xref: /openbmc/linux/drivers/md/dm-bio-prison-v1.c (revision 612a462a)
1 /*
2  * Copyright (C) 2012 Red Hat, Inc.
3  *
4  * This file is released under the GPL.
5  */
6 
7 #include "dm.h"
8 #include "dm-bio-prison-v1.h"
9 #include "dm-bio-prison-v2.h"
10 
11 #include <linux/spinlock.h>
12 #include <linux/mempool.h>
13 #include <linux/module.h>
14 #include <linux/slab.h>
15 
16 /*----------------------------------------------------------------*/
17 
18 #define MIN_CELLS 1024
19 
20 struct dm_bio_prison {
21 	spinlock_t lock;
22 	mempool_t *cell_pool;
23 	struct rb_root cells;
24 };
25 
26 static struct kmem_cache *_cell_cache;
27 
28 /*----------------------------------------------------------------*/
29 
30 /*
31  * @nr_cells should be the number of cells you want in use _concurrently_.
32  * Don't confuse it with the number of distinct keys.
33  */
34 struct dm_bio_prison *dm_bio_prison_create(void)
35 {
36 	struct dm_bio_prison *prison = kmalloc(sizeof(*prison), GFP_KERNEL);
37 
38 	if (!prison)
39 		return NULL;
40 
41 	spin_lock_init(&prison->lock);
42 
43 	prison->cell_pool = mempool_create_slab_pool(MIN_CELLS, _cell_cache);
44 	if (!prison->cell_pool) {
45 		kfree(prison);
46 		return NULL;
47 	}
48 
49 	prison->cells = RB_ROOT;
50 
51 	return prison;
52 }
53 EXPORT_SYMBOL_GPL(dm_bio_prison_create);
54 
55 void dm_bio_prison_destroy(struct dm_bio_prison *prison)
56 {
57 	mempool_destroy(prison->cell_pool);
58 	kfree(prison);
59 }
60 EXPORT_SYMBOL_GPL(dm_bio_prison_destroy);
61 
62 struct dm_bio_prison_cell *dm_bio_prison_alloc_cell(struct dm_bio_prison *prison, gfp_t gfp)
63 {
64 	return mempool_alloc(prison->cell_pool, gfp);
65 }
66 EXPORT_SYMBOL_GPL(dm_bio_prison_alloc_cell);
67 
68 void dm_bio_prison_free_cell(struct dm_bio_prison *prison,
69 			     struct dm_bio_prison_cell *cell)
70 {
71 	mempool_free(cell, prison->cell_pool);
72 }
73 EXPORT_SYMBOL_GPL(dm_bio_prison_free_cell);
74 
75 static void __setup_new_cell(struct dm_cell_key *key,
76 			     struct bio *holder,
77 			     struct dm_bio_prison_cell *cell)
78 {
79        memcpy(&cell->key, key, sizeof(cell->key));
80        cell->holder = holder;
81        bio_list_init(&cell->bios);
82 }
83 
84 static int cmp_keys(struct dm_cell_key *lhs,
85 		    struct dm_cell_key *rhs)
86 {
87 	if (lhs->virtual < rhs->virtual)
88 		return -1;
89 
90 	if (lhs->virtual > rhs->virtual)
91 		return 1;
92 
93 	if (lhs->dev < rhs->dev)
94 		return -1;
95 
96 	if (lhs->dev > rhs->dev)
97 		return 1;
98 
99 	if (lhs->block_end <= rhs->block_begin)
100 		return -1;
101 
102 	if (lhs->block_begin >= rhs->block_end)
103 		return 1;
104 
105 	return 0;
106 }
107 
108 static int __bio_detain(struct dm_bio_prison *prison,
109 			struct dm_cell_key *key,
110 			struct bio *inmate,
111 			struct dm_bio_prison_cell *cell_prealloc,
112 			struct dm_bio_prison_cell **cell_result)
113 {
114 	int r;
115 	struct rb_node **new = &prison->cells.rb_node, *parent = NULL;
116 
117 	while (*new) {
118 		struct dm_bio_prison_cell *cell =
119 			rb_entry(*new, struct dm_bio_prison_cell, node);
120 
121 		r = cmp_keys(key, &cell->key);
122 
123 		parent = *new;
124 		if (r < 0)
125 			new = &((*new)->rb_left);
126 		else if (r > 0)
127 			new = &((*new)->rb_right);
128 		else {
129 			if (inmate)
130 				bio_list_add(&cell->bios, inmate);
131 			*cell_result = cell;
132 			return 1;
133 		}
134 	}
135 
136 	__setup_new_cell(key, inmate, cell_prealloc);
137 	*cell_result = cell_prealloc;
138 
139 	rb_link_node(&cell_prealloc->node, parent, new);
140 	rb_insert_color(&cell_prealloc->node, &prison->cells);
141 
142 	return 0;
143 }
144 
145 static int bio_detain(struct dm_bio_prison *prison,
146 		      struct dm_cell_key *key,
147 		      struct bio *inmate,
148 		      struct dm_bio_prison_cell *cell_prealloc,
149 		      struct dm_bio_prison_cell **cell_result)
150 {
151 	int r;
152 	unsigned long flags;
153 
154 	spin_lock_irqsave(&prison->lock, flags);
155 	r = __bio_detain(prison, key, inmate, cell_prealloc, cell_result);
156 	spin_unlock_irqrestore(&prison->lock, flags);
157 
158 	return r;
159 }
160 
161 int dm_bio_detain(struct dm_bio_prison *prison,
162 		  struct dm_cell_key *key,
163 		  struct bio *inmate,
164 		  struct dm_bio_prison_cell *cell_prealloc,
165 		  struct dm_bio_prison_cell **cell_result)
166 {
167 	return bio_detain(prison, key, inmate, cell_prealloc, cell_result);
168 }
169 EXPORT_SYMBOL_GPL(dm_bio_detain);
170 
171 int dm_get_cell(struct dm_bio_prison *prison,
172 		struct dm_cell_key *key,
173 		struct dm_bio_prison_cell *cell_prealloc,
174 		struct dm_bio_prison_cell **cell_result)
175 {
176 	return bio_detain(prison, key, NULL, cell_prealloc, cell_result);
177 }
178 EXPORT_SYMBOL_GPL(dm_get_cell);
179 
180 /*
181  * @inmates must have been initialised prior to this call
182  */
183 static void __cell_release(struct dm_bio_prison *prison,
184 			   struct dm_bio_prison_cell *cell,
185 			   struct bio_list *inmates)
186 {
187 	rb_erase(&cell->node, &prison->cells);
188 
189 	if (inmates) {
190 		if (cell->holder)
191 			bio_list_add(inmates, cell->holder);
192 		bio_list_merge(inmates, &cell->bios);
193 	}
194 }
195 
196 void dm_cell_release(struct dm_bio_prison *prison,
197 		     struct dm_bio_prison_cell *cell,
198 		     struct bio_list *bios)
199 {
200 	unsigned long flags;
201 
202 	spin_lock_irqsave(&prison->lock, flags);
203 	__cell_release(prison, cell, bios);
204 	spin_unlock_irqrestore(&prison->lock, flags);
205 }
206 EXPORT_SYMBOL_GPL(dm_cell_release);
207 
208 /*
209  * Sometimes we don't want the holder, just the additional bios.
210  */
211 static void __cell_release_no_holder(struct dm_bio_prison *prison,
212 				     struct dm_bio_prison_cell *cell,
213 				     struct bio_list *inmates)
214 {
215 	rb_erase(&cell->node, &prison->cells);
216 	bio_list_merge(inmates, &cell->bios);
217 }
218 
219 void dm_cell_release_no_holder(struct dm_bio_prison *prison,
220 			       struct dm_bio_prison_cell *cell,
221 			       struct bio_list *inmates)
222 {
223 	unsigned long flags;
224 
225 	spin_lock_irqsave(&prison->lock, flags);
226 	__cell_release_no_holder(prison, cell, inmates);
227 	spin_unlock_irqrestore(&prison->lock, flags);
228 }
229 EXPORT_SYMBOL_GPL(dm_cell_release_no_holder);
230 
231 void dm_cell_error(struct dm_bio_prison *prison,
232 		   struct dm_bio_prison_cell *cell, blk_status_t error)
233 {
234 	struct bio_list bios;
235 	struct bio *bio;
236 
237 	bio_list_init(&bios);
238 	dm_cell_release(prison, cell, &bios);
239 
240 	while ((bio = bio_list_pop(&bios))) {
241 		bio->bi_status = error;
242 		bio_endio(bio);
243 	}
244 }
245 EXPORT_SYMBOL_GPL(dm_cell_error);
246 
247 void dm_cell_visit_release(struct dm_bio_prison *prison,
248 			   void (*visit_fn)(void *, struct dm_bio_prison_cell *),
249 			   void *context,
250 			   struct dm_bio_prison_cell *cell)
251 {
252 	unsigned long flags;
253 
254 	spin_lock_irqsave(&prison->lock, flags);
255 	visit_fn(context, cell);
256 	rb_erase(&cell->node, &prison->cells);
257 	spin_unlock_irqrestore(&prison->lock, flags);
258 }
259 EXPORT_SYMBOL_GPL(dm_cell_visit_release);
260 
261 static int __promote_or_release(struct dm_bio_prison *prison,
262 				struct dm_bio_prison_cell *cell)
263 {
264 	if (bio_list_empty(&cell->bios)) {
265 		rb_erase(&cell->node, &prison->cells);
266 		return 1;
267 	}
268 
269 	cell->holder = bio_list_pop(&cell->bios);
270 	return 0;
271 }
272 
273 int dm_cell_promote_or_release(struct dm_bio_prison *prison,
274 			       struct dm_bio_prison_cell *cell)
275 {
276 	int r;
277 	unsigned long flags;
278 
279 	spin_lock_irqsave(&prison->lock, flags);
280 	r = __promote_or_release(prison, cell);
281 	spin_unlock_irqrestore(&prison->lock, flags);
282 
283 	return r;
284 }
285 EXPORT_SYMBOL_GPL(dm_cell_promote_or_release);
286 
287 /*----------------------------------------------------------------*/
288 
289 #define DEFERRED_SET_SIZE 64
290 
291 struct dm_deferred_entry {
292 	struct dm_deferred_set *ds;
293 	unsigned count;
294 	struct list_head work_items;
295 };
296 
297 struct dm_deferred_set {
298 	spinlock_t lock;
299 	unsigned current_entry;
300 	unsigned sweeper;
301 	struct dm_deferred_entry entries[DEFERRED_SET_SIZE];
302 };
303 
304 struct dm_deferred_set *dm_deferred_set_create(void)
305 {
306 	int i;
307 	struct dm_deferred_set *ds;
308 
309 	ds = kmalloc(sizeof(*ds), GFP_KERNEL);
310 	if (!ds)
311 		return NULL;
312 
313 	spin_lock_init(&ds->lock);
314 	ds->current_entry = 0;
315 	ds->sweeper = 0;
316 	for (i = 0; i < DEFERRED_SET_SIZE; i++) {
317 		ds->entries[i].ds = ds;
318 		ds->entries[i].count = 0;
319 		INIT_LIST_HEAD(&ds->entries[i].work_items);
320 	}
321 
322 	return ds;
323 }
324 EXPORT_SYMBOL_GPL(dm_deferred_set_create);
325 
326 void dm_deferred_set_destroy(struct dm_deferred_set *ds)
327 {
328 	kfree(ds);
329 }
330 EXPORT_SYMBOL_GPL(dm_deferred_set_destroy);
331 
332 struct dm_deferred_entry *dm_deferred_entry_inc(struct dm_deferred_set *ds)
333 {
334 	unsigned long flags;
335 	struct dm_deferred_entry *entry;
336 
337 	spin_lock_irqsave(&ds->lock, flags);
338 	entry = ds->entries + ds->current_entry;
339 	entry->count++;
340 	spin_unlock_irqrestore(&ds->lock, flags);
341 
342 	return entry;
343 }
344 EXPORT_SYMBOL_GPL(dm_deferred_entry_inc);
345 
346 static unsigned ds_next(unsigned index)
347 {
348 	return (index + 1) % DEFERRED_SET_SIZE;
349 }
350 
351 static void __sweep(struct dm_deferred_set *ds, struct list_head *head)
352 {
353 	while ((ds->sweeper != ds->current_entry) &&
354 	       !ds->entries[ds->sweeper].count) {
355 		list_splice_init(&ds->entries[ds->sweeper].work_items, head);
356 		ds->sweeper = ds_next(ds->sweeper);
357 	}
358 
359 	if ((ds->sweeper == ds->current_entry) && !ds->entries[ds->sweeper].count)
360 		list_splice_init(&ds->entries[ds->sweeper].work_items, head);
361 }
362 
363 void dm_deferred_entry_dec(struct dm_deferred_entry *entry, struct list_head *head)
364 {
365 	unsigned long flags;
366 
367 	spin_lock_irqsave(&entry->ds->lock, flags);
368 	BUG_ON(!entry->count);
369 	--entry->count;
370 	__sweep(entry->ds, head);
371 	spin_unlock_irqrestore(&entry->ds->lock, flags);
372 }
373 EXPORT_SYMBOL_GPL(dm_deferred_entry_dec);
374 
375 /*
376  * Returns 1 if deferred or 0 if no pending items to delay job.
377  */
378 int dm_deferred_set_add_work(struct dm_deferred_set *ds, struct list_head *work)
379 {
380 	int r = 1;
381 	unsigned long flags;
382 	unsigned next_entry;
383 
384 	spin_lock_irqsave(&ds->lock, flags);
385 	if ((ds->sweeper == ds->current_entry) &&
386 	    !ds->entries[ds->current_entry].count)
387 		r = 0;
388 	else {
389 		list_add(work, &ds->entries[ds->current_entry].work_items);
390 		next_entry = ds_next(ds->current_entry);
391 		if (!ds->entries[next_entry].count)
392 			ds->current_entry = next_entry;
393 	}
394 	spin_unlock_irqrestore(&ds->lock, flags);
395 
396 	return r;
397 }
398 EXPORT_SYMBOL_GPL(dm_deferred_set_add_work);
399 
400 /*----------------------------------------------------------------*/
401 
402 static int __init dm_bio_prison_init_v1(void)
403 {
404 	_cell_cache = KMEM_CACHE(dm_bio_prison_cell, 0);
405 	if (!_cell_cache)
406 		return -ENOMEM;
407 
408 	return 0;
409 }
410 
411 static void dm_bio_prison_exit_v1(void)
412 {
413 	kmem_cache_destroy(_cell_cache);
414 	_cell_cache = NULL;
415 }
416 
417 static int (*_inits[])(void) __initdata = {
418 	dm_bio_prison_init_v1,
419 	dm_bio_prison_init_v2,
420 };
421 
422 static void (*_exits[])(void) = {
423 	dm_bio_prison_exit_v1,
424 	dm_bio_prison_exit_v2,
425 };
426 
427 static int __init dm_bio_prison_init(void)
428 {
429 	const int count = ARRAY_SIZE(_inits);
430 
431 	int r, i;
432 
433 	for (i = 0; i < count; i++) {
434 		r = _inits[i]();
435 		if (r)
436 			goto bad;
437 	}
438 
439 	return 0;
440 
441       bad:
442 	while (i--)
443 		_exits[i]();
444 
445 	return r;
446 }
447 
448 static void __exit dm_bio_prison_exit(void)
449 {
450 	int i = ARRAY_SIZE(_exits);
451 
452 	while (i--)
453 		_exits[i]();
454 }
455 
456 /*
457  * module hooks
458  */
459 module_init(dm_bio_prison_init);
460 module_exit(dm_bio_prison_exit);
461 
462 MODULE_DESCRIPTION(DM_NAME " bio prison");
463 MODULE_AUTHOR("Joe Thornber <dm-devel@redhat.com>");
464 MODULE_LICENSE("GPL");
465