_dentry); out_unlock: mutex_unlock(&new_parent->d_inode->i_mutex); + out_dput: + kfree(dup_name); + dput(new_parent); + dput(old_dentry); + dput(new_dentry); return error; } -int sysfs_move_dir(struct kobject *kobj, struct kobject *new_parent) +int sysfs_move_dir(struct kobject *kobj, struct kobject *new_parent_kobj) { - struct dentry *old_parent_dentry, *new_parent_dentry, *new_dentry; - struct sysfs_dirent *new_parent_sd, *sd; + struct sysfs_dirent *sd = kobj->sd; + struct sysfs_dirent *new_parent_sd; + struct dentry *old_parent, *new_parent = NULL; + struct dentry *old_dentry = NULL, *new_dentry = NULL; int error; - old_parent_dentry = kobj->parent ? - kobj->parent->sd->s_dentry : sysfs_mount->mnt_sb->s_root; - new_parent_dentry = new_parent ? - new_parent->sd->s_dentry : sysfs_mount->mnt_sb->s_root; + BUG_ON(!sd->s_parent); + new_parent_sd = new_parent_kobj->sd ? new_parent_kobj->sd : &sysfs_root; + + /* get dentries */ + old_dentry = sysfs_get_dentry(sd); + if (IS_ERR(old_dentry)) { + error = PTR_ERR(old_dentry); + goto out_dput; + } + old_parent = sd->s_parent->s_dentry; + + new_parent = sysfs_get_dentry(new_parent_sd); + if (IS_ERR(new_parent)) { + error = PTR_ERR(new_parent); + goto out_dput; + } - if (old_parent_dentry->d_inode == new_parent_dentry->d_inode) - return 0; /* nothing to move */ + if (old_parent->d_inode == new_parent->d_inode) { + error = 0; + goto out_dput; /* nothing to move */ + } again: - mutex_lock(&old_parent_dentry->d_inode->i_mutex); - if (!mutex_trylock(&new_parent_dentry->d_inode->i_mutex)) { - mutex_unlock(&old_parent_dentry->d_inode->i_mutex); + mutex_lock(&old_parent->d_inode->i_mutex); + if (!mutex_trylock(&new_parent->d_inode->i_mutex)) { + mutex_unlock(&old_parent->d_inode->i_mutex); goto again; } - new_parent_sd = new_parent_dentry->d_fsdata; - sd = kobj->sd; - - new_dentry = lookup_one_len(kobj->name, new_parent_dentry, - strlen(kobj->name)); + new_dentry = lookup_one_len(kobj->name, new_parent, strlen(kobj->name)); if (IS_ERR(new_dentry)) { error = PTR_ERR(new_dentry); - goto out; + goto out_unlock; } else error = 0; d_add(new_dentry, NULL); @@ -843,10 +857,14 @@ again: sysfs_link_sibling(sd); spin_unlock(&sysfs_lock); -out: - mutex_unlock(&new_parent_dentry->d_inode->i_mutex); - mutex_unlock(&old_parent_dentry->d_inode->i_mutex); + out_unlock: + mutex_unlock(&new_parent->d_inode->i_mutex); + mutex_unlock(&old_parent->d_inode->i_mutex); + out_dput: + dput(new_parent); + dput(old_dentry); + dput(new_dentry); return error; } @@ -1006,12 +1024,20 @@ static loff_t sysfs_dir_lseek(struct fil int sysfs_make_shadowed_dir(struct kobject *kobj, void * (*follow_link)(struct dentry *, struct nameidata *)) { + struct dentry *dentry; struct inode *inode; struct inode_operations *i_op; - inode = kobj->sd->s_dentry->d_inode; - if (inode->i_op != &sysfs_dir_inode_operations) + /* get dentry for @kobj->sd, dentry of a shadowed dir is pinned */ + dentry = sysfs_get_dentry(kobj->sd); + if (IS_ERR(dentry)) + return PTR_ERR(dentry); + + inode = dentry->d_inode; + if (inode->i_op != &sysfs_dir_inode_operations) { + dput(dentry); return -EINVAL; + } i_op = kmalloc(sizeof(*i_op), GFP_KERNEL); if (!i_op) @@ -1038,16 +1064,22 @@ int sysfs_make_shadowed_dir(struct kobje struct sysfs_dirent *sysfs_create_shadow_dir(struct kobject *kobj) { - struct dentry *dir = kobj->sd->s_dentry; - struct inode *inode = dir->d_inode; - struct dentry *parent = dir->d_parent; - struct sysfs_dirent *parent_sd = parent->d_fsdata; - struct dentry *shadow; + struct sysfs_dirent *parent_sd = kobj->sd->s_parent; + struct dentry *dir, *parent, *shadow; + struct inode *inode; struct sysfs_dirent *sd; + dir = sysfs_get_dentry(kobj->sd); + if (IS_ERR(dir)) { + sd = (void *)dir; + goto out; + } + parent = dir->d_parent; + + inode = dir->d_inode; sd = ERR_PTR(-EINVAL); if (!sysfs_is_shadowed_inode(inode)) - goto out; + goto out_dput; shadow = d_alloc(parent, &dir->d_name); if (!shadow) @@ -1066,15 +1098,17 @@ struct sysfs_dirent *sysfs_create_shadow d_instantiate(shadow, igrab(inode)); inc_nlink(inode); inc_nlink(parent->d_inode); - dget(shadow); /* Extra count - pin the dentry in core */ -out: - return sd; -nomem: + goto out_dput; + + nomem: dput(shadow); sd = ERR_PTR(-ENOMEM); - goto out; + out_dput: + dput(dir); + out: + return sd; } /** diff --git a/fs/sysfs/file.c b/fs/sysfs/file.c index f2e3c88..1b0c441 100644 --- a/fs/sysfs/file.c +++ b/fs/sysfs/file.c @@ -362,43 +362,22 @@ static unsigned int sysfs_poll(struct fi return POLLERR|POLLPRI; } - -static struct dentry *step_down(struct dentry *dir, const char * name) -{ - struct dentry * de; - - if (dir == NULL || dir->d_inode == NULL) - return NULL; - - mutex_lock(&dir->d_inode->i_mutex); - de = lookup_one_len(name, dir, strlen(name)); - mutex_unlock(&dir->d_inode->i_mutex); - dput(dir); - if (IS_ERR(de)) - return NULL; - if (de->d_inode == NULL) { - dput(de); - return NULL; - } - return de; -} - void sysfs_notify(struct kobject * k, char *dir, char *attr) { - struct dentry *de = k->sd->s_dentry; - if (de) - dget(de); - if (de && dir) - de = step_down(de, dir); - if (de && attr) - de = step_down(de, attr); - if (de) { - struct sysfs_dirent * sd = de->d_fsdata; - if (sd) - atomic_inc(&sd->s_event); + struct sysfs_dirent *sd = k->sd; + + spin_lock(&sysfs_lock); + + if (sd && dir) + sd = sysfs_find_dirent(sd, dir); + if (sd && attr) + sd = sysfs_find_dirent(sd, attr); + if (sd) { + atomic_inc(&sd->s_event); wake_up_interruptible(&k->poll); - dput(de); } + + spin_unlock(&sysfs_lock); } EXPORT_SYMBOL_GPL(sysfs_notify); @@ -486,30 +465,31 @@ EXPORT_SYMBOL_GPL(sysfs_add_file_to_grou */ int sysfs_update_file(struct kobject * kobj, const struct attribute * attr) { - struct dentry * dir = kobj->sd->s_dentry; - struct dentry * victim; - int res = -ENOENT; - - mutex_lock(&dir->d_inode->i_mutex); - victim = lookup_one_len(attr->name, dir, strlen(attr->name)); - if (!IS_ERR(victim)) { - /* make sure dentry is really there */ - if (victim->d_inode && - (victim->d_parent->d_inode == dir->d_inode)) { - victim->d_inode->i_mtime = CURRENT_TIME; - fsnotify_modify(victim); - res = 0; - } else - d_drop(victim); - - /** - * Drop the reference acquired from lookup_one_len() above. - */ - dput(victim); + struct sysfs_dirent *victim_sd = NULL; + struct dentry *victim = NULL; + int rc; + + rc = -ENOENT; + victim_sd = sysfs_get_dirent(kobj->sd, attr->name); + if (!victim_sd) + goto out; + + victim = sysfs_get_dentry(victim_sd); + if (IS_ERR(victim)) { + rc = PTR_ERR(victim); + victim = NULL; + goto out; } - mutex_unlock(&dir->d_inode->i_mutex); - return res; + mutex_lock(&victim->d_inode->i_mutex); + victim->d_inode->i_mtime = CURRENT_TIME; + fsnotify_modify(victim); + mutex_unlock(&victim->d_inode->i_mutex); + rc = 0; + out: + dput(victim); + sysfs_put(victim_sd); + return rc; } @@ -522,30 +502,34 @@ int sysfs_update_file(struct kobject * k */ int sysfs_chmod_file(struct kobject *kobj, struct attribute *attr, mode_t mode) { - struct dentry *dir = kobj->sd->s_dentry; - struct dentry *victim; + struct sysfs_dirent *victim_sd = NULL; + struct dentry *victim = NULL; struct inode * inode; struct iattr newattrs; - int res = -ENOENT; - - mutex_lock(&dir->d_inode->i_mutex); - victim = lookup_one_len(attr->name, dir, strlen(attr->name)); - if (!IS_ERR(victim)) { - if (victim->d_inode && - (victim->d_parent->d_inode == dir->d_inode)) { - inode = victim->d_inode; - mutex_lock(&inode->i_mutex); - newattrs.ia_mode = (mode & S_IALLUGO) | - (inode->i_mode & ~S_IALLUGO); - newattrs.ia_valid = ATTR_MODE | ATTR_CTIME; - res = notify_change(victim, &newattrs); - mutex_unlock(&inode->i_mutex); - } - dput(victim); + int rc; + + rc = -ENOENT; + victim_sd = sysfs_get_dirent(kobj->sd, attr->name); + if (!victim_sd) + goto out; + + victim = sysfs_get_dentry(victim_sd); + if (IS_ERR(victim)) { + rc = PTR_ERR(victim); + victim = NULL; + goto out; } - mutex_unlock(&dir->d_inode->i_mutex); - return res; + inode = victim->d_inode; + mutex_lock(&inode->i_mutex); + newattrs.ia_mode = (mode & S_IALLUGO) | (inode->i_mode & ~S_IALLUGO); + newattrs.ia_valid = ATTR_MODE | ATTR_CTIME; + rc = notify_change(victim, &newattrs); + mutex_unlock(&inode->i_mutex); + out: + dput(victim); + sysfs_put(victim_sd); + return rc; } EXPORT_SYMBOL_GPL(sysfs_chmod_file); diff --git a/fs/sysfs/inode.c b/fs/sysfs/inode.c index 705c03d..e71d763 100644 --- a/fs/sysfs/inode.c +++ b/fs/sysfs/inode.c @@ -190,15 +190,9 @@ void sysfs_instantiate(struct dentry *de { BUG_ON(!dentry || dentry->d_inode); - if (inode->i_state & I_NEW) { + if (inode->i_state & I_NEW) unlock_new_inode(inode); - if (dentry->d_parent && dentry->d_parent->d_inode) { - struct inode *p_inode = dentry->d_parent->d_inode; - p_inode->i_mtime = p_inode->i_ctime = CURRENT_TIME; - } - } - d_instantiate(dentry, inode); } @@ -232,7 +226,7 @@ void sysfs_drop_dentry(struct sysfs_dire /* get dentry if it's there and dput() didn't kill it yet */ dentry = dget_locked(sd->s_dentry); parent = dentry->d_parent; - } else if (sd->s_parent->s_dentry->d_inode) { + } else if (sd->s_parent->s_dentry && sd->s_parent->s_dentry->d_inode) { /* We need to update the parent even if dentry for the * victim itself doesn't exist. */ @@ -267,8 +261,9 @@ void sysfs_drop_dentry(struct sysfs_dire if (sysfs_type(sd) & SYSFS_DIR) { drop_nlink(dentry->d_inode); drop_nlink(dir); - /* XXX: unpin if directory, this will go away soon */ - dput(dentry); + /* dentries for shadowed inode are pinned, unpin */ + if (sysfs_is_shadowed_inode(dentry->d_inode)) + dput(dentry); } } diff --git a/fs/sysfs/mount.c b/fs/sysfs/mount.c index d5ce3aa..f82c345 100644 --- a/fs/sysfs/mount.c +++ b/fs/sysfs/mount.c @@ -22,7 +22,7 @@ static const struct super_operations sys .drop_inode = sysfs_delete_inode, }; -static struct sysfs_dirent sysfs_root = { +struct sysfs_dirent sysfs_root = { .s_count = ATOMIC_INIT(1), .s_flags = SYSFS_ROOT, .s_mode = S_IFDIR | S_IRWXU | S_IRUGO | S_IXUGO, diff --git a/fs/sysfs/sysfs.h b/fs/sysfs/sysfs.h index b11c448..0aa4b18 100644 --- a/fs/sysfs/sysfs.h +++ b/fs/sysfs/sysfs.h @@ -45,6 +45,7 @@ struct sysfs_dirent { #define SD_DEACTIVATED_BIAS INT_MIN extern struct vfsmount * sysfs_mount; +extern struct sysfs_dirent sysfs_root; extern struct kmem_cache *sysfs_dir_cachep; extern struct dentry *sysfs_get_dentry(struct sysfs_dirent *sd); diff --git a/include/linux/sysfs.h b/include/linux/sysfs.h index 4c43030..2f58ca1 100644 --- a/include/linux/sysfs.h +++ b/include/linux/sysfs.h @@ -81,7 +81,6 @@ struct sysfs_ops { #define SYSFS_KOBJ_ATTR 0x0004 #define SYSFS_KOBJ_BIN_ATTR 0x0008 #define SYSFS_KOBJ_LINK 0x0020 -#define SYSFS_NOT_PINNED (SYSFS_KOBJ_ATTR | SYSFS_KOBJ_BIN_ATTR | SYSFS_KOBJ_LINK) #define SYSFS_COPY_NAME (SYSFS_DIR | SYSFS_KOBJ_LINK) #define SYSFS_FLAG_MASK ~SYSFS_TYPE_MASK -- 1.4.3.4 - To unsubscribe from this list: send the line "unsubscribe linux-ide" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html