Re: [PATCH 1/4] Support named cgroups hierarchies

[Dhaval Giani <dhaval@linux.vnet.ibm.com>]

On Tue, Jul 28, 2009 at 04:26:21PM -0700, Paul Menage wrote:
> Support named cgroups hierarchies
> 
> To simplify referring to cgroup hierarchies in mount statements, and
> to allow disambiguation in the presence of empty hierarchies and
> multiply-bindable subsystems this patch adds support for naming a new
> cgroup hierarchy via the "name=" mount option
> 
> A pre-existing hierarchy may be specified by either name or by
> subsystems; a hierarchy's name cannot be changed by a remount
> operation.
> 
> Example usage:
> 
> # To create a hierarchy called "foo" containing the "cpu" subsystem
> mount -t cgroup -oname=foo,cpu cgroup /mnt/cgroup1
> 
> # To mount the "foo" hierarchy on a second location
> mount -t cgroup -oname=foo cgroup /mnt/cgroup2
> 
> 
> Signed-off-by: Paul Menage <menage@google.com>
> Reviewed-by: Li Zefan <lizf@cn.fujitsu.com>
> 
> ---
> 
>  Documentation/cgroups/cgroups.txt |   20 ++++
>  kernel/cgroup.c                   |  185 +++++++++++++++++++++++++++----------
>  2 files changed, 157 insertions(+), 48 deletions(-)
> 
> diff --git a/Documentation/cgroups/cgroups.txt b/Documentation/cgroups/cgroups.txt
> index 6eb1a97..4bccfc1 100644
> --- a/Documentation/cgroups/cgroups.txt
> +++ b/Documentation/cgroups/cgroups.txt
> @@ -408,6 +408,26 @@ You can attach the current shell task by echoing 0:
> 
>  # echo 0 > tasks
> 
> +2.3 Mounting hierarchies by name
> +--------------------------------
> +
> +Passing the name=<x> option when mounting a cgroups hierarchy
> +associates the given name with the hierarchy.  This can be used when
> +mounting a pre-existing hierarchy, in order to refer to it by name
> +rather than by its set of active subsystems.  Each hierarchy is either
> +nameless, or has a unique name.
> +
> +The name should match [\w.-]+
> +
> +When passing a name=<x> option for a new hierarchy, you need to
> +specify subsystems manually; the legacy behaviour of mounting all
> +subsystems when none are explicitly specified is not supported when
> +you give a subsystem a name.
> +
> +The name of the subsystem appears as part of the hierarchy description
> +in /proc/mounts and /proc/<pid>/cgroups.
> +
> +
>  3. Kernel API
>  =============
> 
> diff --git a/kernel/cgroup.c b/kernel/cgroup.c
> index 18acba7..85573e8 100644
> --- a/kernel/cgroup.c
> +++ b/kernel/cgroup.c
> @@ -23,6 +23,7 @@
>   */
> 
>  #include <linux/cgroup.h>
> +#include <linux/ctype.h>
>  #include <linux/errno.h>
>  #include <linux/fs.h>
>  #include <linux/kernel.h>
> @@ -60,6 +61,8 @@ static struct cgroup_subsys *subsys[] = {
>  #include <linux/cgroup_subsys.h>
>  };
> 
> +#define MAX_CGROUP_ROOT_NAMELEN 64
> +
>  /*
>   * A cgroupfs_root represents the root of a cgroup hierarchy,
>   * and may be associated with a superblock to form an active
> @@ -94,6 +97,9 @@ struct cgroupfs_root {
> 
>  	/* The path to use for release notifications. */
>  	char release_agent_path[PATH_MAX];
> +
> +	/* The name for this hierarchy - may be empty */
> +	char name[MAX_CGROUP_ROOT_NAMELEN];
>  };
> 
>  /*
> @@ -829,6 +835,8 @@ static int cgroup_show_options(struct seq_file *seq, struct vfsmount *vfs)
>  		seq_puts(seq, ",noprefix");
>  	if (strlen(root->release_agent_path))
>  		seq_printf(seq, ",release_agent=%s", root->release_agent_path);
> +	if (strlen(root->name))
> +		seq_printf(seq, ",name=%s", root->name);
>  	mutex_unlock(&cgroup_mutex);
>  	return 0;
>  }
> @@ -837,6 +845,9 @@ struct cgroup_sb_opts {
>  	unsigned long subsys_bits;
>  	unsigned long flags;
>  	char *release_agent;
> +	char *name;
> +
> +	struct cgroupfs_root *new_root;
>  };
> 
>  /* Convert a hierarchy specifier into a bitmask of subsystems and
> @@ -851,9 +862,7 @@ static int parse_cgroupfs_options(char *data,
>  	mask = ~(1UL << cpuset_subsys_id);
>  #endif
> 
> -	opts->subsys_bits = 0;
> -	opts->flags = 0;
> -	opts->release_agent = NULL;
> +	memset(opts, 0, sizeof(*opts));
> 
>  	while ((token = strsep(&o, ",")) != NULL) {
>  		if (!*token)
> @@ -873,11 +882,33 @@ static int parse_cgroupfs_options(char *data,
>  			/* Specifying two release agents is forbidden */
>  			if (opts->release_agent)
>  				return -EINVAL;
> -			opts->release_agent = kzalloc(PATH_MAX, GFP_KERNEL);
> +			opts->release_agent =
> +				kstrndup(token + 14, PATH_MAX, GFP_KERNEL);

I am not sure how it can be acheived, but can we avoid using 14 here (it
took me a moment before I realized it was strlen("release_agent")

>  			if (!opts->release_agent)
>  				return -ENOMEM;
> -			strncpy(opts->release_agent, token + 14, PATH_MAX - 1);
> -			opts->release_agent[PATH_MAX - 1] = 0;
> +		} else if (!strncmp(token, "name=", 5)) {
> +			int i;
> +			const char *name = token + 5;

similarly here as well

> +			/* Can't specify an empty name */
> +			if (!strlen(name))
> +				return -EINVAL;
> +			/* Must match [\w.-]+ */
> +			for (i = 0; i < strlen(name); i++) {
> +				char c = name[i];
> +				if (isalnum(c))
> +					continue;
> +				if ((c == '.') || (c == '-') || (c == '_'))
> +					continue;
> +				return -EINVAL;
> +			}
> +			/* Specifying two names is forbidden */
> +			if (opts->name)
> +				return -EINVAL;
> +			opts->name = kstrndup(name,
> +					      MAX_CGROUP_ROOT_NAMELEN,
> +					      GFP_KERNEL);
> +			if (!opts->name)
> +				return -ENOMEM;
>  		} else {
>  			struct cgroup_subsys *ss;
>  			int i;
> @@ -904,7 +935,7 @@ static int parse_cgroupfs_options(char *data,
>  		return -EINVAL;
> 
>  	/* We can't have an empty hierarchy */
> -	if (!opts->subsys_bits)
> +	if (!opts->subsys_bits && !opts->name)
>  		return -EINVAL;
> 
>  	return 0;
> @@ -932,6 +963,12 @@ static int cgroup_remount(struct super_block *sb, int *flags, char *data)
>  		goto out_unlock;
>  	}
> 
> +	/* Don't allow name to change at remount */
> +	if (opts.name && strcmp(opts.name, root->name)) {
> +		ret = -EINVAL;
> +		goto out_unlock;
> +	}
> +
>  	ret = rebind_subsystems(root, opts.subsys_bits);
>  	if (ret)
>  		goto out_unlock;
> @@ -943,6 +980,7 @@ static int cgroup_remount(struct super_block *sb, int *flags, char *data)
>  		strcpy(root->release_agent_path, opts.release_agent);
>   out_unlock:
>  	kfree(opts.release_agent);
> +	kfree(opts.name);
>  	mutex_unlock(&cgroup_mutex);
>  	mutex_unlock(&cgrp->dentry->d_inode->i_mutex);
>  	unlock_kernel();
> @@ -965,6 +1003,7 @@ static void init_cgroup_housekeeping(struct cgroup *cgrp)
>  	INIT_LIST_HEAD(&cgrp->pids_list);
>  	init_rwsem(&cgrp->pids_mutex);
>  }
> +
>  static void init_cgroup_root(struct cgroupfs_root *root)
>  {
>  	struct cgroup *cgrp = &root->top_cgroup;
> @@ -978,31 +1017,59 @@ static void init_cgroup_root(struct cgroupfs_root *root)
> 
>  static int cgroup_test_super(struct super_block *sb, void *data)
>  {
> -	struct cgroupfs_root *new = data;
> +	struct cgroup_sb_opts *opts = data;
>  	struct cgroupfs_root *root = sb->s_fs_info;
> 
> -	/* First check subsystems */
> -	if (new->subsys_bits != root->subsys_bits)
> -	    return 0;
> +	/* If we asked for a name then it must match */
> +	if (opts->name && strcmp(opts->name, root->name))
> +		return 0;
> 
> -	/* Next check flags */
> -	if (new->flags != root->flags)
> +	/* If we asked for subsystems then they must match */
> +	if (opts->subsys_bits && (opts->subsys_bits != root->subsys_bits))
>  		return 0;
> 
>  	return 1;
>  }
> 
> +static struct cgroupfs_root *cgroup_root_from_opts(struct cgroup_sb_opts *opts)
> +{
> +	struct cgroupfs_root *root;
> +
> +	/* Empty hierarchies aren't supported */
> +	if (!opts->subsys_bits)
> +		return NULL;
> +
> +	root = kzalloc(sizeof(*root), GFP_KERNEL);
> +	if (!root)
> +		return ERR_PTR(-ENOMEM);
> +
> +	init_cgroup_root(root);
> +	root->subsys_bits = opts->subsys_bits;
> +	root->flags = opts->flags;
> +	if (opts->release_agent)
> +		strcpy(root->release_agent_path, opts->release_agent);
> +	if (opts->name)
> +		strcpy(root->name, opts->name);
> +	return root;
> +}
> +
>  static int cgroup_set_super(struct super_block *sb, void *data)
>  {
>  	int ret;
> -	struct cgroupfs_root *root = data;
> +	struct cgroup_sb_opts *opts = data;
> +
> +	/* If we don't have a new root, we can't set up a new sb */
> +	if (!opts->new_root)
> +		return -EINVAL;
> +
> +	BUG_ON(!opts->subsys_bits);
> 
>  	ret = set_anon_super(sb, NULL);
>  	if (ret)
>  		return ret;
> 
> -	sb->s_fs_info = root;
> -	root->sb = sb;
> +	sb->s_fs_info = opts->new_root;
> +	opts->new_root->sb = sb;
> 
>  	sb->s_blocksize = PAGE_CACHE_SIZE;
>  	sb->s_blocksize_bits = PAGE_CACHE_SHIFT;
> @@ -1039,48 +1106,43 @@ static int cgroup_get_sb(struct file_system_type *fs_type,
>  			 void *data, struct vfsmount *mnt)
>  {
>  	struct cgroup_sb_opts opts;
> +	struct cgroupfs_root *root;
>  	int ret = 0;
>  	struct super_block *sb;
> -	struct cgroupfs_root *root;
> -	struct list_head tmp_cg_links;
> +	struct cgroupfs_root *new_root;
> 
>  	/* First find the desired set of subsystems */
>  	ret = parse_cgroupfs_options(data, &opts);
> -	if (ret) {
> -		kfree(opts.release_agent);
> -		return ret;
> -	}
> -
> -	root = kzalloc(sizeof(*root), GFP_KERNEL);
> -	if (!root) {
> -		kfree(opts.release_agent);
> -		return -ENOMEM;
> -	}
> +	if (ret)
> +		goto out_err;
> 
> -	init_cgroup_root(root);
> -	root->subsys_bits = opts.subsys_bits;
> -	root->flags = opts.flags;
> -	if (opts.release_agent) {
> -		strcpy(root->release_agent_path, opts.release_agent);
> -		kfree(opts.release_agent);
> +	/*
> +	 * Allocate a new cgroup root. We may not need it if we're
> +	 * reusing an existing hierarchy.
> +	 */
> +	new_root = cgroup_root_from_opts(&opts);
> +	if (IS_ERR(new_root)) {
> +		ret = PTR_ERR(new_root);
> +		goto out_err;
>  	}
> +	opts.new_root = new_root;
> 
> -	sb = sget(fs_type, cgroup_test_super, cgroup_set_super, root);
> -
> +	/* Locate an existing or new sb for this hierarchy */
> +	sb = sget(fs_type, cgroup_test_super, cgroup_set_super, &opts);
>  	if (IS_ERR(sb)) {
> -		kfree(root);
> -		return PTR_ERR(sb);
> +		ret = PTR_ERR(sb);
> +		kfree(opts.new_root);
> +		goto out_err;
>  	}
> 
> -	if (sb->s_fs_info != root) {
> -		/* Reusing an existing superblock */
> -		BUG_ON(sb->s_root == NULL);
> -		kfree(root);
> -		root = NULL;
> -	} else {
> -		/* New superblock */
> +	root = sb->s_fs_info;
> +	BUG_ON(!root);
> +	if (root == opts.new_root) {
> +		/* We used the new root structure, so this is a new hierarchy */
> +		struct list_head tmp_cg_links;
>  		struct cgroup *root_cgrp = &root->top_cgroup;
>  		struct inode *inode;
> +		struct cgroupfs_root *existing_root;
>  		int i;
> 
>  		BUG_ON(sb->s_root != NULL);
> @@ -1093,6 +1155,18 @@ static int cgroup_get_sb(struct file_system_type *fs_type,
>  		mutex_lock(&inode->i_mutex);
>  		mutex_lock(&cgroup_mutex);
> 
> +		if (strlen(root->name)) {
> +			/* Check for name clashes with existing mounts */
> +			for_each_active_root(existing_root) {
> +				if (!strcmp(existing_root->name, root->name)) {
> +					ret = -EBUSY;
> +					mutex_unlock(&cgroup_mutex);
> +					mutex_unlock(&inode->i_mutex);
> +					goto drop_new_super;
> +				}
> +			}
> +		}
> +
>  		/*
>  		 * We're accessing css_set_count without locking
>  		 * css_set_lock here, but that's OK - it can only be
> @@ -1111,7 +1185,8 @@ static int cgroup_get_sb(struct file_system_type *fs_type,
>  		if (ret == -EBUSY) {
>  			mutex_unlock(&cgroup_mutex);
>  			mutex_unlock(&inode->i_mutex);
> -			goto free_cg_links;
> +			free_cg_links(&tmp_cg_links);
> +			goto drop_new_super;
>  		}
> 
>  		/* EBUSY should be the only error here */
> @@ -1145,15 +1220,26 @@ static int cgroup_get_sb(struct file_system_type *fs_type,
>  		cgroup_populate_dir(root_cgrp);
>  		mutex_unlock(&inode->i_mutex);
>  		mutex_unlock(&cgroup_mutex);
> +	} else {
> +		/*
> +		 * We re-used an existing hierarchy - the new root (if
> +		 * any) is not needed
> +		 */
> +		kfree(opts.new_root);
>  	}
> 
>  	simple_set_mnt(mnt, sb);
> +	kfree(opts.release_agent);
> +	kfree(opts.name);
>  	return 0;
> 
> - free_cg_links:
> -	free_cg_links(&tmp_cg_links);
>   drop_new_super:
>  	deactivate_locked_super(sb);
> +
> + out_err:
> +	kfree(opts.release_agent);
> +	kfree(opts.name);
> +
>  	return ret;
>  }
> 
> @@ -2971,6 +3057,9 @@ static int proc_cgroup_show(struct seq_file *m, void *v)
>  		seq_printf(m, "%lu:", root->subsys_bits);
>  		for_each_subsys(root, ss)
>  			seq_printf(m, "%s%s", count++ ? "," : "", ss->name);
> +		if (strlen(root->name))
> +			seq_printf(m, "%sname=%s", count ? "," : "",
> +				   root->name);
>  		seq_putc(m, ':');
>  		get_first_subsys(&root->top_cgroup, NULL, &subsys_id);
>  		cgrp = task_cgroup(tsk, subsys_id);
> 

-- 
regards,
Dhaval