xref: /openbmc/linux/drivers/md/dm-bio-prison-v1.c (revision 1a59d1b8)
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 	struct rb_root cells;
23 	mempool_t cell_pool;
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 = kzalloc(sizeof(*prison), GFP_KERNEL);
37 	int ret;
38 
39 	if (!prison)
40 		return NULL;
41 
42 	spin_lock_init(&prison->lock);
43 
44 	ret = mempool_init_slab_pool(&prison->cell_pool, MIN_CELLS, _cell_cache);
45 	if (ret) {
46 		kfree(prison);
47 		return NULL;
48 	}
49 
50 	prison->cells = RB_ROOT;
51 
52 	return prison;
53 }
54 EXPORT_SYMBOL_GPL(dm_bio_prison_create);
55 
56 void dm_bio_prison_destroy(struct dm_bio_prison *prison)
57 {
58 	mempool_exit(&prison->cell_pool);
59 	kfree(prison);
60 }
61 EXPORT_SYMBOL_GPL(dm_bio_prison_destroy);
62 
63 struct dm_bio_prison_cell *dm_bio_prison_alloc_cell(struct dm_bio_prison *prison, gfp_t gfp)
64 {
65 	return mempool_alloc(&prison->cell_pool, gfp);
66 }
67 EXPORT_SYMBOL_GPL(dm_bio_prison_alloc_cell);
68 
69 void dm_bio_prison_free_cell(struct dm_bio_prison *prison,
70 			     struct dm_bio_prison_cell *cell)
71 {
72 	mempool_free(cell, &prison->cell_pool);
73 }
74 EXPORT_SYMBOL_GPL(dm_bio_prison_free_cell);
75 
76 static void __setup_new_cell(struct dm_cell_key *key,
77 			     struct bio *holder,
78 			     struct dm_bio_prison_cell *cell)
79 {
80        memcpy(&cell->key, key, sizeof(cell->key));
81        cell->holder = holder;
82        bio_list_init(&cell->bios);
83 }
84 
85 static int cmp_keys(struct dm_cell_key *lhs,
86 		    struct dm_cell_key *rhs)
87 {
88 	if (lhs->virtual < rhs->virtual)
89 		return -1;
90 
91 	if (lhs->virtual > rhs->virtual)
92 		return 1;
93 
94 	if (lhs->dev < rhs->dev)
95 		return -1;
96 
97 	if (lhs->dev > rhs->dev)
98 		return 1;
99 
100 	if (lhs->block_end <= rhs->block_begin)
101 		return -1;
102 
103 	if (lhs->block_begin >= rhs->block_end)
104 		return 1;
105 
106 	return 0;
107 }
108 
109 static int __bio_detain(struct dm_bio_prison *prison,
110 			struct dm_cell_key *key,
111 			struct bio *inmate,
112 			struct dm_bio_prison_cell *cell_prealloc,
113 			struct dm_bio_prison_cell **cell_result)
114 {
115 	int r;
116 	struct rb_node **new = &prison->cells.rb_node, *parent = NULL;
117 
118 	while (*new) {
119 		struct dm_bio_prison_cell *cell =
120 			rb_entry(*new, struct dm_bio_prison_cell, node);
121 
122 		r = cmp_keys(key, &cell->key);
123 
124 		parent = *new;
125 		if (r < 0)
126 			new = &((*new)->rb_left);
127 		else if (r > 0)
128 			new = &((*new)->rb_right);
129 		else {
130 			if (inmate)
131 				bio_list_add(&cell->bios, inmate);
132 			*cell_result = cell;
133 			return 1;
134 		}
135 	}
136 
137 	__setup_new_cell(key, inmate, cell_prealloc);
138 	*cell_result = cell_prealloc;
139 
140 	rb_link_node(&cell_prealloc->node, parent, new);
141 	rb_insert_color(&cell_prealloc->node, &prison->cells);
142 
143 	return 0;
144 }
145 
146 static int bio_detain(struct dm_bio_prison *prison,
147 		      struct dm_cell_key *key,
148 		      struct bio *inmate,
149 		      struct dm_bio_prison_cell *cell_prealloc,
150 		      struct dm_bio_prison_cell **cell_result)
151 {
152 	int r;
153 	unsigned long flags;
154 
155 	spin_lock_irqsave(&prison->lock, flags);
156 	r = __bio_detain(prison, key, inmate, cell_prealloc, cell_result);
157 	spin_unlock_irqrestore(&prison->lock, flags);
158 
159 	return r;
160 }
161 
162 int dm_bio_detain(struct dm_bio_prison *prison,
163 		  struct dm_cell_key *key,
164 		  struct bio *inmate,
165 		  struct dm_bio_prison_cell *cell_prealloc,
166 		  struct dm_bio_prison_cell **cell_result)
167 {
168 	return bio_detain(prison, key, inmate, cell_prealloc, cell_result);
169 }
170 EXPORT_SYMBOL_GPL(dm_bio_detain);
171 
172 int dm_get_cell(struct dm_bio_prison *prison,
173 		struct dm_cell_key *key,
174 		struct dm_bio_prison_cell *cell_prealloc,
175 		struct dm_bio_prison_cell **cell_result)
176 {
177 	return bio_detain(prison, key, NULL, cell_prealloc, cell_result);
178 }
179 EXPORT_SYMBOL_GPL(dm_get_cell);
180 
181 /*
182  * @inmates must have been initialised prior to this call
183  */
184 static void __cell_release(struct dm_bio_prison *prison,
185 			   struct dm_bio_prison_cell *cell,
186 			   struct bio_list *inmates)
187 {
188 	rb_erase(&cell->node, &prison->cells);
189 
190 	if (inmates) {
191 		if (cell->holder)
192 			bio_list_add(inmates, cell->holder);
193 		bio_list_merge(inmates, &cell->bios);
194 	}
195 }
196 
197 void dm_cell_release(struct dm_bio_prison *prison,
198 		     struct dm_bio_prison_cell *cell,
199 		     struct bio_list *bios)
200 {
201 	unsigned long flags;
202 
203 	spin_lock_irqsave(&prison->lock, flags);
204 	__cell_release(prison, cell, bios);
205 	spin_unlock_irqrestore(&prison->lock, flags);
206 }
207 EXPORT_SYMBOL_GPL(dm_cell_release);
208 
209 /*
210  * Sometimes we don't want the holder, just the additional bios.
211  */
212 static void __cell_release_no_holder(struct dm_bio_prison *prison,
213 				     struct dm_bio_prison_cell *cell,
214 				     struct bio_list *inmates)
215 {
216 	rb_erase(&cell->node, &prison->cells);
217 	bio_list_merge(inmates, &cell->bios);
218 }
219 
220 void dm_cell_release_no_holder(struct dm_bio_prison *prison,
221 			       struct dm_bio_prison_cell *cell,
222 			       struct bio_list *inmates)
223 {
224 	unsigned long flags;
225 
226 	spin_lock_irqsave(&prison->lock, flags);
227 	__cell_release_no_holder(prison, cell, inmates);
228 	spin_unlock_irqrestore(&prison->lock, flags);
229 }
230 EXPORT_SYMBOL_GPL(dm_cell_release_no_holder);
231 
232 void dm_cell_error(struct dm_bio_prison *prison,
233 		   struct dm_bio_prison_cell *cell, blk_status_t error)
234 {
235 	struct bio_list bios;
236 	struct bio *bio;
237 
238 	bio_list_init(&bios);
239 	dm_cell_release(prison, cell, &bios);
240 
241 	while ((bio = bio_list_pop(&bios))) {
242 		bio->bi_status = error;
243 		bio_endio(bio);
244 	}
245 }
246 EXPORT_SYMBOL_GPL(dm_cell_error);
247 
248 void dm_cell_visit_release(struct dm_bio_prison *prison,
249 			   void (*visit_fn)(void *, struct dm_bio_prison_cell *),
250 			   void *context,
251 			   struct dm_bio_prison_cell *cell)
252 {
253 	unsigned long flags;
254 
255 	spin_lock_irqsave(&prison->lock, flags);
256 	visit_fn(context, cell);
257 	rb_erase(&cell->node, &prison->cells);
258 	spin_unlock_irqrestore(&prison->lock, flags);
259 }
260 EXPORT_SYMBOL_GPL(dm_cell_visit_release);
261 
262 static int __promote_or_release(struct dm_bio_prison *prison,
263 				struct dm_bio_prison_cell *cell)
264 {
265 	if (bio_list_empty(&cell->bios)) {
266 		rb_erase(&cell->node, &prison->cells);
267 		return 1;
268 	}
269 
270 	cell->holder = bio_list_pop(&cell->bios);
271 	return 0;
272 }
273 
274 int dm_cell_promote_or_release(struct dm_bio_prison *prison,
275 			       struct dm_bio_prison_cell *cell)
276 {
277 	int r;
278 	unsigned long flags;
279 
280 	spin_lock_irqsave(&prison->lock, flags);
281 	r = __promote_or_release(prison, cell);
282 	spin_unlock_irqrestore(&prison->lock, flags);
283 
284 	return r;
285 }
286 EXPORT_SYMBOL_GPL(dm_cell_promote_or_release);
287 
288 /*----------------------------------------------------------------*/
289 
290 #define DEFERRED_SET_SIZE 64
291 
292 struct dm_deferred_entry {
293 	struct dm_deferred_set *ds;
294 	unsigned count;
295 	struct list_head work_items;
296 };
297 
298 struct dm_deferred_set {
299 	spinlock_t lock;
300 	unsigned current_entry;
301 	unsigned sweeper;
302 	struct dm_deferred_entry entries[DEFERRED_SET_SIZE];
303 };
304 
305 struct dm_deferred_set *dm_deferred_set_create(void)
306 {
307 	int i;
308 	struct dm_deferred_set *ds;
309 
310 	ds = kmalloc(sizeof(*ds), GFP_KERNEL);
311 	if (!ds)
312 		return NULL;
313 
314 	spin_lock_init(&ds->lock);
315 	ds->current_entry = 0;
316 	ds->sweeper = 0;
317 	for (i = 0; i < DEFERRED_SET_SIZE; i++) {
318 		ds->entries[i].ds = ds;
319 		ds->entries[i].count = 0;
320 		INIT_LIST_HEAD(&ds->entries[i].work_items);
321 	}
322 
323 	return ds;
324 }
325 EXPORT_SYMBOL_GPL(dm_deferred_set_create);
326 
327 void dm_deferred_set_destroy(struct dm_deferred_set *ds)
328 {
329 	kfree(ds);
330 }
331 EXPORT_SYMBOL_GPL(dm_deferred_set_destroy);
332 
333 struct dm_deferred_entry *dm_deferred_entry_inc(struct dm_deferred_set *ds)
334 {
335 	unsigned long flags;
336 	struct dm_deferred_entry *entry;
337 
338 	spin_lock_irqsave(&ds->lock, flags);
339 	entry = ds->entries + ds->current_entry;
340 	entry->count++;
341 	spin_unlock_irqrestore(&ds->lock, flags);
342 
343 	return entry;
344 }
345 EXPORT_SYMBOL_GPL(dm_deferred_entry_inc);
346 
347 static unsigned ds_next(unsigned index)
348 {
349 	return (index + 1) % DEFERRED_SET_SIZE;
350 }
351 
352 static void __sweep(struct dm_deferred_set *ds, struct list_head *head)
353 {
354 	while ((ds->sweeper != ds->current_entry) &&
355 	       !ds->entries[ds->sweeper].count) {
356 		list_splice_init(&ds->entries[ds->sweeper].work_items, head);
357 		ds->sweeper = ds_next(ds->sweeper);
358 	}
359 
360 	if ((ds->sweeper == ds->current_entry) && !ds->entries[ds->sweeper].count)
361 		list_splice_init(&ds->entries[ds->sweeper].work_items, head);
362 }
363 
364 void dm_deferred_entry_dec(struct dm_deferred_entry *entry, struct list_head *head)
365 {
366 	unsigned long flags;
367 
368 	spin_lock_irqsave(&entry->ds->lock, flags);
369 	BUG_ON(!entry->count);
370 	--entry->count;
371 	__sweep(entry->ds, head);
372 	spin_unlock_irqrestore(&entry->ds->lock, flags);
373 }
374 EXPORT_SYMBOL_GPL(dm_deferred_entry_dec);
375 
376 /*
377  * Returns 1 if deferred or 0 if no pending items to delay job.
378  */
379 int dm_deferred_set_add_work(struct dm_deferred_set *ds, struct list_head *work)
380 {
381 	int r = 1;
382 	unsigned long flags;
383 	unsigned next_entry;
384 
385 	spin_lock_irqsave(&ds->lock, flags);
386 	if ((ds->sweeper == ds->current_entry) &&
387 	    !ds->entries[ds->current_entry].count)
388 		r = 0;
389 	else {
390 		list_add(work, &ds->entries[ds->current_entry].work_items);
391 		next_entry = ds_next(ds->current_entry);
392 		if (!ds->entries[next_entry].count)
393 			ds->current_entry = next_entry;
394 	}
395 	spin_unlock_irqrestore(&ds->lock, flags);
396 
397 	return r;
398 }
399 EXPORT_SYMBOL_GPL(dm_deferred_set_add_work);
400 
401 /*----------------------------------------------------------------*/
402 
403 static int __init dm_bio_prison_init_v1(void)
404 {
405 	_cell_cache = KMEM_CACHE(dm_bio_prison_cell, 0);
406 	if (!_cell_cache)
407 		return -ENOMEM;
408 
409 	return 0;
410 }
411 
412 static void dm_bio_prison_exit_v1(void)
413 {
414 	kmem_cache_destroy(_cell_cache);
415 	_cell_cache = NULL;
416 }
417 
418 static int (*_inits[])(void) __initdata = {
419 	dm_bio_prison_init_v1,
420 	dm_bio_prison_init_v2,
421 };
422 
423 static void (*_exits[])(void) = {
424 	dm_bio_prison_exit_v1,
425 	dm_bio_prison_exit_v2,
426 };
427 
428 static int __init dm_bio_prison_init(void)
429 {
430 	const int count = ARRAY_SIZE(_inits);
431 
432 	int r, i;
433 
434 	for (i = 0; i < count; i++) {
435 		r = _inits[i]();
436 		if (r)
437 			goto bad;
438 	}
439 
440 	return 0;
441 
442       bad:
443 	while (i--)
444 		_exits[i]();
445 
446 	return r;
447 }
448 
449 static void __exit dm_bio_prison_exit(void)
450 {
451 	int i = ARRAY_SIZE(_exits);
452 
453 	while (i--)
454 		_exits[i]();
455 }
456 
457 /*
458  * module hooks
459  */
460 module_init(dm_bio_prison_init);
461 module_exit(dm_bio_prison_exit);
462 
463 MODULE_DESCRIPTION(DM_NAME " bio prison");
464 MODULE_AUTHOR("Joe Thornber <dm-devel@redhat.com>");
465 MODULE_LICENSE("GPL");
466