xref: /openbmc/linux/kernel/pid.c (revision 1da177e4c3f41524e886b7f1b8a0c1fc7321cac2)
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 
30 #define pid_hashfn(nr) hash_long((unsigned long)nr, pidhash_shift)
31 static struct hlist_head *pid_hash[PIDTYPE_MAX];
32 static int pidhash_shift;
33 
34 int pid_max = PID_MAX_DEFAULT;
35 int last_pid;
36 
37 #define RESERVED_PIDS		300
38 
39 int pid_max_min = RESERVED_PIDS + 1;
40 int pid_max_max = PID_MAX_LIMIT;
41 
42 #define PIDMAP_ENTRIES		((PID_MAX_LIMIT + 8*PAGE_SIZE - 1)/PAGE_SIZE/8)
43 #define BITS_PER_PAGE		(PAGE_SIZE*8)
44 #define BITS_PER_PAGE_MASK	(BITS_PER_PAGE-1)
45 #define mk_pid(map, off)	(((map) - pidmap_array)*BITS_PER_PAGE + (off))
46 #define find_next_offset(map, off)					\
47 		find_next_zero_bit((map)->page, BITS_PER_PAGE, off)
48 
49 /*
50  * PID-map pages start out as NULL, they get allocated upon
51  * first use and are never deallocated. This way a low pid_max
52  * value does not cause lots of bitmaps to be allocated, but
53  * the scheme scales to up to 4 million PIDs, runtime.
54  */
55 typedef struct pidmap {
56 	atomic_t nr_free;
57 	void *page;
58 } pidmap_t;
59 
60 static pidmap_t pidmap_array[PIDMAP_ENTRIES] =
61 	 { [ 0 ... PIDMAP_ENTRIES-1 ] = { ATOMIC_INIT(BITS_PER_PAGE), NULL } };
62 
63 static  __cacheline_aligned_in_smp DEFINE_SPINLOCK(pidmap_lock);
64 
65 fastcall void free_pidmap(int pid)
66 {
67 	pidmap_t *map = pidmap_array + pid / BITS_PER_PAGE;
68 	int offset = pid & BITS_PER_PAGE_MASK;
69 
70 	clear_bit(offset, map->page);
71 	atomic_inc(&map->nr_free);
72 }
73 
74 int alloc_pidmap(void)
75 {
76 	int i, offset, max_scan, pid, last = last_pid;
77 	pidmap_t *map;
78 
79 	pid = last + 1;
80 	if (pid >= pid_max)
81 		pid = RESERVED_PIDS;
82 	offset = pid & BITS_PER_PAGE_MASK;
83 	map = &pidmap_array[pid/BITS_PER_PAGE];
84 	max_scan = (pid_max + BITS_PER_PAGE - 1)/BITS_PER_PAGE - !offset;
85 	for (i = 0; i <= max_scan; ++i) {
86 		if (unlikely(!map->page)) {
87 			unsigned long page = get_zeroed_page(GFP_KERNEL);
88 			/*
89 			 * Free the page if someone raced with us
90 			 * installing it:
91 			 */
92 			spin_lock(&pidmap_lock);
93 			if (map->page)
94 				free_page(page);
95 			else
96 				map->page = (void *)page;
97 			spin_unlock(&pidmap_lock);
98 			if (unlikely(!map->page))
99 				break;
100 		}
101 		if (likely(atomic_read(&map->nr_free))) {
102 			do {
103 				if (!test_and_set_bit(offset, map->page)) {
104 					atomic_dec(&map->nr_free);
105 					last_pid = pid;
106 					return pid;
107 				}
108 				offset = find_next_offset(map, offset);
109 				pid = mk_pid(map, offset);
110 			/*
111 			 * find_next_offset() found a bit, the pid from it
112 			 * is in-bounds, and if we fell back to the last
113 			 * bitmap block and the final block was the same
114 			 * as the starting point, pid is before last_pid.
115 			 */
116 			} while (offset < BITS_PER_PAGE && pid < pid_max &&
117 					(i != max_scan || pid < last ||
118 					    !((last+1) & BITS_PER_PAGE_MASK)));
119 		}
120 		if (map < &pidmap_array[(pid_max-1)/BITS_PER_PAGE]) {
121 			++map;
122 			offset = 0;
123 		} else {
124 			map = &pidmap_array[0];
125 			offset = RESERVED_PIDS;
126 			if (unlikely(last == offset))
127 				break;
128 		}
129 		pid = mk_pid(map, offset);
130 	}
131 	return -1;
132 }
133 
134 struct pid * fastcall find_pid(enum pid_type type, int nr)
135 {
136 	struct hlist_node *elem;
137 	struct pid *pid;
138 
139 	hlist_for_each_entry(pid, elem,
140 			&pid_hash[type][pid_hashfn(nr)], pid_chain) {
141 		if (pid->nr == nr)
142 			return pid;
143 	}
144 	return NULL;
145 }
146 
147 int fastcall attach_pid(task_t *task, enum pid_type type, int nr)
148 {
149 	struct pid *pid, *task_pid;
150 
151 	task_pid = &task->pids[type];
152 	pid = find_pid(type, nr);
153 	if (pid == NULL) {
154 		hlist_add_head(&task_pid->pid_chain,
155 				&pid_hash[type][pid_hashfn(nr)]);
156 		INIT_LIST_HEAD(&task_pid->pid_list);
157 	} else {
158 		INIT_HLIST_NODE(&task_pid->pid_chain);
159 		list_add_tail(&task_pid->pid_list, &pid->pid_list);
160 	}
161 	task_pid->nr = nr;
162 
163 	return 0;
164 }
165 
166 static fastcall int __detach_pid(task_t *task, enum pid_type type)
167 {
168 	struct pid *pid, *pid_next;
169 	int nr = 0;
170 
171 	pid = &task->pids[type];
172 	if (!hlist_unhashed(&pid->pid_chain)) {
173 		hlist_del(&pid->pid_chain);
174 
175 		if (list_empty(&pid->pid_list))
176 			nr = pid->nr;
177 		else {
178 			pid_next = list_entry(pid->pid_list.next,
179 						struct pid, pid_list);
180 			/* insert next pid from pid_list to hash */
181 			hlist_add_head(&pid_next->pid_chain,
182 				&pid_hash[type][pid_hashfn(pid_next->nr)]);
183 		}
184 	}
185 
186 	list_del(&pid->pid_list);
187 	pid->nr = 0;
188 
189 	return nr;
190 }
191 
192 void fastcall detach_pid(task_t *task, enum pid_type type)
193 {
194 	int tmp, nr;
195 
196 	nr = __detach_pid(task, type);
197 	if (!nr)
198 		return;
199 
200 	for (tmp = PIDTYPE_MAX; --tmp >= 0; )
201 		if (tmp != type && find_pid(tmp, nr))
202 			return;
203 
204 	free_pidmap(nr);
205 }
206 
207 task_t *find_task_by_pid_type(int type, int nr)
208 {
209 	struct pid *pid;
210 
211 	pid = find_pid(type, nr);
212 	if (!pid)
213 		return NULL;
214 
215 	return pid_task(&pid->pid_list, type);
216 }
217 
218 EXPORT_SYMBOL(find_task_by_pid_type);
219 
220 /*
221  * This function switches the PIDs if a non-leader thread calls
222  * sys_execve() - this must be done without releasing the PID.
223  * (which a detach_pid() would eventually do.)
224  */
225 void switch_exec_pids(task_t *leader, task_t *thread)
226 {
227 	__detach_pid(leader, PIDTYPE_PID);
228 	__detach_pid(leader, PIDTYPE_TGID);
229 	__detach_pid(leader, PIDTYPE_PGID);
230 	__detach_pid(leader, PIDTYPE_SID);
231 
232 	__detach_pid(thread, PIDTYPE_PID);
233 	__detach_pid(thread, PIDTYPE_TGID);
234 
235 	leader->pid = leader->tgid = thread->pid;
236 	thread->pid = thread->tgid;
237 
238 	attach_pid(thread, PIDTYPE_PID, thread->pid);
239 	attach_pid(thread, PIDTYPE_TGID, thread->tgid);
240 	attach_pid(thread, PIDTYPE_PGID, thread->signal->pgrp);
241 	attach_pid(thread, PIDTYPE_SID, thread->signal->session);
242 	list_add_tail(&thread->tasks, &init_task.tasks);
243 
244 	attach_pid(leader, PIDTYPE_PID, leader->pid);
245 	attach_pid(leader, PIDTYPE_TGID, leader->tgid);
246 	attach_pid(leader, PIDTYPE_PGID, leader->signal->pgrp);
247 	attach_pid(leader, PIDTYPE_SID, leader->signal->session);
248 }
249 
250 /*
251  * The pid hash table is scaled according to the amount of memory in the
252  * machine.  From a minimum of 16 slots up to 4096 slots at one gigabyte or
253  * more.
254  */
255 void __init pidhash_init(void)
256 {
257 	int i, j, pidhash_size;
258 	unsigned long megabytes = nr_kernel_pages >> (20 - PAGE_SHIFT);
259 
260 	pidhash_shift = max(4, fls(megabytes * 4));
261 	pidhash_shift = min(12, pidhash_shift);
262 	pidhash_size = 1 << pidhash_shift;
263 
264 	printk("PID hash table entries: %d (order: %d, %Zd bytes)\n",
265 		pidhash_size, pidhash_shift,
266 		PIDTYPE_MAX * pidhash_size * sizeof(struct hlist_head));
267 
268 	for (i = 0; i < PIDTYPE_MAX; i++) {
269 		pid_hash[i] = alloc_bootmem(pidhash_size *
270 					sizeof(*(pid_hash[i])));
271 		if (!pid_hash[i])
272 			panic("Could not alloc pidhash!\n");
273 		for (j = 0; j < pidhash_size; j++)
274 			INIT_HLIST_HEAD(&pid_hash[i][j]);
275 	}
276 }
277 
278 void __init pidmap_init(void)
279 {
280 	int i;
281 
282 	pidmap_array->page = (void *)get_zeroed_page(GFP_KERNEL);
283 	set_bit(0, pidmap_array->page);
284 	atomic_dec(&pidmap_array->nr_free);
285 
286 	/*
287 	 * Allocate PID 0, and hash it via all PID types:
288 	 */
289 
290 	for (i = 0; i < PIDTYPE_MAX; i++)
291 		attach_pid(current, i, 0);
292 }
293