Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1 | /* |
| 2 | * Generic pidhash and scalable, time-bounded PID allocator |
| 3 | * |
Nadia Yvette Chambers | 6d49e35 | 2012-12-06 10:39:54 +0100 | [diff] [blame] | 4 | * (C) 2002-2003 Nadia Yvette Chambers, IBM |
| 5 | * (C) 2004 Nadia Yvette Chambers, Oracle |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 6 | * (C) 2002-2004 Ingo Molnar, Red Hat |
| 7 | * |
| 8 | * pid-structures are backing objects for tasks sharing a given ID to chain |
| 9 | * against. There is very little to them aside from hashing them and |
| 10 | * parking tasks using given ID's on a list. |
| 11 | * |
| 12 | * The hash is always changed with the tasklist_lock write-acquired, |
| 13 | * and the hash is only accessed with the tasklist_lock at least |
| 14 | * read-acquired, so there's no additional SMP locking needed here. |
| 15 | * |
| 16 | * We have a list of bitmap pages, which bitmaps represent the PID space. |
| 17 | * Allocating and freeing PIDs is completely lockless. The worst-case |
| 18 | * allocation scenario when all but one out of 1 million PIDs possible are |
| 19 | * allocated already: the scanning of 32 list entries and at most PAGE_SIZE |
| 20 | * bytes. The typical fastpath is a single successful setbit. Freeing is O(1). |
Pavel Emelyanov | 30e49c2 | 2007-10-18 23:40:10 -0700 | [diff] [blame] | 21 | * |
| 22 | * Pid namespaces: |
| 23 | * (C) 2007 Pavel Emelyanov <xemul@openvz.org>, OpenVZ, SWsoft Inc. |
| 24 | * (C) 2007 Sukadev Bhattiprolu <sukadev@us.ibm.com>, IBM |
| 25 | * Many thanks to Oleg Nesterov for comments and help |
| 26 | * |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 27 | */ |
| 28 | |
| 29 | #include <linux/mm.h> |
Paul Gortmaker | 9984de1 | 2011-05-23 14:51:41 -0400 | [diff] [blame] | 30 | #include <linux/export.h> |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 31 | #include <linux/slab.h> |
| 32 | #include <linux/init.h> |
Franck Bui-Huu | 8252474 | 2008-05-12 21:21:05 +0200 | [diff] [blame] | 33 | #include <linux/rculist.h> |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 34 | #include <linux/bootmem.h> |
| 35 | #include <linux/hash.h> |
Sukadev Bhattiprolu | 61a58c6 | 2006-12-08 02:37:58 -0800 | [diff] [blame] | 36 | #include <linux/pid_namespace.h> |
Sukadev Bhattiprolu | 820e45d | 2007-05-10 22:23:00 -0700 | [diff] [blame] | 37 | #include <linux/init_task.h> |
Sukadev Bhattiprolu | 3eb07c8 | 2007-10-18 23:40:13 -0700 | [diff] [blame] | 38 | #include <linux/syscalls.h> |
David Howells | 0bb80f2 | 2013-04-12 01:50:06 +0100 | [diff] [blame] | 39 | #include <linux/proc_ns.h> |
Eric W. Biederman | 0a01f2c | 2012-08-01 10:33:47 -0700 | [diff] [blame] | 40 | #include <linux/proc_fs.h> |
Ingo Molnar | 2993002 | 2017-02-08 18:51:36 +0100 | [diff] [blame] | 41 | #include <linux/sched/task.h> |
Gargi Sharma | 95846ec | 2017-11-17 15:30:30 -0800 | [diff] [blame] | 42 | #include <linux/idr.h> |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 43 | |
David Howells | e1e871a | 2018-01-02 15:12:01 +0000 | [diff] [blame] | 44 | struct pid init_struct_pid = { |
| 45 | .count = ATOMIC_INIT(1), |
| 46 | .tasks = { |
| 47 | { .first = NULL }, |
| 48 | { .first = NULL }, |
| 49 | { .first = NULL }, |
| 50 | }, |
| 51 | .level = 0, |
| 52 | .numbers = { { |
| 53 | .nr = 0, |
| 54 | .ns = &init_pid_ns, |
| 55 | }, } |
| 56 | }; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 57 | |
| 58 | int pid_max = PID_MAX_DEFAULT; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 59 | |
| 60 | #define RESERVED_PIDS 300 |
| 61 | |
| 62 | int pid_max_min = RESERVED_PIDS + 1; |
| 63 | int pid_max_max = PID_MAX_LIMIT; |
| 64 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 65 | /* |
| 66 | * PID-map pages start out as NULL, they get allocated upon |
| 67 | * first use and are never deallocated. This way a low pid_max |
| 68 | * value does not cause lots of bitmaps to be allocated, but |
| 69 | * the scheme scales to up to 4 million PIDs, runtime. |
| 70 | */ |
Sukadev Bhattiprolu | 61a58c6 | 2006-12-08 02:37:58 -0800 | [diff] [blame] | 71 | struct pid_namespace init_pid_ns = { |
Peter Zijlstra | 1e24edc | 2016-11-14 17:12:23 +0100 | [diff] [blame] | 72 | .kref = KREF_INIT(2), |
Matthew Wilcox | f6bb2a2 | 2018-04-10 16:36:52 -0700 | [diff] [blame] | 73 | .idr = IDR_INIT(init_pid_ns.idr), |
Gargi Sharma | e8cfbc2 | 2017-11-17 15:30:34 -0800 | [diff] [blame] | 74 | .pid_allocated = PIDNS_ADDING, |
Pavel Emelyanov | faacbfd | 2007-10-18 23:40:04 -0700 | [diff] [blame] | 75 | .level = 0, |
| 76 | .child_reaper = &init_task, |
Eric W. Biederman | 49f4d8b | 2012-08-02 04:25:10 -0700 | [diff] [blame] | 77 | .user_ns = &init_user_ns, |
Al Viro | 435d5f4 | 2014-10-31 22:56:04 -0400 | [diff] [blame] | 78 | .ns.inum = PROC_PID_INIT_INO, |
Al Viro | 33c4294 | 2014-11-01 02:32:53 -0400 | [diff] [blame] | 79 | #ifdef CONFIG_PID_NS |
| 80 | .ns.ops = &pidns_operations, |
| 81 | #endif |
Sukadev Bhattiprolu | 3fbc964 | 2006-10-02 02:17:24 -0700 | [diff] [blame] | 82 | }; |
Pavel Emelyanov | 198fe21 | 2007-10-18 23:40:06 -0700 | [diff] [blame] | 83 | EXPORT_SYMBOL_GPL(init_pid_ns); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 84 | |
Eric W. Biederman | 92476d7 | 2006-03-31 02:31:42 -0800 | [diff] [blame] | 85 | /* |
| 86 | * Note: disable interrupts while the pidmap_lock is held as an |
| 87 | * interrupt might come in and do read_lock(&tasklist_lock). |
| 88 | * |
| 89 | * If we don't disable interrupts there is a nasty deadlock between |
| 90 | * detach_pid()->free_pid() and another cpu that does |
| 91 | * spin_lock(&pidmap_lock) followed by an interrupt routine that does |
| 92 | * read_lock(&tasklist_lock); |
| 93 | * |
| 94 | * After we clean up the tasklist_lock and know there are no |
| 95 | * irq handlers that take it we can leave the interrupts enabled. |
| 96 | * For now it is easier to be safe than to prove it can't happen. |
| 97 | */ |
Sukadev Bhattiprolu | 3fbc964 | 2006-10-02 02:17:24 -0700 | [diff] [blame] | 98 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 99 | static __cacheline_aligned_in_smp DEFINE_SPINLOCK(pidmap_lock); |
| 100 | |
Harvey Harrison | 7ad5b3a | 2008-02-08 04:19:53 -0800 | [diff] [blame] | 101 | void put_pid(struct pid *pid) |
Eric W. Biederman | 92476d7 | 2006-03-31 02:31:42 -0800 | [diff] [blame] | 102 | { |
Pavel Emelianov | baf8f0f | 2007-10-18 23:39:48 -0700 | [diff] [blame] | 103 | struct pid_namespace *ns; |
| 104 | |
Eric W. Biederman | 92476d7 | 2006-03-31 02:31:42 -0800 | [diff] [blame] | 105 | if (!pid) |
| 106 | return; |
Pavel Emelianov | baf8f0f | 2007-10-18 23:39:48 -0700 | [diff] [blame] | 107 | |
Pavel Emelyanov | 8ef047a | 2007-10-18 23:40:05 -0700 | [diff] [blame] | 108 | ns = pid->numbers[pid->level].ns; |
Eric W. Biederman | 92476d7 | 2006-03-31 02:31:42 -0800 | [diff] [blame] | 109 | if ((atomic_read(&pid->count) == 1) || |
Pavel Emelyanov | 8ef047a | 2007-10-18 23:40:05 -0700 | [diff] [blame] | 110 | atomic_dec_and_test(&pid->count)) { |
Pavel Emelianov | baf8f0f | 2007-10-18 23:39:48 -0700 | [diff] [blame] | 111 | kmem_cache_free(ns->pid_cachep, pid); |
Pavel Emelyanov | b461cc0 | 2007-10-18 23:40:09 -0700 | [diff] [blame] | 112 | put_pid_ns(ns); |
Pavel Emelyanov | 8ef047a | 2007-10-18 23:40:05 -0700 | [diff] [blame] | 113 | } |
Eric W. Biederman | 92476d7 | 2006-03-31 02:31:42 -0800 | [diff] [blame] | 114 | } |
Eric W. Biederman | bbf7314 | 2006-10-02 02:17:11 -0700 | [diff] [blame] | 115 | EXPORT_SYMBOL_GPL(put_pid); |
Eric W. Biederman | 92476d7 | 2006-03-31 02:31:42 -0800 | [diff] [blame] | 116 | |
| 117 | static void delayed_put_pid(struct rcu_head *rhp) |
| 118 | { |
| 119 | struct pid *pid = container_of(rhp, struct pid, rcu); |
| 120 | put_pid(pid); |
| 121 | } |
| 122 | |
Harvey Harrison | 7ad5b3a | 2008-02-08 04:19:53 -0800 | [diff] [blame] | 123 | void free_pid(struct pid *pid) |
Eric W. Biederman | 92476d7 | 2006-03-31 02:31:42 -0800 | [diff] [blame] | 124 | { |
| 125 | /* We can be called with write_lock_irq(&tasklist_lock) held */ |
Pavel Emelyanov | 8ef047a | 2007-10-18 23:40:05 -0700 | [diff] [blame] | 126 | int i; |
Eric W. Biederman | 92476d7 | 2006-03-31 02:31:42 -0800 | [diff] [blame] | 127 | unsigned long flags; |
| 128 | |
| 129 | spin_lock_irqsave(&pidmap_lock, flags); |
Eric W. Biederman | 0a01f2c | 2012-08-01 10:33:47 -0700 | [diff] [blame] | 130 | for (i = 0; i <= pid->level; i++) { |
| 131 | struct upid *upid = pid->numbers + i; |
Eric W. Biederman | af4b8a8 | 2012-08-01 15:03:42 -0700 | [diff] [blame] | 132 | struct pid_namespace *ns = upid->ns; |
Gargi Sharma | e8cfbc2 | 2017-11-17 15:30:34 -0800 | [diff] [blame] | 133 | switch (--ns->pid_allocated) { |
Eric W. Biederman | a606488 | 2013-08-29 13:56:50 -0700 | [diff] [blame] | 134 | case 2: |
Eric W. Biederman | af4b8a8 | 2012-08-01 15:03:42 -0700 | [diff] [blame] | 135 | case 1: |
| 136 | /* When all that is left in the pid namespace |
| 137 | * is the reaper wake up the reaper. The reaper |
| 138 | * may be sleeping in zap_pid_ns_processes(). |
| 139 | */ |
| 140 | wake_up_process(ns->child_reaper); |
| 141 | break; |
Gargi Sharma | e8cfbc2 | 2017-11-17 15:30:34 -0800 | [diff] [blame] | 142 | case PIDNS_ADDING: |
Oleg Nesterov | 314a8ad | 2013-09-30 13:45:27 -0700 | [diff] [blame] | 143 | /* Handle a fork failure of the first process */ |
| 144 | WARN_ON(ns->child_reaper); |
Gargi Sharma | e8cfbc2 | 2017-11-17 15:30:34 -0800 | [diff] [blame] | 145 | ns->pid_allocated = 0; |
Oleg Nesterov | 314a8ad | 2013-09-30 13:45:27 -0700 | [diff] [blame] | 146 | /* fall through */ |
Eric W. Biederman | af4b8a8 | 2012-08-01 15:03:42 -0700 | [diff] [blame] | 147 | case 0: |
Eric W. Biederman | af4b8a8 | 2012-08-01 15:03:42 -0700 | [diff] [blame] | 148 | schedule_work(&ns->proc_work); |
| 149 | break; |
Eric W. Biederman | 5e1182deb | 2010-07-12 18:50:25 -0700 | [diff] [blame] | 150 | } |
Gargi Sharma | 95846ec | 2017-11-17 15:30:30 -0800 | [diff] [blame] | 151 | |
| 152 | idr_remove(&ns->idr, upid->nr); |
Eric W. Biederman | 0a01f2c | 2012-08-01 10:33:47 -0700 | [diff] [blame] | 153 | } |
Eric W. Biederman | 92476d7 | 2006-03-31 02:31:42 -0800 | [diff] [blame] | 154 | spin_unlock_irqrestore(&pidmap_lock, flags); |
| 155 | |
Eric W. Biederman | 92476d7 | 2006-03-31 02:31:42 -0800 | [diff] [blame] | 156 | call_rcu(&pid->rcu, delayed_put_pid); |
| 157 | } |
| 158 | |
Pavel Emelyanov | 8ef047a | 2007-10-18 23:40:05 -0700 | [diff] [blame] | 159 | struct pid *alloc_pid(struct pid_namespace *ns) |
Eric W. Biederman | 92476d7 | 2006-03-31 02:31:42 -0800 | [diff] [blame] | 160 | { |
| 161 | struct pid *pid; |
| 162 | enum pid_type type; |
Pavel Emelyanov | 8ef047a | 2007-10-18 23:40:05 -0700 | [diff] [blame] | 163 | int i, nr; |
| 164 | struct pid_namespace *tmp; |
Pavel Emelyanov | 198fe21 | 2007-10-18 23:40:06 -0700 | [diff] [blame] | 165 | struct upid *upid; |
Michal Hocko | 35f71bc | 2015-04-16 12:47:38 -0700 | [diff] [blame] | 166 | int retval = -ENOMEM; |
Eric W. Biederman | 92476d7 | 2006-03-31 02:31:42 -0800 | [diff] [blame] | 167 | |
Pavel Emelianov | baf8f0f | 2007-10-18 23:39:48 -0700 | [diff] [blame] | 168 | pid = kmem_cache_alloc(ns->pid_cachep, GFP_KERNEL); |
Eric W. Biederman | 92476d7 | 2006-03-31 02:31:42 -0800 | [diff] [blame] | 169 | if (!pid) |
Michal Hocko | 35f71bc | 2015-04-16 12:47:38 -0700 | [diff] [blame] | 170 | return ERR_PTR(retval); |
Eric W. Biederman | 92476d7 | 2006-03-31 02:31:42 -0800 | [diff] [blame] | 171 | |
Pavel Emelyanov | 8ef047a | 2007-10-18 23:40:05 -0700 | [diff] [blame] | 172 | tmp = ns; |
Eric W. Biederman | 0a01f2c | 2012-08-01 10:33:47 -0700 | [diff] [blame] | 173 | pid->level = ns->level; |
Gargi Sharma | 95846ec | 2017-11-17 15:30:30 -0800 | [diff] [blame] | 174 | |
Pavel Emelyanov | 8ef047a | 2007-10-18 23:40:05 -0700 | [diff] [blame] | 175 | for (i = ns->level; i >= 0; i--) { |
Gargi Sharma | 95846ec | 2017-11-17 15:30:30 -0800 | [diff] [blame] | 176 | int pid_min = 1; |
| 177 | |
| 178 | idr_preload(GFP_KERNEL); |
| 179 | spin_lock_irq(&pidmap_lock); |
| 180 | |
| 181 | /* |
| 182 | * init really needs pid 1, but after reaching the maximum |
| 183 | * wrap back to RESERVED_PIDS |
| 184 | */ |
| 185 | if (idr_get_cursor(&tmp->idr) > RESERVED_PIDS) |
| 186 | pid_min = RESERVED_PIDS; |
| 187 | |
| 188 | /* |
| 189 | * Store a null pointer so find_pid_ns does not find |
| 190 | * a partially initialized PID (see below). |
| 191 | */ |
| 192 | nr = idr_alloc_cyclic(&tmp->idr, NULL, pid_min, |
| 193 | pid_max, GFP_ATOMIC); |
| 194 | spin_unlock_irq(&pidmap_lock); |
| 195 | idr_preload_end(); |
| 196 | |
Arnd Bergmann | 287980e | 2016-05-27 23:23:25 +0200 | [diff] [blame] | 197 | if (nr < 0) { |
Michal Hocko | 35f71bc | 2015-04-16 12:47:38 -0700 | [diff] [blame] | 198 | retval = nr; |
Pavel Emelyanov | 8ef047a | 2007-10-18 23:40:05 -0700 | [diff] [blame] | 199 | goto out_free; |
Michal Hocko | 35f71bc | 2015-04-16 12:47:38 -0700 | [diff] [blame] | 200 | } |
Eric W. Biederman | 92476d7 | 2006-03-31 02:31:42 -0800 | [diff] [blame] | 201 | |
Pavel Emelyanov | 8ef047a | 2007-10-18 23:40:05 -0700 | [diff] [blame] | 202 | pid->numbers[i].nr = nr; |
| 203 | pid->numbers[i].ns = tmp; |
| 204 | tmp = tmp->parent; |
| 205 | } |
| 206 | |
Eric W. Biederman | 0a01f2c | 2012-08-01 10:33:47 -0700 | [diff] [blame] | 207 | if (unlikely(is_child_reaper(pid))) { |
Eric W. Biederman | c0ee554 | 2017-12-22 12:37:43 -0600 | [diff] [blame] | 208 | if (pid_ns_prepare_proc(ns)) |
Eric W. Biederman | 0a01f2c | 2012-08-01 10:33:47 -0700 | [diff] [blame] | 209 | goto out_free; |
| 210 | } |
| 211 | |
Pavel Emelyanov | b461cc0 | 2007-10-18 23:40:09 -0700 | [diff] [blame] | 212 | get_pid_ns(ns); |
Eric W. Biederman | 92476d7 | 2006-03-31 02:31:42 -0800 | [diff] [blame] | 213 | atomic_set(&pid->count, 1); |
Eric W. Biederman | 92476d7 | 2006-03-31 02:31:42 -0800 | [diff] [blame] | 214 | for (type = 0; type < PIDTYPE_MAX; ++type) |
| 215 | INIT_HLIST_HEAD(&pid->tasks[type]); |
| 216 | |
André Goddard Rosa | 417e315 | 2009-12-15 16:47:40 -0800 | [diff] [blame] | 217 | upid = pid->numbers + ns->level; |
Eric W. Biederman | 92476d7 | 2006-03-31 02:31:42 -0800 | [diff] [blame] | 218 | spin_lock_irq(&pidmap_lock); |
Gargi Sharma | e8cfbc2 | 2017-11-17 15:30:34 -0800 | [diff] [blame] | 219 | if (!(ns->pid_allocated & PIDNS_ADDING)) |
Eric W. Biederman | 5e1182deb | 2010-07-12 18:50:25 -0700 | [diff] [blame] | 220 | goto out_unlock; |
Eric W. Biederman | 0a01f2c | 2012-08-01 10:33:47 -0700 | [diff] [blame] | 221 | for ( ; upid >= pid->numbers; --upid) { |
Gargi Sharma | 95846ec | 2017-11-17 15:30:30 -0800 | [diff] [blame] | 222 | /* Make the PID visible to find_pid_ns. */ |
| 223 | idr_replace(&upid->ns->idr, pid, upid->nr); |
Gargi Sharma | e8cfbc2 | 2017-11-17 15:30:34 -0800 | [diff] [blame] | 224 | upid->ns->pid_allocated++; |
Eric W. Biederman | 0a01f2c | 2012-08-01 10:33:47 -0700 | [diff] [blame] | 225 | } |
Eric W. Biederman | 92476d7 | 2006-03-31 02:31:42 -0800 | [diff] [blame] | 226 | spin_unlock_irq(&pidmap_lock); |
| 227 | |
Eric W. Biederman | 92476d7 | 2006-03-31 02:31:42 -0800 | [diff] [blame] | 228 | return pid; |
| 229 | |
Eric W. Biederman | 5e1182deb | 2010-07-12 18:50:25 -0700 | [diff] [blame] | 230 | out_unlock: |
Eric W. Biederman | 6e66688 | 2013-02-12 13:46:23 -0800 | [diff] [blame] | 231 | spin_unlock_irq(&pidmap_lock); |
Oleg Nesterov | 24c037e | 2014-12-10 15:55:25 -0800 | [diff] [blame] | 232 | put_pid_ns(ns); |
| 233 | |
Eric W. Biederman | 92476d7 | 2006-03-31 02:31:42 -0800 | [diff] [blame] | 234 | out_free: |
Gargi Sharma | 95846ec | 2017-11-17 15:30:30 -0800 | [diff] [blame] | 235 | spin_lock_irq(&pidmap_lock); |
Oleg Nesterov | b7127aa | 2008-04-30 00:54:22 -0700 | [diff] [blame] | 236 | while (++i <= ns->level) |
Gargi Sharma | 95846ec | 2017-11-17 15:30:30 -0800 | [diff] [blame] | 237 | idr_remove(&ns->idr, (pid->numbers + i)->nr); |
| 238 | |
Eric W. Biederman | c0ee554 | 2017-12-22 12:37:43 -0600 | [diff] [blame] | 239 | /* On failure to allocate the first pid, reset the state */ |
| 240 | if (ns->pid_allocated == PIDNS_ADDING) |
| 241 | idr_set_cursor(&ns->idr, 0); |
| 242 | |
Gargi Sharma | 95846ec | 2017-11-17 15:30:30 -0800 | [diff] [blame] | 243 | spin_unlock_irq(&pidmap_lock); |
Pavel Emelyanov | 8ef047a | 2007-10-18 23:40:05 -0700 | [diff] [blame] | 244 | |
Pavel Emelianov | baf8f0f | 2007-10-18 23:39:48 -0700 | [diff] [blame] | 245 | kmem_cache_free(ns->pid_cachep, pid); |
Michal Hocko | 35f71bc | 2015-04-16 12:47:38 -0700 | [diff] [blame] | 246 | return ERR_PTR(retval); |
Eric W. Biederman | 92476d7 | 2006-03-31 02:31:42 -0800 | [diff] [blame] | 247 | } |
| 248 | |
Eric W. Biederman | c876ad76 | 2012-12-21 20:27:12 -0800 | [diff] [blame] | 249 | void disable_pid_allocation(struct pid_namespace *ns) |
| 250 | { |
| 251 | spin_lock_irq(&pidmap_lock); |
Gargi Sharma | e8cfbc2 | 2017-11-17 15:30:34 -0800 | [diff] [blame] | 252 | ns->pid_allocated &= ~PIDNS_ADDING; |
Eric W. Biederman | c876ad76 | 2012-12-21 20:27:12 -0800 | [diff] [blame] | 253 | spin_unlock_irq(&pidmap_lock); |
| 254 | } |
| 255 | |
Harvey Harrison | 7ad5b3a | 2008-02-08 04:19:53 -0800 | [diff] [blame] | 256 | struct pid *find_pid_ns(int nr, struct pid_namespace *ns) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 257 | { |
Gargi Sharma | e8cfbc2 | 2017-11-17 15:30:34 -0800 | [diff] [blame] | 258 | return idr_find(&ns->idr, nr); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 259 | } |
Pavel Emelyanov | 198fe21 | 2007-10-18 23:40:06 -0700 | [diff] [blame] | 260 | EXPORT_SYMBOL_GPL(find_pid_ns); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 261 | |
Pavel Emelyanov | 8990571 | 2007-10-18 23:40:19 -0700 | [diff] [blame] | 262 | struct pid *find_vpid(int nr) |
| 263 | { |
Eric W. Biederman | 17cf22c | 2010-03-02 14:51:53 -0800 | [diff] [blame] | 264 | return find_pid_ns(nr, task_active_pid_ns(current)); |
Pavel Emelyanov | 8990571 | 2007-10-18 23:40:19 -0700 | [diff] [blame] | 265 | } |
| 266 | EXPORT_SYMBOL_GPL(find_vpid); |
| 267 | |
Sukadev Bhattiprolu | e713d0d | 2007-05-10 22:22:58 -0700 | [diff] [blame] | 268 | /* |
| 269 | * attach_pid() must be called with the tasklist_lock write-held. |
| 270 | */ |
Oleg Nesterov | 8190773 | 2013-07-03 15:08:31 -0700 | [diff] [blame] | 271 | void attach_pid(struct task_struct *task, enum pid_type type) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 272 | { |
Oleg Nesterov | 8190773 | 2013-07-03 15:08:31 -0700 | [diff] [blame] | 273 | struct pid_link *link = &task->pids[type]; |
| 274 | hlist_add_head_rcu(&link->node, &link->pid->tasks[type]); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 275 | } |
| 276 | |
Oleg Nesterov | 24336eae | 2008-04-30 00:54:26 -0700 | [diff] [blame] | 277 | static void __change_pid(struct task_struct *task, enum pid_type type, |
| 278 | struct pid *new) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 279 | { |
Eric W. Biederman | 92476d7 | 2006-03-31 02:31:42 -0800 | [diff] [blame] | 280 | struct pid_link *link; |
| 281 | struct pid *pid; |
| 282 | int tmp; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 283 | |
Eric W. Biederman | 92476d7 | 2006-03-31 02:31:42 -0800 | [diff] [blame] | 284 | link = &task->pids[type]; |
| 285 | pid = link->pid; |
| 286 | |
| 287 | hlist_del_rcu(&link->node); |
Oleg Nesterov | 24336eae | 2008-04-30 00:54:26 -0700 | [diff] [blame] | 288 | link->pid = new; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 289 | |
| 290 | for (tmp = PIDTYPE_MAX; --tmp >= 0; ) |
Eric W. Biederman | 92476d7 | 2006-03-31 02:31:42 -0800 | [diff] [blame] | 291 | if (!hlist_empty(&pid->tasks[tmp])) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 292 | return; |
| 293 | |
Eric W. Biederman | 92476d7 | 2006-03-31 02:31:42 -0800 | [diff] [blame] | 294 | free_pid(pid); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 295 | } |
| 296 | |
Oleg Nesterov | 24336eae | 2008-04-30 00:54:26 -0700 | [diff] [blame] | 297 | void detach_pid(struct task_struct *task, enum pid_type type) |
| 298 | { |
| 299 | __change_pid(task, type, NULL); |
| 300 | } |
| 301 | |
| 302 | void change_pid(struct task_struct *task, enum pid_type type, |
| 303 | struct pid *pid) |
| 304 | { |
| 305 | __change_pid(task, type, pid); |
Oleg Nesterov | 8190773 | 2013-07-03 15:08:31 -0700 | [diff] [blame] | 306 | attach_pid(task, type); |
Oleg Nesterov | 24336eae | 2008-04-30 00:54:26 -0700 | [diff] [blame] | 307 | } |
| 308 | |
Eric W. Biederman | c18258c | 2006-09-27 01:51:06 -0700 | [diff] [blame] | 309 | /* transfer_pid is an optimization of attach_pid(new), detach_pid(old) */ |
Harvey Harrison | 7ad5b3a | 2008-02-08 04:19:53 -0800 | [diff] [blame] | 310 | void transfer_pid(struct task_struct *old, struct task_struct *new, |
Eric W. Biederman | c18258c | 2006-09-27 01:51:06 -0700 | [diff] [blame] | 311 | enum pid_type type) |
| 312 | { |
| 313 | new->pids[type].pid = old->pids[type].pid; |
| 314 | hlist_replace_rcu(&old->pids[type].node, &new->pids[type].node); |
Eric W. Biederman | c18258c | 2006-09-27 01:51:06 -0700 | [diff] [blame] | 315 | } |
| 316 | |
Harvey Harrison | 7ad5b3a | 2008-02-08 04:19:53 -0800 | [diff] [blame] | 317 | struct task_struct *pid_task(struct pid *pid, enum pid_type type) |
Eric W. Biederman | 92476d7 | 2006-03-31 02:31:42 -0800 | [diff] [blame] | 318 | { |
| 319 | struct task_struct *result = NULL; |
| 320 | if (pid) { |
| 321 | struct hlist_node *first; |
Arnd Bergmann | 67bdbff | 2010-02-25 16:55:13 +0100 | [diff] [blame] | 322 | first = rcu_dereference_check(hlist_first_rcu(&pid->tasks[type]), |
Paul E. McKenney | db1466b | 2010-03-03 07:46:56 -0800 | [diff] [blame] | 323 | lockdep_tasklist_lock_is_held()); |
Eric W. Biederman | 92476d7 | 2006-03-31 02:31:42 -0800 | [diff] [blame] | 324 | if (first) |
| 325 | result = hlist_entry(first, struct task_struct, pids[(type)].node); |
| 326 | } |
| 327 | return result; |
| 328 | } |
Pavel Emelyanov | eccba06 | 2008-02-07 00:13:21 -0800 | [diff] [blame] | 329 | EXPORT_SYMBOL(pid_task); |
Eric W. Biederman | 92476d7 | 2006-03-31 02:31:42 -0800 | [diff] [blame] | 330 | |
| 331 | /* |
Tetsuo Handa | 9728e5d | 2010-03-05 13:42:56 -0800 | [diff] [blame] | 332 | * Must be called under rcu_read_lock(). |
Eric W. Biederman | 92476d7 | 2006-03-31 02:31:42 -0800 | [diff] [blame] | 333 | */ |
Christoph Hellwig | 17f98dc | 2009-06-17 16:27:51 -0700 | [diff] [blame] | 334 | struct task_struct *find_task_by_pid_ns(pid_t nr, struct pid_namespace *ns) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 335 | { |
Paul E. McKenney | f78f5b9 | 2015-06-18 15:50:02 -0700 | [diff] [blame] | 336 | RCU_LOCKDEP_WARN(!rcu_read_lock_held(), |
| 337 | "find_task_by_pid_ns() needs rcu_read_lock() protection"); |
Christoph Hellwig | 17f98dc | 2009-06-17 16:27:51 -0700 | [diff] [blame] | 338 | return pid_task(find_pid_ns(nr, ns), PIDTYPE_PID); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 339 | } |
| 340 | |
Pavel Emelyanov | 228ebcb | 2007-10-18 23:40:16 -0700 | [diff] [blame] | 341 | struct task_struct *find_task_by_vpid(pid_t vnr) |
| 342 | { |
Eric W. Biederman | 17cf22c | 2010-03-02 14:51:53 -0800 | [diff] [blame] | 343 | return find_task_by_pid_ns(vnr, task_active_pid_ns(current)); |
Pavel Emelyanov | 228ebcb | 2007-10-18 23:40:16 -0700 | [diff] [blame] | 344 | } |
Pavel Emelyanov | 228ebcb | 2007-10-18 23:40:16 -0700 | [diff] [blame] | 345 | |
Mike Rapoport | 2ee0826 | 2018-02-06 15:40:17 -0800 | [diff] [blame] | 346 | struct task_struct *find_get_task_by_vpid(pid_t nr) |
| 347 | { |
| 348 | struct task_struct *task; |
| 349 | |
| 350 | rcu_read_lock(); |
| 351 | task = find_task_by_vpid(nr); |
| 352 | if (task) |
| 353 | get_task_struct(task); |
| 354 | rcu_read_unlock(); |
| 355 | |
| 356 | return task; |
| 357 | } |
| 358 | |
Oleg Nesterov | 1a657f78 | 2006-10-02 02:18:59 -0700 | [diff] [blame] | 359 | struct pid *get_task_pid(struct task_struct *task, enum pid_type type) |
| 360 | { |
| 361 | struct pid *pid; |
| 362 | rcu_read_lock(); |
Oleg Nesterov | 2ae448e | 2009-04-02 16:58:36 -0700 | [diff] [blame] | 363 | if (type != PIDTYPE_PID) |
| 364 | task = task->group_leader; |
Eric Dumazet | 81b1a83 | 2015-11-24 11:39:54 -0800 | [diff] [blame] | 365 | pid = get_pid(rcu_dereference(task->pids[type].pid)); |
Oleg Nesterov | 1a657f78 | 2006-10-02 02:18:59 -0700 | [diff] [blame] | 366 | rcu_read_unlock(); |
| 367 | return pid; |
| 368 | } |
Rik van Riel | 77c100c | 2011-02-01 09:51:46 -0500 | [diff] [blame] | 369 | EXPORT_SYMBOL_GPL(get_task_pid); |
Oleg Nesterov | 1a657f78 | 2006-10-02 02:18:59 -0700 | [diff] [blame] | 370 | |
Harvey Harrison | 7ad5b3a | 2008-02-08 04:19:53 -0800 | [diff] [blame] | 371 | struct task_struct *get_pid_task(struct pid *pid, enum pid_type type) |
Eric W. Biederman | 92476d7 | 2006-03-31 02:31:42 -0800 | [diff] [blame] | 372 | { |
| 373 | struct task_struct *result; |
| 374 | rcu_read_lock(); |
| 375 | result = pid_task(pid, type); |
| 376 | if (result) |
| 377 | get_task_struct(result); |
| 378 | rcu_read_unlock(); |
| 379 | return result; |
| 380 | } |
Rik van Riel | 77c100c | 2011-02-01 09:51:46 -0500 | [diff] [blame] | 381 | EXPORT_SYMBOL_GPL(get_pid_task); |
Eric W. Biederman | 92476d7 | 2006-03-31 02:31:42 -0800 | [diff] [blame] | 382 | |
| 383 | struct pid *find_get_pid(pid_t nr) |
| 384 | { |
| 385 | struct pid *pid; |
| 386 | |
| 387 | rcu_read_lock(); |
Pavel Emelyanov | 198fe21 | 2007-10-18 23:40:06 -0700 | [diff] [blame] | 388 | pid = get_pid(find_vpid(nr)); |
Eric W. Biederman | 92476d7 | 2006-03-31 02:31:42 -0800 | [diff] [blame] | 389 | rcu_read_unlock(); |
| 390 | |
| 391 | return pid; |
| 392 | } |
David Sterba | 339caf2 | 2008-07-25 01:48:31 -0700 | [diff] [blame] | 393 | EXPORT_SYMBOL_GPL(find_get_pid); |
Eric W. Biederman | 92476d7 | 2006-03-31 02:31:42 -0800 | [diff] [blame] | 394 | |
Pavel Emelyanov | 7af5729 | 2007-10-18 23:40:06 -0700 | [diff] [blame] | 395 | pid_t pid_nr_ns(struct pid *pid, struct pid_namespace *ns) |
| 396 | { |
| 397 | struct upid *upid; |
| 398 | pid_t nr = 0; |
| 399 | |
| 400 | if (pid && ns->level <= pid->level) { |
| 401 | upid = &pid->numbers[ns->level]; |
| 402 | if (upid->ns == ns) |
| 403 | nr = upid->nr; |
| 404 | } |
| 405 | return nr; |
| 406 | } |
Eric W. Biederman | 4f82f45 | 2012-05-24 10:37:59 -0600 | [diff] [blame] | 407 | EXPORT_SYMBOL_GPL(pid_nr_ns); |
Pavel Emelyanov | 7af5729 | 2007-10-18 23:40:06 -0700 | [diff] [blame] | 408 | |
Eric W. Biederman | 44c4e1b | 2008-02-08 04:19:15 -0800 | [diff] [blame] | 409 | pid_t pid_vnr(struct pid *pid) |
| 410 | { |
Eric W. Biederman | 17cf22c | 2010-03-02 14:51:53 -0800 | [diff] [blame] | 411 | return pid_nr_ns(pid, task_active_pid_ns(current)); |
Eric W. Biederman | 44c4e1b | 2008-02-08 04:19:15 -0800 | [diff] [blame] | 412 | } |
| 413 | EXPORT_SYMBOL_GPL(pid_vnr); |
| 414 | |
Oleg Nesterov | 52ee2df | 2009-04-02 16:58:38 -0700 | [diff] [blame] | 415 | pid_t __task_pid_nr_ns(struct task_struct *task, enum pid_type type, |
| 416 | struct pid_namespace *ns) |
Pavel Emelyanov | 2f2a3a4 | 2007-10-18 23:40:19 -0700 | [diff] [blame] | 417 | { |
Oleg Nesterov | 52ee2df | 2009-04-02 16:58:38 -0700 | [diff] [blame] | 418 | pid_t nr = 0; |
| 419 | |
| 420 | rcu_read_lock(); |
| 421 | if (!ns) |
Eric W. Biederman | 17cf22c | 2010-03-02 14:51:53 -0800 | [diff] [blame] | 422 | ns = task_active_pid_ns(current); |
Oleg Nesterov | 52ee2df | 2009-04-02 16:58:38 -0700 | [diff] [blame] | 423 | if (likely(pid_alive(task))) { |
Oleg Nesterov | dd1c1f2 | 2017-08-21 17:35:02 +0200 | [diff] [blame] | 424 | if (type != PIDTYPE_PID) { |
| 425 | if (type == __PIDTYPE_TGID) |
| 426 | type = PIDTYPE_PID; |
Gargi Sharma | e8cfbc2 | 2017-11-17 15:30:34 -0800 | [diff] [blame] | 427 | |
Oleg Nesterov | 52ee2df | 2009-04-02 16:58:38 -0700 | [diff] [blame] | 428 | task = task->group_leader; |
Oleg Nesterov | dd1c1f2 | 2017-08-21 17:35:02 +0200 | [diff] [blame] | 429 | } |
Eric Dumazet | 81b1a83 | 2015-11-24 11:39:54 -0800 | [diff] [blame] | 430 | nr = pid_nr_ns(rcu_dereference(task->pids[type].pid), ns); |
Oleg Nesterov | 52ee2df | 2009-04-02 16:58:38 -0700 | [diff] [blame] | 431 | } |
| 432 | rcu_read_unlock(); |
| 433 | |
| 434 | return nr; |
Pavel Emelyanov | 2f2a3a4 | 2007-10-18 23:40:19 -0700 | [diff] [blame] | 435 | } |
Oleg Nesterov | 52ee2df | 2009-04-02 16:58:38 -0700 | [diff] [blame] | 436 | EXPORT_SYMBOL(__task_pid_nr_ns); |
Pavel Emelyanov | 2f2a3a4 | 2007-10-18 23:40:19 -0700 | [diff] [blame] | 437 | |
Eric W. Biederman | 61bce0f | 2009-01-07 18:08:49 -0800 | [diff] [blame] | 438 | struct pid_namespace *task_active_pid_ns(struct task_struct *tsk) |
| 439 | { |
| 440 | return ns_of_pid(task_pid(tsk)); |
| 441 | } |
| 442 | EXPORT_SYMBOL_GPL(task_active_pid_ns); |
| 443 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 444 | /* |
Frederik Schwarzer | 025dfda | 2008-10-16 19:02:37 +0200 | [diff] [blame] | 445 | * Used by proc to find the first pid that is greater than or equal to nr. |
Eric W. Biederman | 0804ef4 | 2006-10-02 02:17:04 -0700 | [diff] [blame] | 446 | * |
Pavel Emelyanov | e49859e | 2008-07-25 01:48:36 -0700 | [diff] [blame] | 447 | * If there is a pid at nr this function is exactly the same as find_pid_ns. |
Eric W. Biederman | 0804ef4 | 2006-10-02 02:17:04 -0700 | [diff] [blame] | 448 | */ |
Pavel Emelyanov | 198fe21 | 2007-10-18 23:40:06 -0700 | [diff] [blame] | 449 | struct pid *find_ge_pid(int nr, struct pid_namespace *ns) |
Eric W. Biederman | 0804ef4 | 2006-10-02 02:17:04 -0700 | [diff] [blame] | 450 | { |
Gargi Sharma | 95846ec | 2017-11-17 15:30:30 -0800 | [diff] [blame] | 451 | return idr_get_next(&ns->idr, &nr); |
Eric W. Biederman | 0804ef4 | 2006-10-02 02:17:04 -0700 | [diff] [blame] | 452 | } |
| 453 | |
Gargi Sharma | 95846ec | 2017-11-17 15:30:30 -0800 | [diff] [blame] | 454 | void __init pid_idr_init(void) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 455 | { |
Zhen Lei | 840d6fe | 2016-01-30 10:04:17 +0800 | [diff] [blame] | 456 | /* Verify no one has done anything silly: */ |
Gargi Sharma | e8cfbc2 | 2017-11-17 15:30:34 -0800 | [diff] [blame] | 457 | BUILD_BUG_ON(PID_MAX_LIMIT >= PIDNS_ADDING); |
Eric W. Biederman | c876ad76 | 2012-12-21 20:27:12 -0800 | [diff] [blame] | 458 | |
Hedi Berriche | 72680a1 | 2010-05-26 14:44:06 -0700 | [diff] [blame] | 459 | /* bump default and minimum pid_max based on number of cpus */ |
| 460 | pid_max = min(pid_max_max, max_t(int, pid_max, |
| 461 | PIDS_PER_CPU_DEFAULT * num_possible_cpus())); |
| 462 | pid_max_min = max_t(int, pid_max_min, |
| 463 | PIDS_PER_CPU_MIN * num_possible_cpus()); |
| 464 | pr_info("pid_max: default: %u minimum: %u\n", pid_max, pid_max_min); |
| 465 | |
Gargi Sharma | 95846ec | 2017-11-17 15:30:30 -0800 | [diff] [blame] | 466 | idr_init(&init_pid_ns.idr); |
Eric W. Biederman | 92476d7 | 2006-03-31 02:31:42 -0800 | [diff] [blame] | 467 | |
Pavel Emelyanov | 74bd59b | 2008-02-08 04:18:24 -0800 | [diff] [blame] | 468 | init_pid_ns.pid_cachep = KMEM_CACHE(pid, |
Vladimir Davydov | 5d09705 | 2016-01-14 15:18:21 -0800 | [diff] [blame] | 469 | SLAB_HWCACHE_ALIGN | SLAB_PANIC | SLAB_ACCOUNT); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 470 | } |