namespaces: mqueue ns: move mqueue_mnt into struct ipc_namespace

Move mqueue vfsmount plus a few tunables into the ipc_namespace struct. The CONFIG_IPC_NS boolean and the ipc_namespace struct will serve both the posix message queue namespaces and the SYSV ipc namespaces. The sysctl code will be fixed separately in patch 3. After just this patch, making a change to posix mqueue tunables always changes the values in the initial ipc namespace. Signed-off-by: Cedric Le Goater <clg@fr.ibm.com> Signed-off-by: Serge E. Hallyn <serue@us.ibm.com> Cc: Alexey Dobriyan <adobriyan@gmail.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

namespaces: mqueue ns: move mqueue_mnt into struct ipc_namespace
Move mqueue vfsmount plus a few tunables into the ipc_namespace struct. The CONFIG_IPC_NS boolean and the ipc_namespace struct will serve both the posix message queue namespaces and the SYSV ipc namespaces. The sysctl code will be fixed separately in patch 3. After just this patch, making a change to posix mqueue tunables always changes the values in the initial ipc namespace. Signed-off-by: Cedric Le Goater <clg@fr.ibm.com> Signed-off-by: Serge E. Hallyn <serue@us.ibm.com> Cc: Alexey Dobriyan <adobriyan@gmail.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Serge E. Hallyn · Linus Torvalds
1 parent 909e6d9479
Showing 7 changed files with 145 additions and 71 deletions Side-by-side Diff
include/linux/ipc_namespace.h
init/Kconfig
ipc/mqueue.c
ipc/msgutil.c
ipc/namespace.c
ipc/util.c
ipc/util.h
@@ -44,24 +44,55 @@
 	int		shm_tot;
  
 	struct notifier_block ipcns_nb;
+
+	/* The kern_mount of the mqueuefs sb.  We take a ref on it */
+	struct vfsmount	*mq_mnt;
+
+	/* # queues in this ns, protected by mq_lock */
+	unsigned int    mq_queues_count;
+
+	/* next fields are set through sysctl */
+	unsigned int    mq_queues_max;   /* initialized to DFLT_QUEUESMAX */
+	unsigned int    mq_msg_max;      /* initialized to DFLT_MSGMAX */
+	unsigned int    mq_msgsize_max;  /* initialized to DFLT_MSGSIZEMAX */
+
 };
  
 extern struct ipc_namespace init_ipc_ns;
 extern atomic_t nr_ipc_ns;
  
-#ifdef CONFIG_SYSVIPC
+#if defined(CONFIG_POSIX_MQUEUE) || defined(CONFIG_SYSVIPC)
 #define INIT_IPC_NS(ns)		.ns		= &init_ipc_ns,
+#else
+#define INIT_IPC_NS(ns)
+#endif
  
+#ifdef CONFIG_SYSVIPC
 extern int register_ipcns_notifier(struct ipc_namespace *);
 extern int cond_register_ipcns_notifier(struct ipc_namespace *);
 extern void unregister_ipcns_notifier(struct ipc_namespace *);
 extern int ipcns_notify(unsigned long);
-
 #else /* CONFIG_SYSVIPC */
-#define INIT_IPC_NS(ns)
+static inline int register_ipcns_notifier(struct ipc_namespace *ns)
+{ return 0; }
+static inline int cond_register_ipcns_notifier(struct ipc_namespace *ns)
+{ return 0; }
+static inline void unregister_ipcns_notifier(struct ipc_namespace *ns) { }
+static inline int ipcns_notify(unsigned long l) { return 0; }
 #endif /* CONFIG_SYSVIPC */
  
-#if defined(CONFIG_SYSVIPC) && defined(CONFIG_IPC_NS)
+#ifdef CONFIG_POSIX_MQUEUE
+extern void mq_init_ns(struct ipc_namespace *ns);
+/* default values */
+#define DFLT_QUEUESMAX 256     /* max number of message queues */
+#define DFLT_MSGMAX    10      /* max number of messages in each queue */
+#define HARD_MSGMAX    (131072/sizeof(void *))
+#define DFLT_MSGSIZEMAX 8192   /* max message size */
+#else
+#define mq_init_ns(ns) ((void) 0)
+#endif
+
+#if defined(CONFIG_IPC_NS)
 extern void free_ipc_ns(struct kref *kref);
 extern struct ipc_namespace *copy_ipcs(unsigned long flags,
 				       struct ipc_namespace *ns);
@@ -670,10 +670,10 @@
  
 config IPC_NS
 	bool "IPC namespace"
-	depends on NAMESPACES && SYSVIPC
+	depends on NAMESPACES && (SYSVIPC || POSIX_MQUEUE)
 	help
 	  In this namespace tasks work with IPC ids which correspond to
-	  different IPC objects in different namespaces
+	  different IPC objects in different namespaces.
  
 config USER_NS
 	bool "User namespace (EXPERIMENTAL)"
@@ -31,6 +31,7 @@
 #include <linux/mutex.h>
 #include <linux/nsproxy.h>
 #include <linux/pid.h>
+#include <linux/ipc_namespace.h>
  
 #include <net/sock.h>
 #include "util.h"
@@ -46,12 +47,6 @@
 #define STATE_PENDING	1
 #define STATE_READY	2
  
-/* default values */
-#define DFLT_QUEUESMAX	256	/* max number of message queues */
-#define DFLT_MSGMAX 	10	/* max number of messages in each queue */
-#define HARD_MSGMAX 	(131072/sizeof(void*))
-#define DFLT_MSGSIZEMAX 8192	/* max message size */
-
 /*
  * Define the ranges various user-specified maximum values can
  * be set to.
  
@@ -95,13 +90,7 @@
  
 static spinlock_t mq_lock;
 static struct kmem_cache *mqueue_inode_cachep;
-static struct vfsmount *mqueue_mnt;
  
-static unsigned int queues_count;
-static unsigned int queues_max 	= DFLT_QUEUESMAX;
-static unsigned int msg_max 	= DFLT_MSGMAX;
-static unsigned int msgsize_max = DFLT_MSGSIZEMAX;
-
 static struct ctl_table_header * mq_sysctl_table;
  
 static inline struct mqueue_inode_info *MQUEUE_I(struct inode *inode)
  
@@ -109,11 +98,27 @@
 	return container_of(inode, struct mqueue_inode_info, vfs_inode);
 }
  
+void mq_init_ns(struct ipc_namespace *ns)
+{
+	ns->mq_queues_count  = 0;
+	ns->mq_queues_max    = DFLT_QUEUESMAX;
+	ns->mq_msg_max       = DFLT_MSGMAX;
+	ns->mq_msgsize_max   = DFLT_MSGSIZEMAX;
+	ns->mq_mnt           = mntget(init_ipc_ns.mq_mnt);
+}
+
+void mq_exit_ns(struct ipc_namespace *ns)
+{
+	/* will need to clear out ns->mq_mnt->mnt_sb->s_fs_info here */
+	mntput(ns->mq_mnt);
+}
+
 static struct inode *mqueue_get_inode(struct super_block *sb, int mode,
 							struct mq_attr *attr)
 {
 	struct user_struct *u = current_user();
 	struct inode *inode;
+	struct ipc_namespace *ipc_ns = &init_ipc_ns;
  
 	inode = new_inode(sb);
 	if (inode) {
@@ -141,8 +146,8 @@
 			info->qsize = 0;
 			info->user = NULL;	/* set when all is ok */
 			memset(&info->attr, 0, sizeof(info->attr));
-			info->attr.mq_maxmsg = msg_max;
-			info->attr.mq_msgsize = msgsize_max;
+			info->attr.mq_maxmsg = ipc_ns->mq_msg_max;
+			info->attr.mq_msgsize = ipc_ns->mq_msgsize_max;
 			if (attr) {
 				info->attr.mq_maxmsg = attr->mq_maxmsg;
 				info->attr.mq_msgsize = attr->mq_msgsize;
@@ -242,6 +247,7 @@
 	struct user_struct *user;
 	unsigned long mq_bytes;
 	int i;
+	struct ipc_namespace *ipc_ns = &init_ipc_ns;
  
 	if (S_ISDIR(inode->i_mode)) {
 		clear_inode(inode);
@@ -262,7 +268,7 @@
 	if (user) {
 		spin_lock(&mq_lock);
 		user->mq_bytes -= mq_bytes;
-		queues_count--;
+		ipc_ns->mq_queues_count--;
 		spin_unlock(&mq_lock);
 		free_uid(user);
 	}
  
  
  
  
@@ -274,21 +280,23 @@
 	struct inode *inode;
 	struct mq_attr *attr = dentry->d_fsdata;
 	int error;
+	struct ipc_namespace *ipc_ns = &init_ipc_ns;
  
 	spin_lock(&mq_lock);
-	if (queues_count >= queues_max && !capable(CAP_SYS_RESOURCE)) {
+	if (ipc_ns->mq_queues_count >= ipc_ns->mq_queues_max &&
+			!capable(CAP_SYS_RESOURCE)) {
 		error = -ENOSPC;
-		goto out_lock;
+		goto out_unlock;
 	}
-	queues_count++;
+	ipc_ns->mq_queues_count++;
 	spin_unlock(&mq_lock);
  
 	inode = mqueue_get_inode(dir->i_sb, mode, attr);
 	if (!inode) {
 		error = -ENOMEM;
 		spin_lock(&mq_lock);
-		queues_count--;
-		goto out_lock;
+		ipc_ns->mq_queues_count--;
+		goto out_unlock;
 	}
  
 	dir->i_size += DIRENT_SIZE;
@@ -297,7 +305,7 @@
 	d_instantiate(dentry, inode);
 	dget(dentry);
 	return 0;
-out_lock:
+out_unlock:
 	spin_unlock(&mq_lock);
 	return error;
 }
@@ -562,7 +570,7 @@
 	info->notify_owner = NULL;
 }
  
-static int mq_attr_ok(struct mq_attr *attr)
+static int mq_attr_ok(struct ipc_namespace *ipc_ns, struct mq_attr *attr)
 {
 	if (attr->mq_maxmsg <= 0 || attr->mq_msgsize <= 0)
 		return 0;
@@ -570,8 +578,8 @@
 		if (attr->mq_maxmsg > HARD_MSGMAX)
 			return 0;
 	} else {
-		if (attr->mq_maxmsg > msg_max ||
-				attr->mq_msgsize > msgsize_max)
+		if (attr->mq_maxmsg > ipc_ns->mq_msg_max ||
+				attr->mq_msgsize > ipc_ns->mq_msgsize_max)
 			return 0;
 	}
 	/* check for overflow */
@@ -587,8 +595,9 @@
 /*
  * Invoked when creating a new queue via sys_mq_open
  */
-static struct file *do_create(struct dentry *dir, struct dentry *dentry,
-			int oflag, mode_t mode, struct mq_attr *attr)
+static struct file *do_create(struct ipc_namespace *ipc_ns, struct dentry *dir,
+			struct dentry *dentry, int oflag, mode_t mode,
+			struct mq_attr *attr)
 {
 	const struct cred *cred = current_cred();
 	struct file *result;
  
@@ -596,14 +605,14 @@
  
 	if (attr) {
 		ret = -EINVAL;
-		if (!mq_attr_ok(attr))
+		if (!mq_attr_ok(ipc_ns, attr))
 			goto out;
 		/* store for use during create */
 		dentry->d_fsdata = attr;
 	}
  
 	mode &= ~current_umask();
-	ret = mnt_want_write(mqueue_mnt);
+	ret = mnt_want_write(ipc_ns->mq_mnt);
 	if (ret)
 		goto out;
 	ret = vfs_create(dir->d_inode, dentry, mode, NULL);
  
  
  
  
@@ -611,24 +620,25 @@
 	if (ret)
 		goto out_drop_write;
  
-	result = dentry_open(dentry, mqueue_mnt, oflag, cred);
+	result = dentry_open(dentry, ipc_ns->mq_mnt, oflag, cred);
 	/*
 	 * dentry_open() took a persistent mnt_want_write(),
 	 * so we can now drop this one.
 	 */
-	mnt_drop_write(mqueue_mnt);
+	mnt_drop_write(ipc_ns->mq_mnt);
 	return result;
  
 out_drop_write:
-	mnt_drop_write(mqueue_mnt);
+	mnt_drop_write(ipc_ns->mq_mnt);
 out:
 	dput(dentry);
-	mntput(mqueue_mnt);
+	mntput(ipc_ns->mq_mnt);
 	return ERR_PTR(ret);
 }
  
 /* Opens existing queue */
-static struct file *do_open(struct dentry *dentry, int oflag)
+static struct file *do_open(struct ipc_namespace *ipc_ns,
+				struct dentry *dentry, int oflag)
 {
 	const struct cred *cred = current_cred();
  
  
  
@@ -637,17 +647,17 @@
  
 	if ((oflag & O_ACCMODE) == (O_RDWR | O_WRONLY)) {
 		dput(dentry);
-		mntput(mqueue_mnt);
+		mntput(ipc_ns->mq_mnt);
 		return ERR_PTR(-EINVAL);
 	}
  
 	if (inode_permission(dentry->d_inode, oflag2acc[oflag & O_ACCMODE])) {
 		dput(dentry);
-		mntput(mqueue_mnt);
+		mntput(ipc_ns->mq_mnt);
 		return ERR_PTR(-EACCES);
 	}
  
-	return dentry_open(dentry, mqueue_mnt, oflag, cred);
+	return dentry_open(dentry, ipc_ns->mq_mnt, oflag, cred);
 }
  
 SYSCALL_DEFINE4(mq_open, const char __user *, u_name, int, oflag, mode_t, mode,
@@ -658,6 +668,7 @@
 	char *name;
 	struct mq_attr attr;
 	int fd, error;
+	struct ipc_namespace *ipc_ns = &init_ipc_ns;
  
 	if (u_attr && copy_from_user(&attr, u_attr, sizeof(struct mq_attr)))
 		return -EFAULT;
  
@@ -671,13 +682,13 @@
 	if (fd < 0)
 		goto out_putname;
  
-	mutex_lock(&mqueue_mnt->mnt_root->d_inode->i_mutex);
-	dentry = lookup_one_len(name, mqueue_mnt->mnt_root, strlen(name));
+	mutex_lock(&ipc_ns->mq_mnt->mnt_root->d_inode->i_mutex);
+	dentry = lookup_one_len(name, ipc_ns->mq_mnt->mnt_root, strlen(name));
 	if (IS_ERR(dentry)) {
 		error = PTR_ERR(dentry);
 		goto out_err;
 	}
-	mntget(mqueue_mnt);
+	mntget(ipc_ns->mq_mnt);
  
 	if (oflag & O_CREAT) {
 		if (dentry->d_inode) {	/* entry already exists */
  
@@ -685,10 +696,10 @@
 			error = -EEXIST;
 			if (oflag & O_EXCL)
 				goto out;
-			filp = do_open(dentry, oflag);
+			filp = do_open(ipc_ns, dentry, oflag);
 		} else {
-			filp = do_create(mqueue_mnt->mnt_root, dentry,
-						oflag, mode,
+			filp = do_create(ipc_ns, ipc_ns->mq_mnt->mnt_root,
+						dentry, oflag, mode,
 						u_attr ? &attr : NULL);
 		}
 	} else {
@@ -696,7 +707,7 @@
 		if (!dentry->d_inode)
 			goto out;
 		audit_inode(name, dentry);
-		filp = do_open(dentry, oflag);
+		filp = do_open(ipc_ns, dentry, oflag);
 	}
  
 	if (IS_ERR(filp)) {
  
@@ -709,13 +720,13 @@
  
 out:
 	dput(dentry);
-	mntput(mqueue_mnt);
+	mntput(ipc_ns->mq_mnt);
 out_putfd:
 	put_unused_fd(fd);
 out_err:
 	fd = error;
 out_upsem:
-	mutex_unlock(&mqueue_mnt->mnt_root->d_inode->i_mutex);
+	mutex_unlock(&ipc_ns->mq_mnt->mnt_root->d_inode->i_mutex);
 out_putname:
 	putname(name);
 	return fd;
  
  
@@ -727,14 +738,15 @@
 	char *name;
 	struct dentry *dentry;
 	struct inode *inode = NULL;
+	struct ipc_namespace *ipc_ns = &init_ipc_ns;
  
 	name = getname(u_name);
 	if (IS_ERR(name))
 		return PTR_ERR(name);
  
-	mutex_lock_nested(&mqueue_mnt->mnt_root->d_inode->i_mutex,
+	mutex_lock_nested(&ipc_ns->mq_mnt->mnt_root->d_inode->i_mutex,
 			I_MUTEX_PARENT);
-	dentry = lookup_one_len(name, mqueue_mnt->mnt_root, strlen(name));
+	dentry = lookup_one_len(name, ipc_ns->mq_mnt->mnt_root, strlen(name));
 	if (IS_ERR(dentry)) {
 		err = PTR_ERR(dentry);
 		goto out_unlock;
  
  
@@ -748,16 +760,16 @@
 	inode = dentry->d_inode;
 	if (inode)
 		atomic_inc(&inode->i_count);
-	err = mnt_want_write(mqueue_mnt);
+	err = mnt_want_write(ipc_ns->mq_mnt);
 	if (err)
 		goto out_err;
 	err = vfs_unlink(dentry->d_parent->d_inode, dentry);
-	mnt_drop_write(mqueue_mnt);
+	mnt_drop_write(ipc_ns->mq_mnt);
 out_err:
 	dput(dentry);
  
 out_unlock:
-	mutex_unlock(&mqueue_mnt->mnt_root->d_inode->i_mutex);
+	mutex_unlock(&ipc_ns->mq_mnt->mnt_root->d_inode->i_mutex);
 	putname(name);
 	if (inode)
 		iput(inode);
  
@@ -1214,14 +1226,14 @@
 static ctl_table mq_sysctls[] = {
 	{
 		.procname	= "queues_max",
-		.data		= &queues_max,
+		.data		= &init_ipc_ns.mq_queues_max,
 		.maxlen		= sizeof(int),
 		.mode		= 0644,
 		.proc_handler	= &proc_dointvec,
 	},
 	{
 		.procname	= "msg_max",
-		.data		= &msg_max,
+		.data		= &init_ipc_ns.mq_msg_max,
 		.maxlen		= sizeof(int),
 		.mode		= 0644,
 		.proc_handler	= &proc_dointvec_minmax,
@@ -1230,7 +1242,7 @@
 	},
 	{
 		.procname	= "msgsize_max",
-		.data		= &msgsize_max,
+		.data		= &init_ipc_ns.mq_msgsize_max,
 		.maxlen		= sizeof(int),
 		.mode		= 0644,
 		.proc_handler	= &proc_dointvec_minmax,
  
@@ -1276,13 +1288,13 @@
 	if (error)
 		goto out_sysctl;
  
-	if (IS_ERR(mqueue_mnt = kern_mount(&mqueue_fs_type))) {
-		error = PTR_ERR(mqueue_mnt);
+	init_ipc_ns.mq_mnt = kern_mount(&mqueue_fs_type);
+	if (IS_ERR(init_ipc_ns.mq_mnt)) {
+		error = PTR_ERR(init_ipc_ns.mq_mnt);
 		goto out_filesystem;
 	}
  
 	/* internal initialization - not common for vfs */
-	queues_count = 0;
 	spin_lock_init(&mq_lock);
  
 	return 0;
@@ -13,9 +13,31 @@
 #include <linux/security.h>
 #include <linux/slab.h>
 #include <linux/ipc.h>
+#include <linux/ipc_namespace.h>
 #include <asm/uaccess.h>
  
 #include "util.h"
+
+/*
+ * The next 2 defines are here bc this is the only file
+ * compiled when either CONFIG_SYSVIPC and CONFIG_POSIX_MQUEUE
+ * and not CONFIG_IPC_NS.
+ */
+struct ipc_namespace init_ipc_ns = {
+	.kref = {
+		/* It's not for this patch to change, but should this be 1? */
+		.refcount	= ATOMIC_INIT(2),
+	},
+#ifdef CONFIG_POSIX_MQUEUE
+	.mq_mnt          = NULL,
+	.mq_queues_count = 0,
+	.mq_queues_max   = DFLT_QUEUESMAX,
+	.mq_msg_max      = DFLT_MSGMAX,
+	.mq_msgsize_max  = DFLT_MSGSIZEMAX,
+#endif
+};
+
+atomic_t nr_ipc_ns = ATOMIC_INIT(1);
  
 struct msg_msgseg {
 	struct msg_msgseg* next;
@@ -25,6 +25,7 @@
 	sem_init_ns(ns);
 	msg_init_ns(ns);
 	shm_init_ns(ns);
+	mq_init_ns(ns);
  
 	/*
 	 * msgmni has already been computed for the new ipc ns.
@@ -101,6 +102,7 @@
 	sem_exit_ns(ns);
 	msg_exit_ns(ns);
 	shm_exit_ns(ns);
+	mq_exit_ns(ns);
 	kfree(ns);
 	atomic_dec(&nr_ipc_ns);
  
@@ -47,15 +47,6 @@
 	int (*show)(struct seq_file *, void *);
 };
  
-struct ipc_namespace init_ipc_ns = {
-	.kref = {
-		.refcount	= ATOMIC_INIT(2),
-	},
-};
-
-atomic_t nr_ipc_ns = ATOMIC_INIT(1);
-
-
 #ifdef CONFIG_MEMORY_HOTPLUG
  
 static void ipc_memory_notifier(struct work_struct *work)
@@ -20,6 +20,13 @@
  
 struct ipc_namespace;
  
+#ifdef CONFIG_POSIX_MQUEUE
+void mq_exit_ns(struct ipc_namespace *ns);
+#else
+static inline void mq_exit_ns(struct ipc_namespace *ns) { }
+#endif
+
+#ifdef CONFIG_SYSVIPC
 void sem_init_ns(struct ipc_namespace *ns);
 void msg_init_ns(struct ipc_namespace *ns);
 void shm_init_ns(struct ipc_namespace *ns);
@@ -27,6 +34,15 @@
 void sem_exit_ns(struct ipc_namespace *ns);
 void msg_exit_ns(struct ipc_namespace *ns);
 void shm_exit_ns(struct ipc_namespace *ns);
+#else
+static inline void sem_init_ns(struct ipc_namespace *ns) { }
+static inline void msg_init_ns(struct ipc_namespace *ns) { }
+static inline void shm_init_ns(struct ipc_namespace *ns) { }
+
+static inline void sem_exit_ns(struct ipc_namespace *ns) { }
+static inline void msg_exit_ns(struct ipc_namespace *ns) { }
+static inline void shm_exit_ns(struct ipc_namespace *ns) { }
+#endif
  
 /*
  * Structure that holds the parameters needed by the ipc operations
...	...	@@ -44,24 +44,55 @@
44	44	int shm_tot;
45	45
46	46	struct notifier_block ipcns_nb;
	47	+
	48	+ /* The kern_mount of the mqueuefs sb. We take a ref on it */
	49	+ struct vfsmount *mq_mnt;
	50	+
	51	+ /* # queues in this ns, protected by mq_lock */
	52	+ unsigned int mq_queues_count;
	53	+
	54	+ /* next fields are set through sysctl */
	55	+ unsigned int mq_queues_max; /* initialized to DFLT_QUEUESMAX */
	56	+ unsigned int mq_msg_max; /* initialized to DFLT_MSGMAX */
	57	+ unsigned int mq_msgsize_max; /* initialized to DFLT_MSGSIZEMAX */
	58	+
47	59	};
48	60
49	61	extern struct ipc_namespace init_ipc_ns;
50	62	extern atomic_t nr_ipc_ns;
51	63
52		-#ifdef CONFIG_SYSVIPC
	64	+#if defined(CONFIG_POSIX_MQUEUE) \|\| defined(CONFIG_SYSVIPC)
53	65	#define INIT_IPC_NS(ns) .ns = &init_ipc_ns,
	66	+#else
	67	+#define INIT_IPC_NS(ns)
	68	+#endif
54	69
	70	+#ifdef CONFIG_SYSVIPC
55	71	extern int register_ipcns_notifier(struct ipc_namespace *);
56	72	extern int cond_register_ipcns_notifier(struct ipc_namespace *);
57	73	extern void unregister_ipcns_notifier(struct ipc_namespace *);
58	74	extern int ipcns_notify(unsigned long);
59		-
60	75	#else /* CONFIG_SYSVIPC */
61		-#define INIT_IPC_NS(ns)
	76	+static inline int register_ipcns_notifier(struct ipc_namespace *ns)
	77	+{ return 0; }
	78	+static inline int cond_register_ipcns_notifier(struct ipc_namespace *ns)
	79	+{ return 0; }
	80	+static inline void unregister_ipcns_notifier(struct ipc_namespace *ns) { }
	81	+static inline int ipcns_notify(unsigned long l) { return 0; }
62	82	#endif /* CONFIG_SYSVIPC */
63	83
64		-#if defined(CONFIG_SYSVIPC) && defined(CONFIG_IPC_NS)
	84	+#ifdef CONFIG_POSIX_MQUEUE
	85	+extern void mq_init_ns(struct ipc_namespace *ns);
	86	+/* default values */
	87	+#define DFLT_QUEUESMAX 256 /* max number of message queues */
	88	+#define DFLT_MSGMAX 10 /* max number of messages in each queue */
	89	+#define HARD_MSGMAX (131072/sizeof(void *))
	90	+#define DFLT_MSGSIZEMAX 8192 /* max message size */
	91	+#else
	92	+#define mq_init_ns(ns) ((void) 0)
	93	+#endif
	94	+
	95	+#if defined(CONFIG_IPC_NS)
65	96	extern void free_ipc_ns(struct kref *kref);
66	97	extern struct ipc_namespace *copy_ipcs(unsigned long flags,
67	98	struct ipc_namespace *ns);
...	...	@@ -670,10 +670,10 @@
670	670
671	671	config IPC_NS
672	672	bool "IPC namespace"
673		- depends on NAMESPACES && SYSVIPC
	673	+ depends on NAMESPACES && (SYSVIPC \|\| POSIX_MQUEUE)
674	674	help
675	675	In this namespace tasks work with IPC ids which correspond to
676		- different IPC objects in different namespaces
	676	+ different IPC objects in different namespaces.
677	677
678	678	config USER_NS
679	679	bool "User namespace (EXPERIMENTAL)"
...	...	@@ -31,6 +31,7 @@
31	31	#include <linux/mutex.h>
32	32	#include <linux/nsproxy.h>
33	33	#include <linux/pid.h>
	34	+#include <linux/ipc_namespace.h>
34	35
35	36	#include <net/sock.h>
36	37	#include "util.h"
...	...	@@ -46,12 +47,6 @@
46	47	#define STATE_PENDING 1
47	48	#define STATE_READY 2
48	49
49		-/* default values */
50		-#define DFLT_QUEUESMAX 256 /* max number of message queues */
51		-#define DFLT_MSGMAX 10 /* max number of messages in each queue */
52		-#define HARD_MSGMAX (131072/sizeof(void*))
53		-#define DFLT_MSGSIZEMAX 8192 /* max message size */
54		-
55	50	/*
56	51	* Define the ranges various user-specified maximum values can
57	52	* be set to.
58	53
...	...	@@ -95,13 +90,7 @@
95	90
96	91	static spinlock_t mq_lock;
97	92	static struct kmem_cache *mqueue_inode_cachep;
98		-static struct vfsmount *mqueue_mnt;
99	93
100		-static unsigned int queues_count;
101		-static unsigned int queues_max = DFLT_QUEUESMAX;
102		-static unsigned int msg_max = DFLT_MSGMAX;
103		-static unsigned int msgsize_max = DFLT_MSGSIZEMAX;
104		-
105	94	static struct ctl_table_header * mq_sysctl_table;
106	95
107	96	static inline struct mqueue_inode_info MQUEUE_I(struct inode inode)
108	97
...	...	@@ -109,11 +98,27 @@
109	98	return container_of(inode, struct mqueue_inode_info, vfs_inode);
110	99	}
111	100
	101	+void mq_init_ns(struct ipc_namespace *ns)
	102	+{
	103	+ ns->mq_queues_count = 0;
	104	+ ns->mq_queues_max = DFLT_QUEUESMAX;
	105	+ ns->mq_msg_max = DFLT_MSGMAX;
	106	+ ns->mq_msgsize_max = DFLT_MSGSIZEMAX;
	107	+ ns->mq_mnt = mntget(init_ipc_ns.mq_mnt);
	108	+}
	109	+
	110	+void mq_exit_ns(struct ipc_namespace *ns)
	111	+{
	112	+ /* will need to clear out ns->mq_mnt->mnt_sb->s_fs_info here */
	113	+ mntput(ns->mq_mnt);
	114	+}
	115	+
112	116	static struct inode mqueue_get_inode(struct super_block sb, int mode,
113	117	struct mq_attr *attr)
114	118	{
115	119	struct user_struct *u = current_user();
116	120	struct inode *inode;
	121	+ struct ipc_namespace *ipc_ns = &init_ipc_ns;
117	122
118	123	inode = new_inode(sb);
119	124	if (inode) {
...	...	@@ -141,8 +146,8 @@
141	146	info->qsize = 0;
142	147	info->user = NULL; /* set when all is ok */
143	148	memset(&info->attr, 0, sizeof(info->attr));
144		- info->attr.mq_maxmsg = msg_max;
145		- info->attr.mq_msgsize = msgsize_max;
	149	+ info->attr.mq_maxmsg = ipc_ns->mq_msg_max;
	150	+ info->attr.mq_msgsize = ipc_ns->mq_msgsize_max;
146	151	if (attr) {
147	152	info->attr.mq_maxmsg = attr->mq_maxmsg;
148	153	info->attr.mq_msgsize = attr->mq_msgsize;
...	...	@@ -242,6 +247,7 @@
242	247	struct user_struct *user;
243	248	unsigned long mq_bytes;
244	249	int i;
	250	+ struct ipc_namespace *ipc_ns = &init_ipc_ns;
245	251
246	252	if (S_ISDIR(inode->i_mode)) {
247	253	clear_inode(inode);
...	...	@@ -262,7 +268,7 @@
262	268	if (user) {
263	269	spin_lock(&mq_lock);
264	270	user->mq_bytes -= mq_bytes;
265		- queues_count--;
	271	+ ipc_ns->mq_queues_count--;
266	272	spin_unlock(&mq_lock);
267	273	free_uid(user);
268	274	}
269	275
270	276
271	277
272	278
...	...	@@ -274,21 +280,23 @@
274	280	struct inode *inode;
275	281	struct mq_attr *attr = dentry->d_fsdata;
276	282	int error;
	283	+ struct ipc_namespace *ipc_ns = &init_ipc_ns;
277	284
278	285	spin_lock(&mq_lock);
279		- if (queues_count >= queues_max && !capable(CAP_SYS_RESOURCE)) {
	286	+ if (ipc_ns->mq_queues_count >= ipc_ns->mq_queues_max &&
	287	+ !capable(CAP_SYS_RESOURCE)) {
280	288	error = -ENOSPC;
281		- goto out_lock;
	289	+ goto out_unlock;
282	290	}
283		- queues_count++;
	291	+ ipc_ns->mq_queues_count++;
284	292	spin_unlock(&mq_lock);
285	293
286	294	inode = mqueue_get_inode(dir->i_sb, mode, attr);
287	295	if (!inode) {
288	296	error = -ENOMEM;
289	297	spin_lock(&mq_lock);
290		- queues_count--;
291		- goto out_lock;
	298	+ ipc_ns->mq_queues_count--;
	299	+ goto out_unlock;
292	300	}
293	301
294	302	dir->i_size += DIRENT_SIZE;
...	...	@@ -297,7 +305,7 @@
297	305	d_instantiate(dentry, inode);
298	306	dget(dentry);
299	307	return 0;
300		-out_lock:
	308	+out_unlock:
301	309	spin_unlock(&mq_lock);
302	310	return error;
303	311	}
...	...	@@ -562,7 +570,7 @@
562	570	info->notify_owner = NULL;
563	571	}
564	572
565		-static int mq_attr_ok(struct mq_attr *attr)
	573	+static int mq_attr_ok(struct ipc_namespace ipc_ns, struct mq_attr attr)
566	574	{
567	575	if (attr->mq_maxmsg <= 0 \|\| attr->mq_msgsize <= 0)
568	576	return 0;
...	...	@@ -570,8 +578,8 @@
570	578	if (attr->mq_maxmsg > HARD_MSGMAX)
571	579	return 0;
572	580	} else {
573		- if (attr->mq_maxmsg > msg_max \|\|
574		- attr->mq_msgsize > msgsize_max)
	581	+ if (attr->mq_maxmsg > ipc_ns->mq_msg_max \|\|
	582	+ attr->mq_msgsize > ipc_ns->mq_msgsize_max)
575	583	return 0;
576	584	}
577	585	/* check for overflow */
...	...	@@ -587,8 +595,9 @@
587	595	/*
588	596	* Invoked when creating a new queue via sys_mq_open
589	597	*/
590		-static struct file do_create(struct dentry dir, struct dentry *dentry,
591		- int oflag, mode_t mode, struct mq_attr *attr)
	598	+static struct file do_create(struct ipc_namespace ipc_ns, struct dentry *dir,
	599	+ struct dentry *dentry, int oflag, mode_t mode,
	600	+ struct mq_attr *attr)
592	601	{
593	602	const struct cred *cred = current_cred();
594	603	struct file *result;
595	604
...	...	@@ -596,14 +605,14 @@
596	605
597	606	if (attr) {
598	607	ret = -EINVAL;
599		- if (!mq_attr_ok(attr))
	608	+ if (!mq_attr_ok(ipc_ns, attr))
600	609	goto out;
601	610	/* store for use during create */
602	611	dentry->d_fsdata = attr;
603	612	}
604	613
605	614	mode &= ~current_umask();
606		- ret = mnt_want_write(mqueue_mnt);
	615	+ ret = mnt_want_write(ipc_ns->mq_mnt);
607	616	if (ret)
608	617	goto out;
609	618	ret = vfs_create(dir->d_inode, dentry, mode, NULL);
610	619
611	620
612	621
613	622
...	...	@@ -611,24 +620,25 @@
611	620	if (ret)
612	621	goto out_drop_write;
613	622
614		- result = dentry_open(dentry, mqueue_mnt, oflag, cred);
	623	+ result = dentry_open(dentry, ipc_ns->mq_mnt, oflag, cred);
615	624	/*
616	625	* dentry_open() took a persistent mnt_want_write(),
617	626	* so we can now drop this one.
618	627	*/
619		- mnt_drop_write(mqueue_mnt);
	628	+ mnt_drop_write(ipc_ns->mq_mnt);
620	629	return result;
621	630
622	631	out_drop_write:
623		- mnt_drop_write(mqueue_mnt);
	632	+ mnt_drop_write(ipc_ns->mq_mnt);
624	633	out:
625	634	dput(dentry);
626		- mntput(mqueue_mnt);
	635	+ mntput(ipc_ns->mq_mnt);
627	636	return ERR_PTR(ret);
628	637	}
629	638
630	639	/* Opens existing queue */
631		-static struct file do_open(struct dentry dentry, int oflag)
	640	+static struct file do_open(struct ipc_namespace ipc_ns,
	641	+ struct dentry *dentry, int oflag)
632	642	{
633	643	const struct cred *cred = current_cred();
634	644
635	645
636	646
...	...	@@ -637,17 +647,17 @@
637	647
638	648	if ((oflag & O_ACCMODE) == (O_RDWR \| O_WRONLY)) {
639	649	dput(dentry);
640		- mntput(mqueue_mnt);
	650	+ mntput(ipc_ns->mq_mnt);
641	651	return ERR_PTR(-EINVAL);
642	652	}
643	653
644	654	if (inode_permission(dentry->d_inode, oflag2acc[oflag & O_ACCMODE])) {
645	655	dput(dentry);
646		- mntput(mqueue_mnt);
	656	+ mntput(ipc_ns->mq_mnt);
647	657	return ERR_PTR(-EACCES);
648	658	}
649	659
650		- return dentry_open(dentry, mqueue_mnt, oflag, cred);
	660	+ return dentry_open(dentry, ipc_ns->mq_mnt, oflag, cred);
651	661	}
652	662
653	663	SYSCALL_DEFINE4(mq_open, const char __user *, u_name, int, oflag, mode_t, mode,
...	...	@@ -658,6 +668,7 @@
658	668	char *name;
659	669	struct mq_attr attr;
660	670	int fd, error;
	671	+ struct ipc_namespace *ipc_ns = &init_ipc_ns;
661	672
662	673	if (u_attr && copy_from_user(&attr, u_attr, sizeof(struct mq_attr)))
663	674	return -EFAULT;
664	675
...	...	@@ -671,13 +682,13 @@
671	682	if (fd < 0)
672	683	goto out_putname;
673	684
674		- mutex_lock(&mqueue_mnt->mnt_root->d_inode->i_mutex);
675		- dentry = lookup_one_len(name, mqueue_mnt->mnt_root, strlen(name));
	685	+ mutex_lock(&ipc_ns->mq_mnt->mnt_root->d_inode->i_mutex);
	686	+ dentry = lookup_one_len(name, ipc_ns->mq_mnt->mnt_root, strlen(name));
676	687	if (IS_ERR(dentry)) {
677	688	error = PTR_ERR(dentry);
678	689	goto out_err;
679	690	}
680		- mntget(mqueue_mnt);
	691	+ mntget(ipc_ns->mq_mnt);
681	692
682	693	if (oflag & O_CREAT) {
683	694	if (dentry->d_inode) { /* entry already exists */
684	695
...	...	@@ -685,10 +696,10 @@
685	696	error = -EEXIST;
686	697	if (oflag & O_EXCL)
687	698	goto out;
688		- filp = do_open(dentry, oflag);
	699	+ filp = do_open(ipc_ns, dentry, oflag);
689	700	} else {
690		- filp = do_create(mqueue_mnt->mnt_root, dentry,
691		- oflag, mode,
	701	+ filp = do_create(ipc_ns, ipc_ns->mq_mnt->mnt_root,
	702	+ dentry, oflag, mode,
692	703	u_attr ? &attr : NULL);
693	704	}
694	705	} else {
...	...	@@ -696,7 +707,7 @@
696	707	if (!dentry->d_inode)
697	708	goto out;
698	709	audit_inode(name, dentry);
699		- filp = do_open(dentry, oflag);
	710	+ filp = do_open(ipc_ns, dentry, oflag);
700	711	}
701	712
702	713	if (IS_ERR(filp)) {
703	714
...	...	@@ -709,13 +720,13 @@
709	720
710	721	out:
711	722	dput(dentry);
712		- mntput(mqueue_mnt);
	723	+ mntput(ipc_ns->mq_mnt);
713	724	out_putfd:
714	725	put_unused_fd(fd);
715	726	out_err:
716	727	fd = error;
717	728	out_upsem:
718		- mutex_unlock(&mqueue_mnt->mnt_root->d_inode->i_mutex);
	729	+ mutex_unlock(&ipc_ns->mq_mnt->mnt_root->d_inode->i_mutex);
719	730	out_putname:
720	731	putname(name);
721	732	return fd;
722	733
723	734
...	...	@@ -727,14 +738,15 @@
727	738	char *name;
728	739	struct dentry *dentry;
729	740	struct inode *inode = NULL;
	741	+ struct ipc_namespace *ipc_ns = &init_ipc_ns;
730	742
731	743	name = getname(u_name);
732	744	if (IS_ERR(name))
733	745	return PTR_ERR(name);
734	746
735		- mutex_lock_nested(&mqueue_mnt->mnt_root->d_inode->i_mutex,
	747	+ mutex_lock_nested(&ipc_ns->mq_mnt->mnt_root->d_inode->i_mutex,
736	748	I_MUTEX_PARENT);
737		- dentry = lookup_one_len(name, mqueue_mnt->mnt_root, strlen(name));
	749	+ dentry = lookup_one_len(name, ipc_ns->mq_mnt->mnt_root, strlen(name));
738	750	if (IS_ERR(dentry)) {
739	751	err = PTR_ERR(dentry);
740	752	goto out_unlock;
741	753
742	754
...	...	@@ -748,16 +760,16 @@
748	760	inode = dentry->d_inode;
749	761	if (inode)
750	762	atomic_inc(&inode->i_count);
751		- err = mnt_want_write(mqueue_mnt);
	763	+ err = mnt_want_write(ipc_ns->mq_mnt);
752	764	if (err)
753	765	goto out_err;
754	766	err = vfs_unlink(dentry->d_parent->d_inode, dentry);
755		- mnt_drop_write(mqueue_mnt);
	767	+ mnt_drop_write(ipc_ns->mq_mnt);
756	768	out_err:
757	769	dput(dentry);
758	770
759	771	out_unlock:
760		- mutex_unlock(&mqueue_mnt->mnt_root->d_inode->i_mutex);
	772	+ mutex_unlock(&ipc_ns->mq_mnt->mnt_root->d_inode->i_mutex);
761	773	putname(name);
762	774	if (inode)
763	775	iput(inode);
764	776
...	...	@@ -1214,14 +1226,14 @@
1214	1226	static ctl_table mq_sysctls[] = {
1215	1227	{
1216	1228	.procname = "queues_max",
1217		- .data = &queues_max,
	1229	+ .data = &init_ipc_ns.mq_queues_max,
1218	1230	.maxlen = sizeof(int),
1219	1231	.mode = 0644,
1220	1232	.proc_handler = &proc_dointvec,
1221	1233	},
1222	1234	{
1223	1235	.procname = "msg_max",
1224		- .data = &msg_max,
	1236	+ .data = &init_ipc_ns.mq_msg_max,
1225	1237	.maxlen = sizeof(int),
1226	1238	.mode = 0644,
1227	1239	.proc_handler = &proc_dointvec_minmax,
...	...	@@ -1230,7 +1242,7 @@
1230	1242	},
1231	1243	{
1232	1244	.procname = "msgsize_max",
1233		- .data = &msgsize_max,
	1245	+ .data = &init_ipc_ns.mq_msgsize_max,
1234	1246	.maxlen = sizeof(int),
1235	1247	.mode = 0644,
1236	1248	.proc_handler = &proc_dointvec_minmax,
1237	1249
...	...	@@ -1276,13 +1288,13 @@
1276	1288	if (error)
1277	1289	goto out_sysctl;
1278	1290
1279		- if (IS_ERR(mqueue_mnt = kern_mount(&mqueue_fs_type))) {
1280		- error = PTR_ERR(mqueue_mnt);
	1291	+ init_ipc_ns.mq_mnt = kern_mount(&mqueue_fs_type);
	1292	+ if (IS_ERR(init_ipc_ns.mq_mnt)) {
	1293	+ error = PTR_ERR(init_ipc_ns.mq_mnt);
1281	1294	goto out_filesystem;
1282	1295	}
1283	1296
1284	1297	/* internal initialization - not common for vfs */
1285		- queues_count = 0;
1286	1298	spin_lock_init(&mq_lock);
1287	1299
1288	1300	return 0;
...	...	@@ -13,9 +13,31 @@
13	13	#include <linux/security.h>
14	14	#include <linux/slab.h>
15	15	#include <linux/ipc.h>
	16	+#include <linux/ipc_namespace.h>
16	17	#include <asm/uaccess.h>
17	18
18	19	#include "util.h"
	20	+
	21	+/*
	22	+ * The next 2 defines are here bc this is the only file
	23	+ * compiled when either CONFIG_SYSVIPC and CONFIG_POSIX_MQUEUE
	24	+ * and not CONFIG_IPC_NS.
	25	+ */
	26	+struct ipc_namespace init_ipc_ns = {
	27	+ .kref = {
	28	+ /* It's not for this patch to change, but should this be 1? */
	29	+ .refcount = ATOMIC_INIT(2),
	30	+ },
	31	+#ifdef CONFIG_POSIX_MQUEUE
	32	+ .mq_mnt = NULL,
	33	+ .mq_queues_count = 0,
	34	+ .mq_queues_max = DFLT_QUEUESMAX,
	35	+ .mq_msg_max = DFLT_MSGMAX,
	36	+ .mq_msgsize_max = DFLT_MSGSIZEMAX,
	37	+#endif
	38	+};
	39	+
	40	+atomic_t nr_ipc_ns = ATOMIC_INIT(1);
19	41
20	42	struct msg_msgseg {
21	43	struct msg_msgseg* next;
...	...	@@ -25,6 +25,7 @@
25	25	sem_init_ns(ns);
26	26	msg_init_ns(ns);
27	27	shm_init_ns(ns);
	28	+ mq_init_ns(ns);
28	29
29	30	/*
30	31	* msgmni has already been computed for the new ipc ns.
...	...	@@ -101,6 +102,7 @@
101	102	sem_exit_ns(ns);
102	103	msg_exit_ns(ns);
103	104	shm_exit_ns(ns);
	105	+ mq_exit_ns(ns);
104	106	kfree(ns);
105	107	atomic_dec(&nr_ipc_ns);
106	108
...	...	@@ -47,15 +47,6 @@
47	47	int (show)(struct seq_file , void *);
48	48	};
49	49
50		-struct ipc_namespace init_ipc_ns = {
51		- .kref = {
52		- .refcount = ATOMIC_INIT(2),
53		- },
54		-};
55		-
56		-atomic_t nr_ipc_ns = ATOMIC_INIT(1);
57		-
58		-
59	50	#ifdef CONFIG_MEMORY_HOTPLUG
60	51
61	52	static void ipc_memory_notifier(struct work_struct *work)
...	...	@@ -20,6 +20,13 @@
20	20
21	21	struct ipc_namespace;
22	22
	23	+#ifdef CONFIG_POSIX_MQUEUE
	24	+void mq_exit_ns(struct ipc_namespace *ns);
	25	+#else
	26	+static inline void mq_exit_ns(struct ipc_namespace *ns) { }
	27	+#endif
	28	+
	29	+#ifdef CONFIG_SYSVIPC
23	30	void sem_init_ns(struct ipc_namespace *ns);
24	31	void msg_init_ns(struct ipc_namespace *ns);
25	32	void shm_init_ns(struct ipc_namespace *ns);
...	...	@@ -27,6 +34,15 @@
27	34	void sem_exit_ns(struct ipc_namespace *ns);
28	35	void msg_exit_ns(struct ipc_namespace *ns);
29	36	void shm_exit_ns(struct ipc_namespace *ns);
	37	+#else
	38	+static inline void sem_init_ns(struct ipc_namespace *ns) { }
	39	+static inline void msg_init_ns(struct ipc_namespace *ns) { }
	40	+static inline void shm_init_ns(struct ipc_namespace *ns) { }
	41	+
	42	+static inline void sem_exit_ns(struct ipc_namespace *ns) { }
	43	+static inline void msg_exit_ns(struct ipc_namespace *ns) { }
	44	+static inline void shm_exit_ns(struct ipc_namespace *ns) { }
	45	+#endif
30	46
31	47	/*
32	48	* Structure that holds the parameters needed by the ipc operations