userns: Make the count of user namespaces per user

Add a structure that is per user and per user ns and use it to hold
the count of user namespaces.  This makes prevents one user from
creating denying service to another user by creating the maximum
number of user namespaces.

Rename the sysctl export of the maximum count from
/proc/sys/userns/max_user_namespaces to /proc/sys/user/max_user_namespaces
to reflect that the count is now per user.

Signed-off-by: "Eric W. Biederman" <ebiederm@xmission.com>

4 files changed, 117 insertions, 30 deletions

diff --git a/include/linux/user_namespace.h b/include/linux/user_namespace.h
index 6421cca2daa9..826de7a12a20 100644
--- a/include/linux/user_namespace.h
+++ b/include/linux/user_namespace.h
@@ -22,6 +22,7 @@ struct uid_gid_map {	/* 64 bytes -- 1 cache line */
 
 #define USERNS_INIT_FLAGS USERNS_SETGROUPS_ALLOWED
 
+struct ucounts;
 struct user_namespace {
         struct uid_gid_map      uid_map;
         struct uid_gid_map      gid_map;
@@ -44,15 +45,24 @@ struct user_namespace {
         struct ctl_table_set    set;
         struct ctl_table_header *sysctls;
 #endif
+        struct ucounts          *ucounts;
         int max_user_namespaces;
+};
+
+struct ucounts {
+        struct hlist_node node;
+        struct user_namespace *ns;
+        kuid_t uid;
+        atomic_t count;
         atomic_t user_namespaces;
 };
 
 extern struct user_namespace init_user_ns;
-extern bool setup_userns_sysctls(struct user_namespace *ns);
-extern void retire_userns_sysctls(struct user_namespace *ns);
-extern bool inc_user_namespaces(struct user_namespace *ns);
-extern void dec_user_namespaces(struct user_namespace *ns);
+
+bool setup_userns_sysctls(struct user_namespace *ns);
+void retire_userns_sysctls(struct user_namespace *ns);
+struct ucounts *inc_user_namespaces(struct user_namespace *ns, kuid_t uid);
+void dec_user_namespaces(struct ucounts *ucounts);
 
 #ifdef CONFIG_USER_NS
 
diff --git a/kernel/fork.c b/kernel/fork.c
index daa6a82b4900..d8cde533ace3 100644
--- a/kernel/fork.c
+++ b/kernel/fork.c
@@ -322,7 +322,7 @@ void __init fork_init(void)
         init_task.signal->rlim[RLIMIT_SIGPENDING] =
                 init_task.signal->rlim[RLIMIT_NPROC];
 
-        init_user_ns.max_user_namespaces = max_threads;
+        init_user_ns.max_user_namespaces = max_threads/2;
 }
 
 int __weak arch_dup_task_struct(struct task_struct *dst,
diff --git a/kernel/ucount.c b/kernel/ucount.c
index 6c2205c0befd..33c418718304 100644
--- a/kernel/ucount.c
+++ b/kernel/ucount.c
@@ -8,8 +8,20 @@
 #include <linux/stat.h>
 #include <linux/sysctl.h>
 #include <linux/slab.h>
+#include <linux/hash.h>
 #include <linux/user_namespace.h>
 
+#define UCOUNTS_HASHTABLE_BITS 10
+static struct hlist_head ucounts_hashtable[(1 << UCOUNTS_HASHTABLE_BITS)];
+static DEFINE_SPINLOCK(ucounts_lock);
+
+#define ucounts_hashfn(ns, uid)                                         \
+        hash_long((unsigned long)__kuid_val(uid) + (unsigned long)(ns), \
+                  UCOUNTS_HASHTABLE_BITS)
+#define ucounts_hashentry(ns, uid)      \
+        (ucounts_hashtable + ucounts_hashfn(ns, uid))
+
+
 #ifdef CONFIG_SYSCTL
 static struct ctl_table_set *
 set_lookup(struct ctl_table_root *root)
@@ -45,7 +57,7 @@ static struct ctl_table_root set_root = {
 
 static int zero = 0;
 static int int_max = INT_MAX;
-static struct ctl_table userns_table[] = {
+static struct ctl_table user_table[] = {
         {
                 .procname       = "max_user_namespaces",
                 .data           = &init_user_ns.max_user_namespaces,
@@ -64,11 +76,11 @@ bool setup_userns_sysctls(struct user_namespace *ns)
 #ifdef CONFIG_SYSCTL
         struct ctl_table *tbl;
         setup_sysctl_set(&ns->set, &set_root, set_is_seen);
-        tbl = kmemdup(userns_table, sizeof(userns_table), GFP_KERNEL);
+        tbl = kmemdup(user_table, sizeof(user_table), GFP_KERNEL);
         if (tbl) {
                 tbl[0].data = &ns->max_user_namespaces;
 
-                ns->sysctls = __register_sysctl_table(&ns->set, "userns", tbl);
+                ns->sysctls = __register_sysctl_table(&ns->set, "user", tbl);
         }
         if (!ns->sysctls) {
                 kfree(tbl);
@@ -91,6 +103,61 @@ void retire_userns_sysctls(struct user_namespace *ns)
 #endif
 }
 
+static struct ucounts *find_ucounts(struct user_namespace *ns, kuid_t uid, struct hlist_head *hashent)
+{
+        struct ucounts *ucounts;
+
+        hlist_for_each_entry(ucounts, hashent, node) {
+                if (uid_eq(ucounts->uid, uid) && (ucounts->ns == ns))
+                        return ucounts;
+        }
+        return NULL;
+}
+
+static struct ucounts *get_ucounts(struct user_namespace *ns, kuid_t uid)
+{
+        struct hlist_head *hashent = ucounts_hashentry(ns, uid);
+        struct ucounts *ucounts, *new;
+
+        spin_lock(&ucounts_lock);
+        ucounts = find_ucounts(ns, uid, hashent);
+        if (!ucounts) {
+                spin_unlock(&ucounts_lock);
+
+                new = kzalloc(sizeof(*new), GFP_KERNEL);
+                if (!new)
+                        return NULL;
+
+                new->ns = ns;
+                new->uid = uid;
+                atomic_set(&new->count, 0);
+
+                spin_lock(&ucounts_lock);
+                ucounts = find_ucounts(ns, uid, hashent);
+                if (ucounts) {
+                        kfree(new);
+                } else {
+                        hlist_add_head(&new->node, hashent);
+                        ucounts = new;
+                }
+        }
+        if (!atomic_add_unless(&ucounts->count, 1, INT_MAX))
+                ucounts = NULL;
+        spin_unlock(&ucounts_lock);
+        return ucounts;
+}
+
+static void put_ucounts(struct ucounts *ucounts)
+{
+        if (atomic_dec_and_test(&ucounts->count)) {
+                spin_lock(&ucounts_lock);
+                hlist_del_init(&ucounts->node);
+                spin_unlock(&ucounts_lock);
+
+                kfree(ucounts);
+        }
+}
+
 static inline bool atomic_inc_below(atomic_t *v, int u)
 {
         int c, old;
@@ -105,44 +172,51 @@ static inline bool atomic_inc_below(atomic_t *v, int u)
         }
 }
 
-bool inc_user_namespaces(struct user_namespace *ns)
+struct ucounts *inc_user_namespaces(struct user_namespace *ns, kuid_t uid)
 {
-        struct user_namespace *pos, *bad;
-        for (pos = ns; pos; pos = pos->parent) {
-                int max = READ_ONCE(pos->max_user_namespaces);
-                if (!atomic_inc_below(&pos->user_namespaces, max))
+        struct ucounts *ucounts, *iter, *bad;
+        struct user_namespace *tns;
+        ucounts = get_ucounts(ns, uid);
+        for (iter = ucounts; iter; iter = tns->ucounts) {
+                int max;
+                tns = iter->ns;
+                max = READ_ONCE(tns->max_user_namespaces);
+                if (!atomic_inc_below(&iter->user_namespaces, max))
                         goto fail;
         }
-        return true;
+        return ucounts;
 fail:
-        bad = pos;
-        for (pos = ns; pos != bad; pos = pos->parent)
-                atomic_dec(&pos->user_namespaces);
+        bad = iter;
+        for (iter = ucounts; iter != bad; iter = iter->ns->ucounts)
+                atomic_dec(&iter->user_namespaces);
 
-        return false;
+        put_ucounts(ucounts);
+        return NULL;
 }
 
-void dec_user_namespaces(struct user_namespace *ns)
+void dec_user_namespaces(struct ucounts *ucounts)
 {
-        struct user_namespace *pos;
-        for (pos = ns; pos; pos = pos->parent) {
-                int dec = atomic_dec_if_positive(&pos->user_namespaces);
+        struct ucounts *iter;
+        for (iter = ucounts; iter; iter = iter->ns->ucounts) {
+                int dec = atomic_dec_if_positive(&iter->user_namespaces);
                 WARN_ON_ONCE(dec < 0);
         }
+        put_ucounts(ucounts);
 }
 
+
 static __init int user_namespace_sysctl_init(void)
 {
 #ifdef CONFIG_SYSCTL
-        static struct ctl_table_header *userns_header;
+        static struct ctl_table_header *user_header;
         static struct ctl_table empty[1];
         /*
-         * It is necessary to register the userns directory in the
+         * It is necessary to register the user directory in the
          * default set so that registrations in the child sets work
          * properly.
          */
-        userns_header = register_sysctl("userns", empty);
-        BUG_ON(!userns_header);
+        user_header = register_sysctl("user", empty);
+        BUG_ON(!user_header);
         BUG_ON(!setup_userns_sysctls(&init_user_ns));
 #endif
         return 0;
diff --git a/kernel/user_namespace.c b/kernel/user_namespace.c
index 7d87017a0040..58c67e5f851c 100644
--- a/kernel/user_namespace.c
+++ b/kernel/user_namespace.c
@@ -31,7 +31,6 @@ static bool new_idmap_permitted(const struct file *file,
                                 struct uid_gid_map *map);
 static void free_user_ns(struct work_struct *work);
 
-
 static void set_cred_user_ns(struct cred *cred, struct user_namespace *user_ns)
 {
         /* Start with the same capabilities as init but useless for doing
@@ -64,13 +63,15 @@ int create_user_ns(struct cred *new)
         struct user_namespace *ns, *parent_ns = new->user_ns;
         kuid_t owner = new->euid;
         kgid_t group = new->egid;
+        struct ucounts *ucounts;
         int ret;
 
         ret = -EUSERS;
         if (parent_ns->level > 32)
                 goto fail;
 
-        if (!inc_user_namespaces(parent_ns))
+        ucounts = inc_user_namespaces(parent_ns, owner);
+        if (!ucounts)
                 goto fail;
 
         /*
@@ -110,6 +111,7 @@ int create_user_ns(struct cred *new)
         ns->group = group;
         INIT_WORK(&ns->work, free_user_ns);
         ns->max_user_namespaces = INT_MAX;
+        ns->ucounts = ucounts;
 
         /* Inherit USERNS_SETGROUPS_ALLOWED from our parent */
         mutex_lock(&userns_state_mutex);
@@ -133,7 +135,7 @@ fail_keyring:
 fail_free:
         kmem_cache_free(user_ns_cachep, ns);
 fail_dec:
-        dec_user_namespaces(parent_ns);
+        dec_user_namespaces(ucounts);
 fail:
         return ret;
 }
@@ -164,6 +166,7 @@ static void free_user_ns(struct work_struct *work)
                 container_of(work, struct user_namespace, work);
 
         do {
+                struct ucounts *ucounts = ns->ucounts;
                 parent = ns->parent;
                 retire_userns_sysctls(ns);
 #ifdef CONFIG_PERSISTENT_KEYRINGS
@@ -171,7 +174,7 @@ static void free_user_ns(struct work_struct *work)
 #endif
                 ns_free_inum(&ns->ns);
                 kmem_cache_free(user_ns_cachep, ns);
-                dec_user_namespaces(parent);
+                dec_user_namespaces(ucounts);
                 ns = parent;
         } while (atomic_dec_and_test(&parent->count));
 }