1 // SPDX-License-Identifier: GPL-2.0
2 /*
3  * Ldisc rw semaphore
4  *
5  * The ldisc semaphore is semantically a rw_semaphore but which enforces
6  * an alternate policy, namely:
7  *   1) Supports lock wait timeouts
8  *   2) Write waiter has priority
9  *   3) Downgrading is not supported
10  *
11  * Implementation notes:
12  *   1) Upper half of semaphore count is a wait count (differs from rwsem
13  *	in that rwsem normalizes the upper half to the wait bias)
14  *   2) Lacks overflow checking
15  *
16  * The generic counting was copied and modified from include/asm-generic/rwsem.h
17  * by Paul Mackerras <paulus@samba.org>.
18  *
19  * The scheduling policy was copied and modified from lib/rwsem.c
20  * Written by David Howells (dhowells@redhat.com).
21  *
22  * This implementation incorporates the write lock stealing work of
23  * Michel Lespinasse <walken@google.com>.
24  *
25  * Copyright (C) 2013 Peter Hurley <peter@hurleysoftware.com>
26  */
27 
28 #include <linux/list.h>
29 #include <linux/spinlock.h>
30 #include <linux/atomic.h>
31 #include <linux/tty.h>
32 #include <linux/sched.h>
33 #include <linux/sched/debug.h>
34 #include <linux/sched/task.h>
35 
36 
37 #ifdef CONFIG_DEBUG_LOCK_ALLOC
38 # define __acq(l, s, t, r, c, n, i)		\
39 				lock_acquire(&(l)->dep_map, s, t, r, c, n, i)
40 # define __rel(l, n, i)				\
41 				lock_release(&(l)->dep_map, n, i)
42 #define lockdep_acquire(l, s, t, i)		__acq(l, s, t, 0, 1, NULL, i)
43 #define lockdep_acquire_nest(l, s, t, n, i)	__acq(l, s, t, 0, 1, n, i)
44 #define lockdep_acquire_read(l, s, t, i)	__acq(l, s, t, 1, 1, NULL, i)
45 #define lockdep_release(l, n, i)		__rel(l, n, i)
46 #else
47 # define lockdep_acquire(l, s, t, i)		do { } while (0)
48 # define lockdep_acquire_nest(l, s, t, n, i)	do { } while (0)
49 # define lockdep_acquire_read(l, s, t, i)	do { } while (0)
50 # define lockdep_release(l, n, i)		do { } while (0)
51 #endif
52 
53 #ifdef CONFIG_LOCK_STAT
54 # define lock_stat(_lock, stat)		lock_##stat(&(_lock)->dep_map, _RET_IP_)
55 #else
56 # define lock_stat(_lock, stat)		do { } while (0)
57 #endif
58 
59 
60 #if BITS_PER_LONG == 64
61 # define LDSEM_ACTIVE_MASK	0xffffffffL
62 #else
63 # define LDSEM_ACTIVE_MASK	0x0000ffffL
64 #endif
65 
66 #define LDSEM_UNLOCKED		0L
67 #define LDSEM_ACTIVE_BIAS	1L
68 #define LDSEM_WAIT_BIAS		(-LDSEM_ACTIVE_MASK-1)
69 #define LDSEM_READ_BIAS		LDSEM_ACTIVE_BIAS
70 #define LDSEM_WRITE_BIAS	(LDSEM_WAIT_BIAS + LDSEM_ACTIVE_BIAS)
71 
72 struct ldsem_waiter {
73 	struct list_head list;
74 	struct task_struct *task;
75 };
76 
77 /*
78  * Initialize an ldsem:
79  */
__init_ldsem(struct ld_semaphore * sem,const char * name,struct lock_class_key * key)80 void __init_ldsem(struct ld_semaphore *sem, const char *name,
81 		  struct lock_class_key *key)
82 {
83 #ifdef CONFIG_DEBUG_LOCK_ALLOC
84 	/*
85 	 * Make sure we are not reinitializing a held semaphore:
86 	 */
87 	debug_check_no_locks_freed((void *)sem, sizeof(*sem));
88 	lockdep_init_map(&sem->dep_map, name, key, 0);
89 #endif
90 	atomic_long_set(&sem->count, LDSEM_UNLOCKED);
91 	sem->wait_readers = 0;
92 	raw_spin_lock_init(&sem->wait_lock);
93 	INIT_LIST_HEAD(&sem->read_wait);
94 	INIT_LIST_HEAD(&sem->write_wait);
95 }
96 
__ldsem_wake_readers(struct ld_semaphore * sem)97 static void __ldsem_wake_readers(struct ld_semaphore *sem)
98 {
99 	struct ldsem_waiter *waiter, *next;
100 	struct task_struct *tsk;
101 	long adjust, count;
102 
103 	/*
104 	 * Try to grant read locks to all readers on the read wait list.
105 	 * Note the 'active part' of the count is incremented by
106 	 * the number of readers before waking any processes up.
107 	 */
108 	adjust = sem->wait_readers * (LDSEM_ACTIVE_BIAS - LDSEM_WAIT_BIAS);
109 	count = atomic_long_add_return(adjust, &sem->count);
110 	do {
111 		if (count > 0)
112 			break;
113 		if (atomic_long_try_cmpxchg(&sem->count, &count, count - adjust))
114 			return;
115 	} while (1);
116 
117 	list_for_each_entry_safe(waiter, next, &sem->read_wait, list) {
118 		tsk = waiter->task;
119 		smp_mb();
120 		waiter->task = NULL;
121 		wake_up_process(tsk);
122 		put_task_struct(tsk);
123 	}
124 	INIT_LIST_HEAD(&sem->read_wait);
125 	sem->wait_readers = 0;
126 }
127 
writer_trylock(struct ld_semaphore * sem)128 static inline int writer_trylock(struct ld_semaphore *sem)
129 {
130 	/*
131 	 * Only wake this writer if the active part of the count can be
132 	 * transitioned from 0 -> 1
133 	 */
134 	long count = atomic_long_add_return(LDSEM_ACTIVE_BIAS, &sem->count);
135 	do {
136 		if ((count & LDSEM_ACTIVE_MASK) == LDSEM_ACTIVE_BIAS)
137 			return 1;
138 		if (atomic_long_try_cmpxchg(&sem->count, &count, count - LDSEM_ACTIVE_BIAS))
139 			return 0;
140 	} while (1);
141 }
142 
__ldsem_wake_writer(struct ld_semaphore * sem)143 static void __ldsem_wake_writer(struct ld_semaphore *sem)
144 {
145 	struct ldsem_waiter *waiter;
146 
147 	waiter = list_entry(sem->write_wait.next, struct ldsem_waiter, list);
148 	wake_up_process(waiter->task);
149 }
150 
151 /*
152  * handle the lock release when processes blocked on it that can now run
153  * - if we come here from up_xxxx(), then:
154  *   - the 'active part' of count (&0x0000ffff) reached 0 (but may have changed)
155  *   - the 'waiting part' of count (&0xffff0000) is -ve (and will still be so)
156  * - the spinlock must be held by the caller
157  * - woken process blocks are discarded from the list after having task zeroed
158  */
__ldsem_wake(struct ld_semaphore * sem)159 static void __ldsem_wake(struct ld_semaphore *sem)
160 {
161 	if (!list_empty(&sem->write_wait))
162 		__ldsem_wake_writer(sem);
163 	else if (!list_empty(&sem->read_wait))
164 		__ldsem_wake_readers(sem);
165 }
166 
ldsem_wake(struct ld_semaphore * sem)167 static void ldsem_wake(struct ld_semaphore *sem)
168 {
169 	unsigned long flags;
170 
171 	raw_spin_lock_irqsave(&sem->wait_lock, flags);
172 	__ldsem_wake(sem);
173 	raw_spin_unlock_irqrestore(&sem->wait_lock, flags);
174 }
175 
176 /*
177  * wait for the read lock to be granted
178  */
179 static struct ld_semaphore __sched *
down_read_failed(struct ld_semaphore * sem,long count,long timeout)180 down_read_failed(struct ld_semaphore *sem, long count, long timeout)
181 {
182 	struct ldsem_waiter waiter;
183 	long adjust = -LDSEM_ACTIVE_BIAS + LDSEM_WAIT_BIAS;
184 
185 	/* set up my own style of waitqueue */
186 	raw_spin_lock_irq(&sem->wait_lock);
187 
188 	/*
189 	 * Try to reverse the lock attempt but if the count has changed
190 	 * so that reversing fails, check if there are are no waiters,
191 	 * and early-out if not
192 	 */
193 	do {
194 		if (atomic_long_try_cmpxchg(&sem->count, &count, count + adjust)) {
195 			count += adjust;
196 			break;
197 		}
198 		if (count > 0) {
199 			raw_spin_unlock_irq(&sem->wait_lock);
200 			return sem;
201 		}
202 	} while (1);
203 
204 	list_add_tail(&waiter.list, &sem->read_wait);
205 	sem->wait_readers++;
206 
207 	waiter.task = current;
208 	get_task_struct(current);
209 
210 	/* if there are no active locks, wake the new lock owner(s) */
211 	if ((count & LDSEM_ACTIVE_MASK) == 0)
212 		__ldsem_wake(sem);
213 
214 	raw_spin_unlock_irq(&sem->wait_lock);
215 
216 	/* wait to be given the lock */
217 	for (;;) {
218 		set_current_state(TASK_UNINTERRUPTIBLE);
219 
220 		if (!waiter.task)
221 			break;
222 		if (!timeout)
223 			break;
224 		timeout = schedule_timeout(timeout);
225 	}
226 
227 	__set_current_state(TASK_RUNNING);
228 
229 	if (!timeout) {
230 		/*
231 		 * Lock timed out but check if this task was just
232 		 * granted lock ownership - if so, pretend there
233 		 * was no timeout; otherwise, cleanup lock wait.
234 		 */
235 		raw_spin_lock_irq(&sem->wait_lock);
236 		if (waiter.task) {
237 			atomic_long_add_return(-LDSEM_WAIT_BIAS, &sem->count);
238 			list_del(&waiter.list);
239 			raw_spin_unlock_irq(&sem->wait_lock);
240 			put_task_struct(waiter.task);
241 			return NULL;
242 		}
243 		raw_spin_unlock_irq(&sem->wait_lock);
244 	}
245 
246 	return sem;
247 }
248 
249 /*
250  * wait for the write lock to be granted
251  */
252 static struct ld_semaphore __sched *
down_write_failed(struct ld_semaphore * sem,long count,long timeout)253 down_write_failed(struct ld_semaphore *sem, long count, long timeout)
254 {
255 	struct ldsem_waiter waiter;
256 	long adjust = -LDSEM_ACTIVE_BIAS;
257 	int locked = 0;
258 
259 	/* set up my own style of waitqueue */
260 	raw_spin_lock_irq(&sem->wait_lock);
261 
262 	/*
263 	 * Try to reverse the lock attempt but if the count has changed
264 	 * so that reversing fails, check if the lock is now owned,
265 	 * and early-out if so.
266 	 */
267 	do {
268 		if (atomic_long_try_cmpxchg(&sem->count, &count, count + adjust))
269 			break;
270 		if ((count & LDSEM_ACTIVE_MASK) == LDSEM_ACTIVE_BIAS) {
271 			raw_spin_unlock_irq(&sem->wait_lock);
272 			return sem;
273 		}
274 	} while (1);
275 
276 	list_add_tail(&waiter.list, &sem->write_wait);
277 
278 	waiter.task = current;
279 
280 	set_current_state(TASK_UNINTERRUPTIBLE);
281 	for (;;) {
282 		if (!timeout)
283 			break;
284 		raw_spin_unlock_irq(&sem->wait_lock);
285 		timeout = schedule_timeout(timeout);
286 		raw_spin_lock_irq(&sem->wait_lock);
287 		set_current_state(TASK_UNINTERRUPTIBLE);
288 		locked = writer_trylock(sem);
289 		if (locked)
290 			break;
291 	}
292 
293 	if (!locked)
294 		atomic_long_add_return(-LDSEM_WAIT_BIAS, &sem->count);
295 	list_del(&waiter.list);
296 	raw_spin_unlock_irq(&sem->wait_lock);
297 
298 	__set_current_state(TASK_RUNNING);
299 
300 	/* lock wait may have timed out */
301 	if (!locked)
302 		return NULL;
303 	return sem;
304 }
305 
306 
307 
__ldsem_down_read_nested(struct ld_semaphore * sem,int subclass,long timeout)308 static int __ldsem_down_read_nested(struct ld_semaphore *sem,
309 					   int subclass, long timeout)
310 {
311 	long count;
312 
313 	lockdep_acquire_read(sem, subclass, 0, _RET_IP_);
314 
315 	count = atomic_long_add_return(LDSEM_READ_BIAS, &sem->count);
316 	if (count <= 0) {
317 		lock_stat(sem, contended);
318 		if (!down_read_failed(sem, count, timeout)) {
319 			lockdep_release(sem, 1, _RET_IP_);
320 			return 0;
321 		}
322 	}
323 	lock_stat(sem, acquired);
324 	return 1;
325 }
326 
__ldsem_down_write_nested(struct ld_semaphore * sem,int subclass,long timeout)327 static int __ldsem_down_write_nested(struct ld_semaphore *sem,
328 					    int subclass, long timeout)
329 {
330 	long count;
331 
332 	lockdep_acquire(sem, subclass, 0, _RET_IP_);
333 
334 	count = atomic_long_add_return(LDSEM_WRITE_BIAS, &sem->count);
335 	if ((count & LDSEM_ACTIVE_MASK) != LDSEM_ACTIVE_BIAS) {
336 		lock_stat(sem, contended);
337 		if (!down_write_failed(sem, count, timeout)) {
338 			lockdep_release(sem, 1, _RET_IP_);
339 			return 0;
340 		}
341 	}
342 	lock_stat(sem, acquired);
343 	return 1;
344 }
345 
346 
347 /*
348  * lock for reading -- returns 1 if successful, 0 if timed out
349  */
ldsem_down_read(struct ld_semaphore * sem,long timeout)350 int __sched ldsem_down_read(struct ld_semaphore *sem, long timeout)
351 {
352 	might_sleep();
353 	return __ldsem_down_read_nested(sem, 0, timeout);
354 }
355 
356 /*
357  * trylock for reading -- returns 1 if successful, 0 if contention
358  */
ldsem_down_read_trylock(struct ld_semaphore * sem)359 int ldsem_down_read_trylock(struct ld_semaphore *sem)
360 {
361 	long count = atomic_long_read(&sem->count);
362 
363 	while (count >= 0) {
364 		if (atomic_long_try_cmpxchg(&sem->count, &count, count + LDSEM_READ_BIAS)) {
365 			lockdep_acquire_read(sem, 0, 1, _RET_IP_);
366 			lock_stat(sem, acquired);
367 			return 1;
368 		}
369 	}
370 	return 0;
371 }
372 
373 /*
374  * lock for writing -- returns 1 if successful, 0 if timed out
375  */
ldsem_down_write(struct ld_semaphore * sem,long timeout)376 int __sched ldsem_down_write(struct ld_semaphore *sem, long timeout)
377 {
378 	might_sleep();
379 	return __ldsem_down_write_nested(sem, 0, timeout);
380 }
381 
382 /*
383  * trylock for writing -- returns 1 if successful, 0 if contention
384  */
ldsem_down_write_trylock(struct ld_semaphore * sem)385 int ldsem_down_write_trylock(struct ld_semaphore *sem)
386 {
387 	long count = atomic_long_read(&sem->count);
388 
389 	while ((count & LDSEM_ACTIVE_MASK) == 0) {
390 		if (atomic_long_try_cmpxchg(&sem->count, &count, count + LDSEM_WRITE_BIAS)) {
391 			lockdep_acquire(sem, 0, 1, _RET_IP_);
392 			lock_stat(sem, acquired);
393 			return 1;
394 		}
395 	}
396 	return 0;
397 }
398 
399 /*
400  * release a read lock
401  */
ldsem_up_read(struct ld_semaphore * sem)402 void ldsem_up_read(struct ld_semaphore *sem)
403 {
404 	long count;
405 
406 	lockdep_release(sem, 1, _RET_IP_);
407 
408 	count = atomic_long_add_return(-LDSEM_READ_BIAS, &sem->count);
409 	if (count < 0 && (count & LDSEM_ACTIVE_MASK) == 0)
410 		ldsem_wake(sem);
411 }
412 
413 /*
414  * release a write lock
415  */
ldsem_up_write(struct ld_semaphore * sem)416 void ldsem_up_write(struct ld_semaphore *sem)
417 {
418 	long count;
419 
420 	lockdep_release(sem, 1, _RET_IP_);
421 
422 	count = atomic_long_add_return(-LDSEM_WRITE_BIAS, &sem->count);
423 	if (count < 0)
424 		ldsem_wake(sem);
425 }
426 
427 
428 #ifdef CONFIG_DEBUG_LOCK_ALLOC
429 
ldsem_down_read_nested(struct ld_semaphore * sem,int subclass,long timeout)430 int ldsem_down_read_nested(struct ld_semaphore *sem, int subclass, long timeout)
431 {
432 	might_sleep();
433 	return __ldsem_down_read_nested(sem, subclass, timeout);
434 }
435 
ldsem_down_write_nested(struct ld_semaphore * sem,int subclass,long timeout)436 int ldsem_down_write_nested(struct ld_semaphore *sem, int subclass,
437 			    long timeout)
438 {
439 	might_sleep();
440 	return __ldsem_down_write_nested(sem, subclass, timeout);
441 }
442 
443 #endif
444