Eric W. Biederman | dbec284 | 2016-07-30 13:58:49 -0500 | [diff] [blame] | 1 | /* |
| 2 | * This program is free software; you can redistribute it and/or |
| 3 | * modify it under the terms of the GNU General Public License as |
| 4 | * published by the Free Software Foundation, version 2 of the |
| 5 | * License. |
| 6 | */ |
| 7 | |
| 8 | #include <linux/stat.h> |
| 9 | #include <linux/sysctl.h> |
| 10 | #include <linux/slab.h> |
Ingo Molnar | 5b825c3 | 2017-02-02 17:54:15 +0100 | [diff] [blame] | 11 | #include <linux/cred.h> |
Eric W. Biederman | f6b2db1a | 2016-08-08 13:54:50 -0500 | [diff] [blame] | 12 | #include <linux/hash.h> |
Randy Dunlap | 514c603 | 2018-04-05 16:25:34 -0700 | [diff] [blame] | 13 | #include <linux/kmemleak.h> |
Eric W. Biederman | dbec284 | 2016-07-30 13:58:49 -0500 | [diff] [blame] | 14 | #include <linux/user_namespace.h> |
| 15 | |
Eric W. Biederman | f6b2db1a | 2016-08-08 13:54:50 -0500 | [diff] [blame] | 16 | #define UCOUNTS_HASHTABLE_BITS 10 |
| 17 | static struct hlist_head ucounts_hashtable[(1 << UCOUNTS_HASHTABLE_BITS)]; |
| 18 | static DEFINE_SPINLOCK(ucounts_lock); |
| 19 | |
| 20 | #define ucounts_hashfn(ns, uid) \ |
| 21 | hash_long((unsigned long)__kuid_val(uid) + (unsigned long)(ns), \ |
| 22 | UCOUNTS_HASHTABLE_BITS) |
| 23 | #define ucounts_hashentry(ns, uid) \ |
| 24 | (ucounts_hashtable + ucounts_hashfn(ns, uid)) |
| 25 | |
| 26 | |
Eric W. Biederman | dbec284 | 2016-07-30 13:58:49 -0500 | [diff] [blame] | 27 | #ifdef CONFIG_SYSCTL |
| 28 | static struct ctl_table_set * |
| 29 | set_lookup(struct ctl_table_root *root) |
| 30 | { |
| 31 | return ¤t_user_ns()->set; |
| 32 | } |
| 33 | |
| 34 | static int set_is_seen(struct ctl_table_set *set) |
| 35 | { |
| 36 | return ¤t_user_ns()->set == set; |
| 37 | } |
| 38 | |
| 39 | static int set_permissions(struct ctl_table_header *head, |
| 40 | struct ctl_table *table) |
| 41 | { |
| 42 | struct user_namespace *user_ns = |
| 43 | container_of(head->set, struct user_namespace, set); |
| 44 | int mode; |
| 45 | |
| 46 | /* Allow users with CAP_SYS_RESOURCE unrestrained access */ |
| 47 | if (ns_capable(user_ns, CAP_SYS_RESOURCE)) |
| 48 | mode = (table->mode & S_IRWXU) >> 6; |
| 49 | else |
| 50 | /* Allow all others at most read-only access */ |
| 51 | mode = table->mode & S_IROTH; |
| 52 | return (mode << 6) | (mode << 3) | mode; |
| 53 | } |
| 54 | |
| 55 | static struct ctl_table_root set_root = { |
| 56 | .lookup = set_lookup, |
| 57 | .permissions = set_permissions, |
| 58 | }; |
| 59 | |
Eric W. Biederman | b376c3e | 2016-08-08 13:41:24 -0500 | [diff] [blame] | 60 | static int zero = 0; |
| 61 | static int int_max = INT_MAX; |
Nikolay Borisov | 1cce1ee | 2016-12-14 15:56:33 +0200 | [diff] [blame] | 62 | #define UCOUNT_ENTRY(name) \ |
Eric W. Biederman | 25f9c08 | 2016-08-08 14:41:52 -0500 | [diff] [blame] | 63 | { \ |
| 64 | .procname = name, \ |
| 65 | .maxlen = sizeof(int), \ |
| 66 | .mode = 0644, \ |
| 67 | .proc_handler = proc_dointvec_minmax, \ |
| 68 | .extra1 = &zero, \ |
| 69 | .extra2 = &int_max, \ |
| 70 | } |
Eric W. Biederman | f6b2db1a | 2016-08-08 13:54:50 -0500 | [diff] [blame] | 71 | static struct ctl_table user_table[] = { |
Eric W. Biederman | 25f9c08 | 2016-08-08 14:41:52 -0500 | [diff] [blame] | 72 | UCOUNT_ENTRY("max_user_namespaces"), |
Eric W. Biederman | f333c70 | 2016-08-08 14:08:36 -0500 | [diff] [blame] | 73 | UCOUNT_ENTRY("max_pid_namespaces"), |
Eric W. Biederman | f7af3d1 | 2016-08-08 14:11:25 -0500 | [diff] [blame] | 74 | UCOUNT_ENTRY("max_uts_namespaces"), |
Eric W. Biederman | aba3566 | 2016-08-08 14:20:23 -0500 | [diff] [blame] | 75 | UCOUNT_ENTRY("max_ipc_namespaces"), |
Eric W. Biederman | 7032866 | 2016-08-08 14:33:23 -0500 | [diff] [blame] | 76 | UCOUNT_ENTRY("max_net_namespaces"), |
Eric W. Biederman | 537f7cc | 2016-08-08 14:37:37 -0500 | [diff] [blame] | 77 | UCOUNT_ENTRY("max_mnt_namespaces"), |
Eric W. Biederman | d08311d | 2016-08-08 14:25:30 -0500 | [diff] [blame] | 78 | UCOUNT_ENTRY("max_cgroup_namespaces"), |
Nikolay Borisov | 1cce1ee | 2016-12-14 15:56:33 +0200 | [diff] [blame] | 79 | #ifdef CONFIG_INOTIFY_USER |
| 80 | UCOUNT_ENTRY("max_inotify_instances"), |
| 81 | UCOUNT_ENTRY("max_inotify_watches"), |
| 82 | #endif |
Eric W. Biederman | dbec284 | 2016-07-30 13:58:49 -0500 | [diff] [blame] | 83 | { } |
| 84 | }; |
| 85 | #endif /* CONFIG_SYSCTL */ |
| 86 | |
| 87 | bool setup_userns_sysctls(struct user_namespace *ns) |
| 88 | { |
| 89 | #ifdef CONFIG_SYSCTL |
| 90 | struct ctl_table *tbl; |
| 91 | setup_sysctl_set(&ns->set, &set_root, set_is_seen); |
Eric W. Biederman | f6b2db1a | 2016-08-08 13:54:50 -0500 | [diff] [blame] | 92 | tbl = kmemdup(user_table, sizeof(user_table), GFP_KERNEL); |
Eric W. Biederman | dbec284 | 2016-07-30 13:58:49 -0500 | [diff] [blame] | 93 | if (tbl) { |
Eric W. Biederman | 25f9c08 | 2016-08-08 14:41:52 -0500 | [diff] [blame] | 94 | int i; |
| 95 | for (i = 0; i < UCOUNT_COUNTS; i++) { |
| 96 | tbl[i].data = &ns->ucount_max[i]; |
| 97 | } |
Eric W. Biederman | f6b2db1a | 2016-08-08 13:54:50 -0500 | [diff] [blame] | 98 | ns->sysctls = __register_sysctl_table(&ns->set, "user", tbl); |
Eric W. Biederman | dbec284 | 2016-07-30 13:58:49 -0500 | [diff] [blame] | 99 | } |
| 100 | if (!ns->sysctls) { |
| 101 | kfree(tbl); |
| 102 | retire_sysctl_set(&ns->set); |
| 103 | return false; |
| 104 | } |
| 105 | #endif |
| 106 | return true; |
| 107 | } |
| 108 | |
| 109 | void retire_userns_sysctls(struct user_namespace *ns) |
| 110 | { |
| 111 | #ifdef CONFIG_SYSCTL |
| 112 | struct ctl_table *tbl; |
| 113 | |
| 114 | tbl = ns->sysctls->ctl_table_arg; |
| 115 | unregister_sysctl_table(ns->sysctls); |
| 116 | retire_sysctl_set(&ns->set); |
| 117 | kfree(tbl); |
| 118 | #endif |
| 119 | } |
| 120 | |
Eric W. Biederman | f6b2db1a | 2016-08-08 13:54:50 -0500 | [diff] [blame] | 121 | static struct ucounts *find_ucounts(struct user_namespace *ns, kuid_t uid, struct hlist_head *hashent) |
| 122 | { |
| 123 | struct ucounts *ucounts; |
| 124 | |
| 125 | hlist_for_each_entry(ucounts, hashent, node) { |
| 126 | if (uid_eq(ucounts->uid, uid) && (ucounts->ns == ns)) |
| 127 | return ucounts; |
| 128 | } |
| 129 | return NULL; |
| 130 | } |
| 131 | |
| 132 | static struct ucounts *get_ucounts(struct user_namespace *ns, kuid_t uid) |
| 133 | { |
| 134 | struct hlist_head *hashent = ucounts_hashentry(ns, uid); |
| 135 | struct ucounts *ucounts, *new; |
| 136 | |
Nikolay Borisov | 880a385 | 2017-01-20 15:21:35 +0200 | [diff] [blame] | 137 | spin_lock_irq(&ucounts_lock); |
Eric W. Biederman | f6b2db1a | 2016-08-08 13:54:50 -0500 | [diff] [blame] | 138 | ucounts = find_ucounts(ns, uid, hashent); |
| 139 | if (!ucounts) { |
Nikolay Borisov | 880a385 | 2017-01-20 15:21:35 +0200 | [diff] [blame] | 140 | spin_unlock_irq(&ucounts_lock); |
Eric W. Biederman | f6b2db1a | 2016-08-08 13:54:50 -0500 | [diff] [blame] | 141 | |
| 142 | new = kzalloc(sizeof(*new), GFP_KERNEL); |
| 143 | if (!new) |
| 144 | return NULL; |
| 145 | |
| 146 | new->ns = ns; |
| 147 | new->uid = uid; |
Eric W. Biederman | 040757f | 2017-03-05 15:03:22 -0600 | [diff] [blame] | 148 | new->count = 0; |
Eric W. Biederman | f6b2db1a | 2016-08-08 13:54:50 -0500 | [diff] [blame] | 149 | |
Nikolay Borisov | 880a385 | 2017-01-20 15:21:35 +0200 | [diff] [blame] | 150 | spin_lock_irq(&ucounts_lock); |
Eric W. Biederman | f6b2db1a | 2016-08-08 13:54:50 -0500 | [diff] [blame] | 151 | ucounts = find_ucounts(ns, uid, hashent); |
| 152 | if (ucounts) { |
| 153 | kfree(new); |
| 154 | } else { |
| 155 | hlist_add_head(&new->node, hashent); |
| 156 | ucounts = new; |
| 157 | } |
| 158 | } |
Eric W. Biederman | 040757f | 2017-03-05 15:03:22 -0600 | [diff] [blame] | 159 | if (ucounts->count == INT_MAX) |
Eric W. Biederman | f6b2db1a | 2016-08-08 13:54:50 -0500 | [diff] [blame] | 160 | ucounts = NULL; |
Eric W. Biederman | 040757f | 2017-03-05 15:03:22 -0600 | [diff] [blame] | 161 | else |
| 162 | ucounts->count += 1; |
Nikolay Borisov | 880a385 | 2017-01-20 15:21:35 +0200 | [diff] [blame] | 163 | spin_unlock_irq(&ucounts_lock); |
Eric W. Biederman | f6b2db1a | 2016-08-08 13:54:50 -0500 | [diff] [blame] | 164 | return ucounts; |
| 165 | } |
| 166 | |
| 167 | static void put_ucounts(struct ucounts *ucounts) |
| 168 | { |
Nikolay Borisov | 880a385 | 2017-01-20 15:21:35 +0200 | [diff] [blame] | 169 | unsigned long flags; |
| 170 | |
Eric W. Biederman | 040757f | 2017-03-05 15:03:22 -0600 | [diff] [blame] | 171 | spin_lock_irqsave(&ucounts_lock, flags); |
| 172 | ucounts->count -= 1; |
| 173 | if (!ucounts->count) |
Eric W. Biederman | f6b2db1a | 2016-08-08 13:54:50 -0500 | [diff] [blame] | 174 | hlist_del_init(&ucounts->node); |
Eric W. Biederman | 040757f | 2017-03-05 15:03:22 -0600 | [diff] [blame] | 175 | else |
| 176 | ucounts = NULL; |
| 177 | spin_unlock_irqrestore(&ucounts_lock, flags); |
Eric W. Biederman | f6b2db1a | 2016-08-08 13:54:50 -0500 | [diff] [blame] | 178 | |
Eric W. Biederman | 040757f | 2017-03-05 15:03:22 -0600 | [diff] [blame] | 179 | kfree(ucounts); |
Eric W. Biederman | f6b2db1a | 2016-08-08 13:54:50 -0500 | [diff] [blame] | 180 | } |
| 181 | |
Eric W. Biederman | b376c3e | 2016-08-08 13:41:24 -0500 | [diff] [blame] | 182 | static inline bool atomic_inc_below(atomic_t *v, int u) |
| 183 | { |
| 184 | int c, old; |
| 185 | c = atomic_read(v); |
| 186 | for (;;) { |
| 187 | if (unlikely(c >= u)) |
| 188 | return false; |
| 189 | old = atomic_cmpxchg(v, c, c+1); |
| 190 | if (likely(old == c)) |
| 191 | return true; |
| 192 | c = old; |
| 193 | } |
| 194 | } |
| 195 | |
Eric W. Biederman | 25f9c08 | 2016-08-08 14:41:52 -0500 | [diff] [blame] | 196 | struct ucounts *inc_ucount(struct user_namespace *ns, kuid_t uid, |
| 197 | enum ucount_type type) |
Eric W. Biederman | b376c3e | 2016-08-08 13:41:24 -0500 | [diff] [blame] | 198 | { |
Eric W. Biederman | f6b2db1a | 2016-08-08 13:54:50 -0500 | [diff] [blame] | 199 | struct ucounts *ucounts, *iter, *bad; |
| 200 | struct user_namespace *tns; |
| 201 | ucounts = get_ucounts(ns, uid); |
| 202 | for (iter = ucounts; iter; iter = tns->ucounts) { |
| 203 | int max; |
| 204 | tns = iter->ns; |
Eric W. Biederman | 25f9c08 | 2016-08-08 14:41:52 -0500 | [diff] [blame] | 205 | max = READ_ONCE(tns->ucount_max[type]); |
| 206 | if (!atomic_inc_below(&iter->ucount[type], max)) |
Eric W. Biederman | b376c3e | 2016-08-08 13:41:24 -0500 | [diff] [blame] | 207 | goto fail; |
| 208 | } |
Eric W. Biederman | f6b2db1a | 2016-08-08 13:54:50 -0500 | [diff] [blame] | 209 | return ucounts; |
Eric W. Biederman | b376c3e | 2016-08-08 13:41:24 -0500 | [diff] [blame] | 210 | fail: |
Eric W. Biederman | f6b2db1a | 2016-08-08 13:54:50 -0500 | [diff] [blame] | 211 | bad = iter; |
| 212 | for (iter = ucounts; iter != bad; iter = iter->ns->ucounts) |
Eric W. Biederman | 25f9c08 | 2016-08-08 14:41:52 -0500 | [diff] [blame] | 213 | atomic_dec(&iter->ucount[type]); |
Eric W. Biederman | b376c3e | 2016-08-08 13:41:24 -0500 | [diff] [blame] | 214 | |
Eric W. Biederman | f6b2db1a | 2016-08-08 13:54:50 -0500 | [diff] [blame] | 215 | put_ucounts(ucounts); |
| 216 | return NULL; |
Eric W. Biederman | b376c3e | 2016-08-08 13:41:24 -0500 | [diff] [blame] | 217 | } |
| 218 | |
Eric W. Biederman | 25f9c08 | 2016-08-08 14:41:52 -0500 | [diff] [blame] | 219 | void dec_ucount(struct ucounts *ucounts, enum ucount_type type) |
Eric W. Biederman | b376c3e | 2016-08-08 13:41:24 -0500 | [diff] [blame] | 220 | { |
Eric W. Biederman | f6b2db1a | 2016-08-08 13:54:50 -0500 | [diff] [blame] | 221 | struct ucounts *iter; |
| 222 | for (iter = ucounts; iter; iter = iter->ns->ucounts) { |
Eric W. Biederman | 25f9c08 | 2016-08-08 14:41:52 -0500 | [diff] [blame] | 223 | int dec = atomic_dec_if_positive(&iter->ucount[type]); |
Eric W. Biederman | b376c3e | 2016-08-08 13:41:24 -0500 | [diff] [blame] | 224 | WARN_ON_ONCE(dec < 0); |
| 225 | } |
Eric W. Biederman | f6b2db1a | 2016-08-08 13:54:50 -0500 | [diff] [blame] | 226 | put_ucounts(ucounts); |
Eric W. Biederman | b376c3e | 2016-08-08 13:41:24 -0500 | [diff] [blame] | 227 | } |
| 228 | |
Eric W. Biederman | dbec284 | 2016-07-30 13:58:49 -0500 | [diff] [blame] | 229 | static __init int user_namespace_sysctl_init(void) |
| 230 | { |
| 231 | #ifdef CONFIG_SYSCTL |
Eric W. Biederman | f6b2db1a | 2016-08-08 13:54:50 -0500 | [diff] [blame] | 232 | static struct ctl_table_header *user_header; |
Eric W. Biederman | dbec284 | 2016-07-30 13:58:49 -0500 | [diff] [blame] | 233 | static struct ctl_table empty[1]; |
| 234 | /* |
Eric W. Biederman | f6b2db1a | 2016-08-08 13:54:50 -0500 | [diff] [blame] | 235 | * It is necessary to register the user directory in the |
Eric W. Biederman | dbec284 | 2016-07-30 13:58:49 -0500 | [diff] [blame] | 236 | * default set so that registrations in the child sets work |
| 237 | * properly. |
| 238 | */ |
Eric W. Biederman | f6b2db1a | 2016-08-08 13:54:50 -0500 | [diff] [blame] | 239 | user_header = register_sysctl("user", empty); |
Luis R. Rodriguez | ed5bd7d | 2017-02-08 14:30:50 -0800 | [diff] [blame] | 240 | kmemleak_ignore(user_header); |
Eric W. Biederman | f6b2db1a | 2016-08-08 13:54:50 -0500 | [diff] [blame] | 241 | BUG_ON(!user_header); |
Eric W. Biederman | dbec284 | 2016-07-30 13:58:49 -0500 | [diff] [blame] | 242 | BUG_ON(!setup_userns_sysctls(&init_user_ns)); |
| 243 | #endif |
| 244 | return 0; |
| 245 | } |
| 246 | subsys_initcall(user_namespace_sysctl_init); |