xref: /linux/kernel/pid.c (revision b0148a98ec5151fec82064d95f11eb9efbc628ea)
1 /*
2  * Generic pidhash and scalable, time-bounded PID allocator
3  *
4  * (C) 2002-2003 William Irwin, IBM
5  * (C) 2004 William Irwin, Oracle
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).
21  */
22 
23 #include <linux/mm.h>
24 #include <linux/module.h>
25 #include <linux/slab.h>
26 #include <linux/init.h>
27 #include <linux/bootmem.h>
28 #include <linux/hash.h>
29 #include <linux/pid_namespace.h>
30 
31 #define pid_hashfn(nr) hash_long((unsigned long)nr, pidhash_shift)
32 static struct hlist_head *pid_hash;
33 static int pidhash_shift;
34 static struct kmem_cache *pid_cachep;
35 
36 int pid_max = PID_MAX_DEFAULT;
37 
38 #define RESERVED_PIDS		300
39 
40 int pid_max_min = RESERVED_PIDS + 1;
41 int pid_max_max = PID_MAX_LIMIT;
42 
43 #define BITS_PER_PAGE		(PAGE_SIZE*8)
44 #define BITS_PER_PAGE_MASK	(BITS_PER_PAGE-1)
45 
46 static inline int mk_pid(struct pid_namespace *pid_ns,
47 		struct pidmap *map, int off)
48 {
49 	return (map - pid_ns->pidmap)*BITS_PER_PAGE + off;
50 }
51 
52 #define find_next_offset(map, off)					\
53 		find_next_zero_bit((map)->page, BITS_PER_PAGE, off)
54 
55 /*
56  * PID-map pages start out as NULL, they get allocated upon
57  * first use and are never deallocated. This way a low pid_max
58  * value does not cause lots of bitmaps to be allocated, but
59  * the scheme scales to up to 4 million PIDs, runtime.
60  */
61 struct pid_namespace init_pid_ns = {
62 	.kref = {
63 		.refcount       = ATOMIC_INIT(2),
64 	},
65 	.pidmap = {
66 		[ 0 ... PIDMAP_ENTRIES-1] = { ATOMIC_INIT(BITS_PER_PAGE), NULL }
67 	},
68 	.last_pid = 0,
69 	.child_reaper = &init_task
70 };
71 
72 /*
73  * Note: disable interrupts while the pidmap_lock is held as an
74  * interrupt might come in and do read_lock(&tasklist_lock).
75  *
76  * If we don't disable interrupts there is a nasty deadlock between
77  * detach_pid()->free_pid() and another cpu that does
78  * spin_lock(&pidmap_lock) followed by an interrupt routine that does
79  * read_lock(&tasklist_lock);
80  *
81  * After we clean up the tasklist_lock and know there are no
82  * irq handlers that take it we can leave the interrupts enabled.
83  * For now it is easier to be safe than to prove it can't happen.
84  */
85 
86 static  __cacheline_aligned_in_smp DEFINE_SPINLOCK(pidmap_lock);
87 
88 static fastcall void free_pidmap(struct pid_namespace *pid_ns, int pid)
89 {
90 	struct pidmap *map = pid_ns->pidmap + pid / BITS_PER_PAGE;
91 	int offset = pid & BITS_PER_PAGE_MASK;
92 
93 	clear_bit(offset, map->page);
94 	atomic_inc(&map->nr_free);
95 }
96 
97 static int alloc_pidmap(struct pid_namespace *pid_ns)
98 {
99 	int i, offset, max_scan, pid, last = pid_ns->last_pid;
100 	struct pidmap *map;
101 
102 	pid = last + 1;
103 	if (pid >= pid_max)
104 		pid = RESERVED_PIDS;
105 	offset = pid & BITS_PER_PAGE_MASK;
106 	map = &pid_ns->pidmap[pid/BITS_PER_PAGE];
107 	max_scan = (pid_max + BITS_PER_PAGE - 1)/BITS_PER_PAGE - !offset;
108 	for (i = 0; i <= max_scan; ++i) {
109 		if (unlikely(!map->page)) {
110 			void *page = kzalloc(PAGE_SIZE, GFP_KERNEL);
111 			/*
112 			 * Free the page if someone raced with us
113 			 * installing it:
114 			 */
115 			spin_lock_irq(&pidmap_lock);
116 			if (map->page)
117 				kfree(page);
118 			else
119 				map->page = page;
120 			spin_unlock_irq(&pidmap_lock);
121 			if (unlikely(!map->page))
122 				break;
123 		}
124 		if (likely(atomic_read(&map->nr_free))) {
125 			do {
126 				if (!test_and_set_bit(offset, map->page)) {
127 					atomic_dec(&map->nr_free);
128 					pid_ns->last_pid = pid;
129 					return pid;
130 				}
131 				offset = find_next_offset(map, offset);
132 				pid = mk_pid(pid_ns, map, offset);
133 			/*
134 			 * find_next_offset() found a bit, the pid from it
135 			 * is in-bounds, and if we fell back to the last
136 			 * bitmap block and the final block was the same
137 			 * as the starting point, pid is before last_pid.
138 			 */
139 			} while (offset < BITS_PER_PAGE && pid < pid_max &&
140 					(i != max_scan || pid < last ||
141 					    !((last+1) & BITS_PER_PAGE_MASK)));
142 		}
143 		if (map < &pid_ns->pidmap[(pid_max-1)/BITS_PER_PAGE]) {
144 			++map;
145 			offset = 0;
146 		} else {
147 			map = &pid_ns->pidmap[0];
148 			offset = RESERVED_PIDS;
149 			if (unlikely(last == offset))
150 				break;
151 		}
152 		pid = mk_pid(pid_ns, map, offset);
153 	}
154 	return -1;
155 }
156 
157 static int next_pidmap(struct pid_namespace *pid_ns, int last)
158 {
159 	int offset;
160 	struct pidmap *map, *end;
161 
162 	offset = (last + 1) & BITS_PER_PAGE_MASK;
163 	map = &pid_ns->pidmap[(last + 1)/BITS_PER_PAGE];
164 	end = &pid_ns->pidmap[PIDMAP_ENTRIES];
165 	for (; map < end; map++, offset = 0) {
166 		if (unlikely(!map->page))
167 			continue;
168 		offset = find_next_bit((map)->page, BITS_PER_PAGE, offset);
169 		if (offset < BITS_PER_PAGE)
170 			return mk_pid(pid_ns, map, offset);
171 	}
172 	return -1;
173 }
174 
175 fastcall void put_pid(struct pid *pid)
176 {
177 	if (!pid)
178 		return;
179 	if ((atomic_read(&pid->count) == 1) ||
180 	     atomic_dec_and_test(&pid->count))
181 		kmem_cache_free(pid_cachep, pid);
182 }
183 EXPORT_SYMBOL_GPL(put_pid);
184 
185 static void delayed_put_pid(struct rcu_head *rhp)
186 {
187 	struct pid *pid = container_of(rhp, struct pid, rcu);
188 	put_pid(pid);
189 }
190 
191 fastcall void free_pid(struct pid *pid)
192 {
193 	/* We can be called with write_lock_irq(&tasklist_lock) held */
194 	unsigned long flags;
195 
196 	spin_lock_irqsave(&pidmap_lock, flags);
197 	hlist_del_rcu(&pid->pid_chain);
198 	spin_unlock_irqrestore(&pidmap_lock, flags);
199 
200 	free_pidmap(&init_pid_ns, pid->nr);
201 	call_rcu(&pid->rcu, delayed_put_pid);
202 }
203 
204 struct pid *alloc_pid(void)
205 {
206 	struct pid *pid;
207 	enum pid_type type;
208 	int nr = -1;
209 
210 	pid = kmem_cache_alloc(pid_cachep, GFP_KERNEL);
211 	if (!pid)
212 		goto out;
213 
214 	nr = alloc_pidmap(current->nsproxy->pid_ns);
215 	if (nr < 0)
216 		goto out_free;
217 
218 	atomic_set(&pid->count, 1);
219 	pid->nr = nr;
220 	for (type = 0; type < PIDTYPE_MAX; ++type)
221 		INIT_HLIST_HEAD(&pid->tasks[type]);
222 
223 	spin_lock_irq(&pidmap_lock);
224 	hlist_add_head_rcu(&pid->pid_chain, &pid_hash[pid_hashfn(pid->nr)]);
225 	spin_unlock_irq(&pidmap_lock);
226 
227 out:
228 	return pid;
229 
230 out_free:
231 	kmem_cache_free(pid_cachep, pid);
232 	pid = NULL;
233 	goto out;
234 }
235 
236 struct pid * fastcall find_pid(int nr)
237 {
238 	struct hlist_node *elem;
239 	struct pid *pid;
240 
241 	hlist_for_each_entry_rcu(pid, elem,
242 			&pid_hash[pid_hashfn(nr)], pid_chain) {
243 		if (pid->nr == nr)
244 			return pid;
245 	}
246 	return NULL;
247 }
248 EXPORT_SYMBOL_GPL(find_pid);
249 
250 int fastcall attach_pid(struct task_struct *task, enum pid_type type, int nr)
251 {
252 	struct pid_link *link;
253 	struct pid *pid;
254 
255 	link = &task->pids[type];
256 	link->pid = pid = find_pid(nr);
257 	hlist_add_head_rcu(&link->node, &pid->tasks[type]);
258 
259 	return 0;
260 }
261 
262 void fastcall detach_pid(struct task_struct *task, enum pid_type type)
263 {
264 	struct pid_link *link;
265 	struct pid *pid;
266 	int tmp;
267 
268 	link = &task->pids[type];
269 	pid = link->pid;
270 
271 	hlist_del_rcu(&link->node);
272 	link->pid = NULL;
273 
274 	for (tmp = PIDTYPE_MAX; --tmp >= 0; )
275 		if (!hlist_empty(&pid->tasks[tmp]))
276 			return;
277 
278 	free_pid(pid);
279 }
280 
281 /* transfer_pid is an optimization of attach_pid(new), detach_pid(old) */
282 void fastcall transfer_pid(struct task_struct *old, struct task_struct *new,
283 			   enum pid_type type)
284 {
285 	new->pids[type].pid = old->pids[type].pid;
286 	hlist_replace_rcu(&old->pids[type].node, &new->pids[type].node);
287 	old->pids[type].pid = NULL;
288 }
289 
290 struct task_struct * fastcall pid_task(struct pid *pid, enum pid_type type)
291 {
292 	struct task_struct *result = NULL;
293 	if (pid) {
294 		struct hlist_node *first;
295 		first = rcu_dereference(pid->tasks[type].first);
296 		if (first)
297 			result = hlist_entry(first, struct task_struct, pids[(type)].node);
298 	}
299 	return result;
300 }
301 
302 /*
303  * Must be called under rcu_read_lock() or with tasklist_lock read-held.
304  */
305 struct task_struct *find_task_by_pid_type(int type, int nr)
306 {
307 	return pid_task(find_pid(nr), type);
308 }
309 
310 EXPORT_SYMBOL(find_task_by_pid_type);
311 
312 struct pid *get_task_pid(struct task_struct *task, enum pid_type type)
313 {
314 	struct pid *pid;
315 	rcu_read_lock();
316 	pid = get_pid(task->pids[type].pid);
317 	rcu_read_unlock();
318 	return pid;
319 }
320 
321 struct task_struct *fastcall get_pid_task(struct pid *pid, enum pid_type type)
322 {
323 	struct task_struct *result;
324 	rcu_read_lock();
325 	result = pid_task(pid, type);
326 	if (result)
327 		get_task_struct(result);
328 	rcu_read_unlock();
329 	return result;
330 }
331 
332 struct pid *find_get_pid(pid_t nr)
333 {
334 	struct pid *pid;
335 
336 	rcu_read_lock();
337 	pid = get_pid(find_pid(nr));
338 	rcu_read_unlock();
339 
340 	return pid;
341 }
342 
343 /*
344  * Used by proc to find the first pid that is greater then or equal to nr.
345  *
346  * If there is a pid at nr this function is exactly the same as find_pid.
347  */
348 struct pid *find_ge_pid(int nr)
349 {
350 	struct pid *pid;
351 
352 	do {
353 		pid = find_pid(nr);
354 		if (pid)
355 			break;
356 		nr = next_pidmap(current->nsproxy->pid_ns, nr);
357 	} while (nr > 0);
358 
359 	return pid;
360 }
361 EXPORT_SYMBOL_GPL(find_get_pid);
362 
363 int copy_pid_ns(int flags, struct task_struct *tsk)
364 {
365 	struct pid_namespace *old_ns = tsk->nsproxy->pid_ns;
366 	int err = 0;
367 
368 	if (!old_ns)
369 		return 0;
370 
371 	get_pid_ns(old_ns);
372 	return err;
373 }
374 
375 void free_pid_ns(struct kref *kref)
376 {
377 	struct pid_namespace *ns;
378 
379 	ns = container_of(kref, struct pid_namespace, kref);
380 	kfree(ns);
381 }
382 
383 /*
384  * The pid hash table is scaled according to the amount of memory in the
385  * machine.  From a minimum of 16 slots up to 4096 slots at one gigabyte or
386  * more.
387  */
388 void __init pidhash_init(void)
389 {
390 	int i, pidhash_size;
391 	unsigned long megabytes = nr_kernel_pages >> (20 - PAGE_SHIFT);
392 
393 	pidhash_shift = max(4, fls(megabytes * 4));
394 	pidhash_shift = min(12, pidhash_shift);
395 	pidhash_size = 1 << pidhash_shift;
396 
397 	printk("PID hash table entries: %d (order: %d, %Zd bytes)\n",
398 		pidhash_size, pidhash_shift,
399 		pidhash_size * sizeof(struct hlist_head));
400 
401 	pid_hash = alloc_bootmem(pidhash_size *	sizeof(*(pid_hash)));
402 	if (!pid_hash)
403 		panic("Could not alloc pidhash!\n");
404 	for (i = 0; i < pidhash_size; i++)
405 		INIT_HLIST_HEAD(&pid_hash[i]);
406 }
407 
408 void __init pidmap_init(void)
409 {
410 	init_pid_ns.pidmap[0].page = kzalloc(PAGE_SIZE, GFP_KERNEL);
411 	/* Reserve PID 0. We never call free_pidmap(0) */
412 	set_bit(0, init_pid_ns.pidmap[0].page);
413 	atomic_dec(&init_pid_ns.pidmap[0].nr_free);
414 
415 	pid_cachep = kmem_cache_create("pid", sizeof(struct pid),
416 					__alignof__(struct pid),
417 					SLAB_PANIC, NULL, NULL);
418 }
419