1 // SPDX-License-Identifier: GPL-2.0-only
2 /*
3 * Copyright 2023 Red Hat
4 */
5
6 /*
7 * This file contains the main entry points for normal operations on a vdo as well as functions for
8 * constructing and destroying vdo instances (in memory).
9 */
10
11 /**
12 * DOC:
13 *
14 * A read_only_notifier has a single completion which is used to perform read-only notifications,
15 * however, vdo_enter_read_only_mode() may be called from any thread. A pair of fields, protected
16 * by a spinlock, are used to control the read-only mode entry process. The first field holds the
17 * read-only error. The second is the state field, which may hold any of the four special values
18 * enumerated here.
19 *
20 * When vdo_enter_read_only_mode() is called from some vdo thread, if the read_only_error field
21 * already contains an error (i.e. its value is not VDO_SUCCESS), then some other error has already
22 * initiated the read-only process, and nothing more is done. Otherwise, the new error is stored in
23 * the read_only_error field, and the state field is consulted. If the state is MAY_NOTIFY, it is
24 * set to NOTIFYING, and the notification process begins. If the state is MAY_NOT_NOTIFY, then
25 * notifications are currently disallowed, generally due to the vdo being suspended. In this case,
26 * the nothing more will be done until the vdo is resumed, at which point the notification will be
27 * performed. In any other case, the vdo is already read-only, and there is nothing more to do.
28 */
29
30 #include "vdo.h"
31
32 #include <linux/completion.h>
33 #include <linux/device-mapper.h>
34 #include <linux/kernel.h>
35 #include <linux/lz4.h>
36 #include <linux/module.h>
37 #include <linux/mutex.h>
38 #include <linux/spinlock.h>
39 #include <linux/types.h>
40
41 #include "logger.h"
42 #include "memory-alloc.h"
43 #include "permassert.h"
44 #include "string-utils.h"
45
46 #include "block-map.h"
47 #include "completion.h"
48 #include "data-vio.h"
49 #include "dedupe.h"
50 #include "encodings.h"
51 #include "funnel-workqueue.h"
52 #include "io-submitter.h"
53 #include "logical-zone.h"
54 #include "packer.h"
55 #include "physical-zone.h"
56 #include "recovery-journal.h"
57 #include "slab-depot.h"
58 #include "statistics.h"
59 #include "status-codes.h"
60 #include "vio.h"
61
62 #define PARANOID_THREAD_CONSISTENCY_CHECKS 0
63
64 struct sync_completion {
65 struct vdo_completion vdo_completion;
66 struct completion completion;
67 };
68
69 /* A linked list is adequate for the small number of entries we expect. */
70 struct device_registry {
71 struct list_head links;
72 /* TODO: Convert to rcu per kernel recommendation. */
73 rwlock_t lock;
74 };
75
76 static struct device_registry registry;
77
78 /**
79 * vdo_initialize_device_registry_once() - Initialize the necessary structures for the device
80 * registry.
81 */
vdo_initialize_device_registry_once(void)82 void vdo_initialize_device_registry_once(void)
83 {
84 INIT_LIST_HEAD(®istry.links);
85 rwlock_init(®istry.lock);
86 }
87
88 /** vdo_is_equal() - Implements vdo_filter_fn. */
vdo_is_equal(struct vdo * vdo,const void * context)89 static bool vdo_is_equal(struct vdo *vdo, const void *context)
90 {
91 return (vdo == context);
92 }
93
94 /**
95 * filter_vdos_locked() - Find a vdo in the registry if it exists there.
96 * @filter: The filter function to apply to devices.
97 * @context: A bit of context to provide the filter.
98 *
99 * Context: Must be called holding the lock.
100 *
101 * Return: the vdo object found, if any.
102 */
filter_vdos_locked(vdo_filter_fn filter,const void * context)103 static struct vdo * __must_check filter_vdos_locked(vdo_filter_fn filter,
104 const void *context)
105 {
106 struct vdo *vdo;
107
108 list_for_each_entry(vdo, ®istry.links, registration) {
109 if (filter(vdo, context))
110 return vdo;
111 }
112
113 return NULL;
114 }
115
116 /**
117 * vdo_find_matching() - Find and return the first (if any) vdo matching a given filter function.
118 * @filter: The filter function to apply to vdos.
119 * @context: A bit of context to provide the filter.
120 */
vdo_find_matching(vdo_filter_fn filter,const void * context)121 struct vdo *vdo_find_matching(vdo_filter_fn filter, const void *context)
122 {
123 struct vdo *vdo;
124
125 read_lock(®istry.lock);
126 vdo = filter_vdos_locked(filter, context);
127 read_unlock(®istry.lock);
128
129 return vdo;
130 }
131
start_vdo_request_queue(void * ptr)132 static void start_vdo_request_queue(void *ptr)
133 {
134 struct vdo_thread *thread = vdo_get_work_queue_owner(vdo_get_current_work_queue());
135
136 vdo_register_allocating_thread(&thread->allocating_thread,
137 &thread->vdo->allocations_allowed);
138 }
139
finish_vdo_request_queue(void * ptr)140 static void finish_vdo_request_queue(void *ptr)
141 {
142 vdo_unregister_allocating_thread();
143 }
144
145 #ifdef MODULE
146 #define MODULE_NAME THIS_MODULE->name
147 #else
148 #define MODULE_NAME "dm-vdo"
149 #endif /* MODULE */
150
151 static const struct vdo_work_queue_type default_queue_type = {
152 .start = start_vdo_request_queue,
153 .finish = finish_vdo_request_queue,
154 .max_priority = VDO_DEFAULT_Q_MAX_PRIORITY,
155 .default_priority = VDO_DEFAULT_Q_COMPLETION_PRIORITY,
156 };
157
158 static const struct vdo_work_queue_type bio_ack_q_type = {
159 .start = NULL,
160 .finish = NULL,
161 .max_priority = BIO_ACK_Q_MAX_PRIORITY,
162 .default_priority = BIO_ACK_Q_ACK_PRIORITY,
163 };
164
165 static const struct vdo_work_queue_type cpu_q_type = {
166 .start = NULL,
167 .finish = NULL,
168 .max_priority = CPU_Q_MAX_PRIORITY,
169 .default_priority = CPU_Q_MAX_PRIORITY,
170 };
171
uninitialize_thread_config(struct thread_config * config)172 static void uninitialize_thread_config(struct thread_config *config)
173 {
174 vdo_free(vdo_forget(config->logical_threads));
175 vdo_free(vdo_forget(config->physical_threads));
176 vdo_free(vdo_forget(config->hash_zone_threads));
177 vdo_free(vdo_forget(config->bio_threads));
178 memset(config, 0, sizeof(struct thread_config));
179 }
180
assign_thread_ids(struct thread_config * config,thread_id_t thread_ids[],zone_count_t count)181 static void assign_thread_ids(struct thread_config *config,
182 thread_id_t thread_ids[], zone_count_t count)
183 {
184 zone_count_t zone;
185
186 for (zone = 0; zone < count; zone++)
187 thread_ids[zone] = config->thread_count++;
188 }
189
190 /**
191 * initialize_thread_config() - Initialize the thread mapping
192 *
193 * If the logical, physical, and hash zone counts are all 0, a single thread will be shared by all
194 * three plus the packer and recovery journal. Otherwise, there must be at least one of each type,
195 * and each will have its own thread, as will the packer and recovery journal.
196 *
197 * Return: VDO_SUCCESS or an error.
198 */
initialize_thread_config(struct thread_count_config counts,struct thread_config * config)199 static int __must_check initialize_thread_config(struct thread_count_config counts,
200 struct thread_config *config)
201 {
202 int result;
203 bool single = ((counts.logical_zones + counts.physical_zones + counts.hash_zones) == 0);
204
205 config->bio_thread_count = counts.bio_threads;
206 if (single) {
207 config->logical_zone_count = 1;
208 config->physical_zone_count = 1;
209 config->hash_zone_count = 1;
210 } else {
211 config->logical_zone_count = counts.logical_zones;
212 config->physical_zone_count = counts.physical_zones;
213 config->hash_zone_count = counts.hash_zones;
214 }
215
216 result = vdo_allocate(config->logical_zone_count, thread_id_t,
217 "logical thread array", &config->logical_threads);
218 if (result != VDO_SUCCESS) {
219 uninitialize_thread_config(config);
220 return result;
221 }
222
223 result = vdo_allocate(config->physical_zone_count, thread_id_t,
224 "physical thread array", &config->physical_threads);
225 if (result != VDO_SUCCESS) {
226 uninitialize_thread_config(config);
227 return result;
228 }
229
230 result = vdo_allocate(config->hash_zone_count, thread_id_t,
231 "hash thread array", &config->hash_zone_threads);
232 if (result != VDO_SUCCESS) {
233 uninitialize_thread_config(config);
234 return result;
235 }
236
237 result = vdo_allocate(config->bio_thread_count, thread_id_t,
238 "bio thread array", &config->bio_threads);
239 if (result != VDO_SUCCESS) {
240 uninitialize_thread_config(config);
241 return result;
242 }
243
244 if (single) {
245 config->logical_threads[0] = config->thread_count;
246 config->physical_threads[0] = config->thread_count;
247 config->hash_zone_threads[0] = config->thread_count++;
248 } else {
249 config->admin_thread = config->thread_count;
250 config->journal_thread = config->thread_count++;
251 config->packer_thread = config->thread_count++;
252 assign_thread_ids(config, config->logical_threads, counts.logical_zones);
253 assign_thread_ids(config, config->physical_threads, counts.physical_zones);
254 assign_thread_ids(config, config->hash_zone_threads, counts.hash_zones);
255 }
256
257 config->dedupe_thread = config->thread_count++;
258 config->bio_ack_thread =
259 ((counts.bio_ack_threads > 0) ? config->thread_count++ : VDO_INVALID_THREAD_ID);
260 config->cpu_thread = config->thread_count++;
261 assign_thread_ids(config, config->bio_threads, counts.bio_threads);
262 return VDO_SUCCESS;
263 }
264
265 /**
266 * read_geometry_block() - Synchronously read the geometry block from a vdo's underlying block
267 * device.
268 * @vdo: The vdo whose geometry is to be read.
269 *
270 * Return: VDO_SUCCESS or an error code.
271 */
read_geometry_block(struct vdo * vdo)272 static int __must_check read_geometry_block(struct vdo *vdo)
273 {
274 struct vio *vio;
275 char *block;
276 int result;
277
278 result = vdo_allocate(VDO_BLOCK_SIZE, u8, __func__, &block);
279 if (result != VDO_SUCCESS)
280 return result;
281
282 result = create_metadata_vio(vdo, VIO_TYPE_GEOMETRY, VIO_PRIORITY_HIGH, NULL,
283 block, &vio);
284 if (result != VDO_SUCCESS) {
285 vdo_free(block);
286 return result;
287 }
288
289 /*
290 * This is only safe because, having not already loaded the geometry, the vdo's geometry's
291 * bio_offset field is 0, so the fact that vio_reset_bio() will subtract that offset from
292 * the supplied pbn is not a problem.
293 */
294 result = vio_reset_bio(vio, block, NULL, REQ_OP_READ,
295 VDO_GEOMETRY_BLOCK_LOCATION);
296 if (result != VDO_SUCCESS) {
297 free_vio(vdo_forget(vio));
298 vdo_free(block);
299 return result;
300 }
301
302 bio_set_dev(vio->bio, vdo_get_backing_device(vdo));
303 submit_bio_wait(vio->bio);
304 result = blk_status_to_errno(vio->bio->bi_status);
305 free_vio(vdo_forget(vio));
306 if (result != 0) {
307 vdo_log_error_strerror(result, "synchronous read failed");
308 vdo_free(block);
309 return -EIO;
310 }
311
312 result = vdo_parse_geometry_block((u8 *) block, &vdo->geometry);
313 vdo_free(block);
314 return result;
315 }
316
get_zone_thread_name(const thread_id_t thread_ids[],zone_count_t count,thread_id_t id,const char * prefix,char * buffer,size_t buffer_length)317 static bool get_zone_thread_name(const thread_id_t thread_ids[], zone_count_t count,
318 thread_id_t id, const char *prefix,
319 char *buffer, size_t buffer_length)
320 {
321 if (id >= thread_ids[0]) {
322 thread_id_t index = id - thread_ids[0];
323
324 if (index < count) {
325 snprintf(buffer, buffer_length, "%s%d", prefix, index);
326 return true;
327 }
328 }
329
330 return false;
331 }
332
333 /**
334 * get_thread_name() - Format the name of the worker thread desired to support a given work queue.
335 * @thread_config: The thread configuration.
336 * @thread_id: The thread id.
337 * @buffer: Where to put the formatted name.
338 * @buffer_length: Size of the output buffer.
339 *
340 * The physical layer may add a prefix identifying the product; the output from this function
341 * should just identify the thread.
342 */
get_thread_name(const struct thread_config * thread_config,thread_id_t thread_id,char * buffer,size_t buffer_length)343 static void get_thread_name(const struct thread_config *thread_config,
344 thread_id_t thread_id, char *buffer, size_t buffer_length)
345 {
346 if (thread_id == thread_config->journal_thread) {
347 if (thread_config->packer_thread == thread_id) {
348 /*
349 * This is the "single thread" config where one thread is used for the
350 * journal, packer, logical, physical, and hash zones. In that case, it is
351 * known as the "request queue."
352 */
353 snprintf(buffer, buffer_length, "reqQ");
354 return;
355 }
356
357 snprintf(buffer, buffer_length, "journalQ");
358 return;
359 } else if (thread_id == thread_config->admin_thread) {
360 /* Theoretically this could be different from the journal thread. */
361 snprintf(buffer, buffer_length, "adminQ");
362 return;
363 } else if (thread_id == thread_config->packer_thread) {
364 snprintf(buffer, buffer_length, "packerQ");
365 return;
366 } else if (thread_id == thread_config->dedupe_thread) {
367 snprintf(buffer, buffer_length, "dedupeQ");
368 return;
369 } else if (thread_id == thread_config->bio_ack_thread) {
370 snprintf(buffer, buffer_length, "ackQ");
371 return;
372 } else if (thread_id == thread_config->cpu_thread) {
373 snprintf(buffer, buffer_length, "cpuQ");
374 return;
375 }
376
377 if (get_zone_thread_name(thread_config->logical_threads,
378 thread_config->logical_zone_count,
379 thread_id, "logQ", buffer, buffer_length))
380 return;
381
382 if (get_zone_thread_name(thread_config->physical_threads,
383 thread_config->physical_zone_count,
384 thread_id, "physQ", buffer, buffer_length))
385 return;
386
387 if (get_zone_thread_name(thread_config->hash_zone_threads,
388 thread_config->hash_zone_count,
389 thread_id, "hashQ", buffer, buffer_length))
390 return;
391
392 if (get_zone_thread_name(thread_config->bio_threads,
393 thread_config->bio_thread_count,
394 thread_id, "bioQ", buffer, buffer_length))
395 return;
396
397 /* Some sort of misconfiguration? */
398 snprintf(buffer, buffer_length, "reqQ%d", thread_id);
399 }
400
401 /**
402 * vdo_make_thread() - Construct a single vdo work_queue and its associated thread (or threads for
403 * round-robin queues).
404 * @vdo: The vdo which owns the thread.
405 * @thread_id: The id of the thread to create (as determined by the thread_config).
406 * @type: The description of the work queue for this thread.
407 * @queue_count: The number of actual threads/queues contained in the "thread".
408 * @contexts: An array of queue_count contexts, one for each individual queue; may be NULL.
409 *
410 * Each "thread" constructed by this method is represented by a unique thread id in the thread
411 * config, and completions can be enqueued to the queue and run on the threads comprising this
412 * entity.
413 *
414 * Return: VDO_SUCCESS or an error.
415 */
vdo_make_thread(struct vdo * vdo,thread_id_t thread_id,const struct vdo_work_queue_type * type,unsigned int queue_count,void * contexts[])416 int vdo_make_thread(struct vdo *vdo, thread_id_t thread_id,
417 const struct vdo_work_queue_type *type,
418 unsigned int queue_count, void *contexts[])
419 {
420 struct vdo_thread *thread = &vdo->threads[thread_id];
421 char queue_name[MAX_VDO_WORK_QUEUE_NAME_LEN];
422
423 if (type == NULL)
424 type = &default_queue_type;
425
426 if (thread->queue != NULL) {
427 return VDO_ASSERT(vdo_work_queue_type_is(thread->queue, type),
428 "already constructed vdo thread %u is of the correct type",
429 thread_id);
430 }
431
432 thread->vdo = vdo;
433 thread->thread_id = thread_id;
434 get_thread_name(&vdo->thread_config, thread_id, queue_name, sizeof(queue_name));
435 return vdo_make_work_queue(vdo->thread_name_prefix, queue_name, thread,
436 type, queue_count, contexts, &thread->queue);
437 }
438
439 /**
440 * register_vdo() - Register a VDO; it must not already be registered.
441 * @vdo: The vdo to register.
442 *
443 * Return: VDO_SUCCESS or an error.
444 */
register_vdo(struct vdo * vdo)445 static int register_vdo(struct vdo *vdo)
446 {
447 int result;
448
449 write_lock(®istry.lock);
450 result = VDO_ASSERT(filter_vdos_locked(vdo_is_equal, vdo) == NULL,
451 "VDO not already registered");
452 if (result == VDO_SUCCESS) {
453 INIT_LIST_HEAD(&vdo->registration);
454 list_add_tail(&vdo->registration, ®istry.links);
455 }
456 write_unlock(®istry.lock);
457
458 return result;
459 }
460
461 /**
462 * initialize_vdo() - Do the portion of initializing a vdo which will clean up after itself on
463 * error.
464 * @vdo: The vdo being initialized
465 * @config: The configuration of the vdo
466 * @instance: The instance number of the vdo
467 * @reason: The buffer to hold the failure reason on error
468 */
initialize_vdo(struct vdo * vdo,struct device_config * config,unsigned int instance,char ** reason)469 static int initialize_vdo(struct vdo *vdo, struct device_config *config,
470 unsigned int instance, char **reason)
471 {
472 int result;
473 zone_count_t i;
474
475 vdo->device_config = config;
476 vdo->starting_sector_offset = config->owning_target->begin;
477 vdo->instance = instance;
478 vdo->allocations_allowed = true;
479 vdo_set_admin_state_code(&vdo->admin.state, VDO_ADMIN_STATE_NEW);
480 INIT_LIST_HEAD(&vdo->device_config_list);
481 vdo_initialize_completion(&vdo->admin.completion, vdo, VDO_ADMIN_COMPLETION);
482 init_completion(&vdo->admin.callback_sync);
483 mutex_init(&vdo->stats_mutex);
484 result = read_geometry_block(vdo);
485 if (result != VDO_SUCCESS) {
486 *reason = "Could not load geometry block";
487 return result;
488 }
489
490 result = initialize_thread_config(config->thread_counts, &vdo->thread_config);
491 if (result != VDO_SUCCESS) {
492 *reason = "Cannot create thread configuration";
493 return result;
494 }
495
496 vdo_log_info("zones: %d logical, %d physical, %d hash; total threads: %d",
497 config->thread_counts.logical_zones,
498 config->thread_counts.physical_zones,
499 config->thread_counts.hash_zones, vdo->thread_config.thread_count);
500
501 /* Compression context storage */
502 result = vdo_allocate(config->thread_counts.cpu_threads, char *, "LZ4 context",
503 &vdo->compression_context);
504 if (result != VDO_SUCCESS) {
505 *reason = "cannot allocate LZ4 context";
506 return result;
507 }
508
509 for (i = 0; i < config->thread_counts.cpu_threads; i++) {
510 result = vdo_allocate(LZ4_MEM_COMPRESS, char, "LZ4 context",
511 &vdo->compression_context[i]);
512 if (result != VDO_SUCCESS) {
513 *reason = "cannot allocate LZ4 context";
514 return result;
515 }
516 }
517
518 result = register_vdo(vdo);
519 if (result != VDO_SUCCESS) {
520 *reason = "Cannot add VDO to device registry";
521 return result;
522 }
523
524 vdo_set_admin_state_code(&vdo->admin.state, VDO_ADMIN_STATE_INITIALIZED);
525 return result;
526 }
527
528 /**
529 * vdo_make() - Allocate and initialize a vdo.
530 * @instance: Device instantiation counter.
531 * @config: The device configuration.
532 * @reason: The reason for any failure during this call.
533 * @vdo_ptr: A pointer to hold the created vdo.
534 *
535 * Return: VDO_SUCCESS or an error.
536 */
vdo_make(unsigned int instance,struct device_config * config,char ** reason,struct vdo ** vdo_ptr)537 int vdo_make(unsigned int instance, struct device_config *config, char **reason,
538 struct vdo **vdo_ptr)
539 {
540 int result;
541 struct vdo *vdo;
542
543 /* Initialize with a generic failure reason to prevent returning garbage. */
544 *reason = "Unspecified error";
545
546 result = vdo_allocate(1, struct vdo, __func__, &vdo);
547 if (result != VDO_SUCCESS) {
548 *reason = "Cannot allocate VDO";
549 return result;
550 }
551
552 result = initialize_vdo(vdo, config, instance, reason);
553 if (result != VDO_SUCCESS) {
554 vdo_destroy(vdo);
555 return result;
556 }
557
558 /* From here on, the caller will clean up if there is an error. */
559 *vdo_ptr = vdo;
560
561 snprintf(vdo->thread_name_prefix, sizeof(vdo->thread_name_prefix),
562 "%s%u", MODULE_NAME, instance);
563 BUG_ON(vdo->thread_name_prefix[0] == '\0');
564 result = vdo_allocate(vdo->thread_config.thread_count,
565 struct vdo_thread, __func__, &vdo->threads);
566 if (result != VDO_SUCCESS) {
567 *reason = "Cannot allocate thread structures";
568 return result;
569 }
570
571 result = vdo_make_thread(vdo, vdo->thread_config.admin_thread,
572 &default_queue_type, 1, NULL);
573 if (result != VDO_SUCCESS) {
574 *reason = "Cannot make admin thread";
575 return result;
576 }
577
578 result = vdo_make_flusher(vdo);
579 if (result != VDO_SUCCESS) {
580 *reason = "Cannot make flusher zones";
581 return result;
582 }
583
584 result = vdo_make_packer(vdo, DEFAULT_PACKER_BINS, &vdo->packer);
585 if (result != VDO_SUCCESS) {
586 *reason = "Cannot make packer zones";
587 return result;
588 }
589
590 BUG_ON(vdo->device_config->logical_block_size <= 0);
591 BUG_ON(vdo->device_config->owned_device == NULL);
592 result = make_data_vio_pool(vdo, MAXIMUM_VDO_USER_VIOS,
593 MAXIMUM_VDO_USER_VIOS * 3 / 4,
594 &vdo->data_vio_pool);
595 if (result != VDO_SUCCESS) {
596 *reason = "Cannot allocate data_vio pool";
597 return result;
598 }
599
600 result = vdo_make_io_submitter(config->thread_counts.bio_threads,
601 config->thread_counts.bio_rotation_interval,
602 get_data_vio_pool_request_limit(vdo->data_vio_pool),
603 vdo, &vdo->io_submitter);
604 if (result != VDO_SUCCESS) {
605 *reason = "bio submission initialization failed";
606 return result;
607 }
608
609 if (vdo_uses_bio_ack_queue(vdo)) {
610 result = vdo_make_thread(vdo, vdo->thread_config.bio_ack_thread,
611 &bio_ack_q_type,
612 config->thread_counts.bio_ack_threads, NULL);
613 if (result != VDO_SUCCESS) {
614 *reason = "bio ack queue initialization failed";
615 return result;
616 }
617 }
618
619 result = vdo_make_thread(vdo, vdo->thread_config.cpu_thread, &cpu_q_type,
620 config->thread_counts.cpu_threads,
621 (void **) vdo->compression_context);
622 if (result != VDO_SUCCESS) {
623 *reason = "CPU queue initialization failed";
624 return result;
625 }
626
627 return VDO_SUCCESS;
628 }
629
finish_vdo(struct vdo * vdo)630 static void finish_vdo(struct vdo *vdo)
631 {
632 int i;
633
634 if (vdo->threads == NULL)
635 return;
636
637 vdo_cleanup_io_submitter(vdo->io_submitter);
638 vdo_finish_dedupe_index(vdo->hash_zones);
639
640 for (i = 0; i < vdo->thread_config.thread_count; i++)
641 vdo_finish_work_queue(vdo->threads[i].queue);
642 }
643
644 /**
645 * free_listeners() - Free the list of read-only listeners associated with a thread.
646 * @thread_data: The thread holding the list to free.
647 */
free_listeners(struct vdo_thread * thread)648 static void free_listeners(struct vdo_thread *thread)
649 {
650 struct read_only_listener *listener, *next;
651
652 for (listener = vdo_forget(thread->listeners); listener != NULL; listener = next) {
653 next = vdo_forget(listener->next);
654 vdo_free(listener);
655 }
656 }
657
uninitialize_super_block(struct vdo_super_block * super_block)658 static void uninitialize_super_block(struct vdo_super_block *super_block)
659 {
660 free_vio_components(&super_block->vio);
661 vdo_free(super_block->buffer);
662 }
663
664 /**
665 * unregister_vdo() - Remove a vdo from the device registry.
666 * @vdo: The vdo to remove.
667 */
unregister_vdo(struct vdo * vdo)668 static void unregister_vdo(struct vdo *vdo)
669 {
670 write_lock(®istry.lock);
671 if (filter_vdos_locked(vdo_is_equal, vdo) == vdo)
672 list_del_init(&vdo->registration);
673
674 write_unlock(®istry.lock);
675 }
676
677 /**
678 * vdo_destroy() - Destroy a vdo instance.
679 * @vdo: The vdo to destroy (may be NULL).
680 */
vdo_destroy(struct vdo * vdo)681 void vdo_destroy(struct vdo *vdo)
682 {
683 unsigned int i;
684
685 if (vdo == NULL)
686 return;
687
688 /* A running VDO should never be destroyed without suspending first. */
689 BUG_ON(vdo_get_admin_state(vdo)->normal);
690
691 vdo->allocations_allowed = true;
692
693 finish_vdo(vdo);
694 unregister_vdo(vdo);
695 free_data_vio_pool(vdo->data_vio_pool);
696 vdo_free_io_submitter(vdo_forget(vdo->io_submitter));
697 vdo_free_flusher(vdo_forget(vdo->flusher));
698 vdo_free_packer(vdo_forget(vdo->packer));
699 vdo_free_recovery_journal(vdo_forget(vdo->recovery_journal));
700 vdo_free_slab_depot(vdo_forget(vdo->depot));
701 vdo_uninitialize_layout(&vdo->layout);
702 vdo_uninitialize_layout(&vdo->next_layout);
703 if (vdo->partition_copier)
704 dm_kcopyd_client_destroy(vdo_forget(vdo->partition_copier));
705 uninitialize_super_block(&vdo->super_block);
706 vdo_free_block_map(vdo_forget(vdo->block_map));
707 vdo_free_hash_zones(vdo_forget(vdo->hash_zones));
708 vdo_free_physical_zones(vdo_forget(vdo->physical_zones));
709 vdo_free_logical_zones(vdo_forget(vdo->logical_zones));
710
711 if (vdo->threads != NULL) {
712 for (i = 0; i < vdo->thread_config.thread_count; i++) {
713 free_listeners(&vdo->threads[i]);
714 vdo_free_work_queue(vdo_forget(vdo->threads[i].queue));
715 }
716 vdo_free(vdo_forget(vdo->threads));
717 }
718
719 uninitialize_thread_config(&vdo->thread_config);
720
721 if (vdo->compression_context != NULL) {
722 for (i = 0; i < vdo->device_config->thread_counts.cpu_threads; i++)
723 vdo_free(vdo_forget(vdo->compression_context[i]));
724
725 vdo_free(vdo_forget(vdo->compression_context));
726 }
727 vdo_free(vdo);
728 }
729
initialize_super_block(struct vdo * vdo,struct vdo_super_block * super_block)730 static int initialize_super_block(struct vdo *vdo, struct vdo_super_block *super_block)
731 {
732 int result;
733
734 result = vdo_allocate(VDO_BLOCK_SIZE, char, "encoded super block",
735 (char **) &vdo->super_block.buffer);
736 if (result != VDO_SUCCESS)
737 return result;
738
739 return allocate_vio_components(vdo, VIO_TYPE_SUPER_BLOCK,
740 VIO_PRIORITY_METADATA, NULL, 1,
741 (char *) super_block->buffer,
742 &vdo->super_block.vio);
743 }
744
745 /**
746 * finish_reading_super_block() - Continue after loading the super block.
747 * @completion: The super block vio.
748 *
749 * This callback is registered in vdo_load_super_block().
750 */
finish_reading_super_block(struct vdo_completion * completion)751 static void finish_reading_super_block(struct vdo_completion *completion)
752 {
753 struct vdo_super_block *super_block =
754 container_of(as_vio(completion), struct vdo_super_block, vio);
755
756 vdo_continue_completion(vdo_forget(completion->parent),
757 vdo_decode_super_block(super_block->buffer));
758 }
759
760 /**
761 * handle_super_block_read_error() - Handle an error reading the super block.
762 * @completion: The super block vio.
763 *
764 * This error handler is registered in vdo_load_super_block().
765 */
handle_super_block_read_error(struct vdo_completion * completion)766 static void handle_super_block_read_error(struct vdo_completion *completion)
767 {
768 vio_record_metadata_io_error(as_vio(completion));
769 finish_reading_super_block(completion);
770 }
771
read_super_block_endio(struct bio * bio)772 static void read_super_block_endio(struct bio *bio)
773 {
774 struct vio *vio = bio->bi_private;
775 struct vdo_completion *parent = vio->completion.parent;
776
777 continue_vio_after_io(vio, finish_reading_super_block,
778 parent->callback_thread_id);
779 }
780
781 /**
782 * vdo_load_super_block() - Allocate a super block and read its contents from storage.
783 * @vdo: The vdo containing the super block on disk.
784 * @parent: The completion to notify after loading the super block.
785 */
vdo_load_super_block(struct vdo * vdo,struct vdo_completion * parent)786 void vdo_load_super_block(struct vdo *vdo, struct vdo_completion *parent)
787 {
788 int result;
789
790 result = initialize_super_block(vdo, &vdo->super_block);
791 if (result != VDO_SUCCESS) {
792 vdo_continue_completion(parent, result);
793 return;
794 }
795
796 vdo->super_block.vio.completion.parent = parent;
797 vdo_submit_metadata_vio(&vdo->super_block.vio,
798 vdo_get_data_region_start(vdo->geometry),
799 read_super_block_endio,
800 handle_super_block_read_error,
801 REQ_OP_READ);
802 }
803
804 /**
805 * vdo_get_backing_device() - Get the block device object underlying a vdo.
806 * @vdo: The vdo.
807 *
808 * Return: The vdo's current block device.
809 */
vdo_get_backing_device(const struct vdo * vdo)810 struct block_device *vdo_get_backing_device(const struct vdo *vdo)
811 {
812 return vdo->device_config->owned_device->bdev;
813 }
814
815 /**
816 * vdo_get_device_name() - Get the device name associated with the vdo target.
817 * @target: The target device interface.
818 *
819 * Return: The block device name.
820 */
vdo_get_device_name(const struct dm_target * target)821 const char *vdo_get_device_name(const struct dm_target *target)
822 {
823 return dm_device_name(dm_table_get_md(target->table));
824 }
825
826 /**
827 * vdo_synchronous_flush() - Issue a flush request and wait for it to complete.
828 * @vdo: The vdo.
829 *
830 * Return: VDO_SUCCESS or an error.
831 */
vdo_synchronous_flush(struct vdo * vdo)832 int vdo_synchronous_flush(struct vdo *vdo)
833 {
834 int result;
835 struct bio bio;
836
837 bio_init(&bio, vdo_get_backing_device(vdo), NULL, 0,
838 REQ_OP_WRITE | REQ_PREFLUSH);
839 submit_bio_wait(&bio);
840 result = blk_status_to_errno(bio.bi_status);
841
842 atomic64_inc(&vdo->stats.flush_out);
843 if (result != 0) {
844 vdo_log_error_strerror(result, "synchronous flush failed");
845 result = -EIO;
846 }
847
848 bio_uninit(&bio);
849 return result;
850 }
851
852 /**
853 * vdo_get_state() - Get the current state of the vdo.
854 * @vdo: The vdo.
855
856 * Context: This method may be called from any thread.
857 *
858 * Return: The current state of the vdo.
859 */
vdo_get_state(const struct vdo * vdo)860 enum vdo_state vdo_get_state(const struct vdo *vdo)
861 {
862 enum vdo_state state = atomic_read(&vdo->state);
863
864 /* pairs with barriers where state field is changed */
865 smp_rmb();
866 return state;
867 }
868
869 /**
870 * vdo_set_state() - Set the current state of the vdo.
871 * @vdo: The vdo whose state is to be set.
872 * @state: The new state of the vdo.
873 *
874 * Context: This method may be called from any thread.
875 */
vdo_set_state(struct vdo * vdo,enum vdo_state state)876 void vdo_set_state(struct vdo *vdo, enum vdo_state state)
877 {
878 /* pairs with barrier in vdo_get_state */
879 smp_wmb();
880 atomic_set(&vdo->state, state);
881 }
882
883 /**
884 * vdo_get_admin_state() - Get the admin state of the vdo.
885 * @vdo: The vdo.
886 *
887 * Return: The code for the vdo's current admin state.
888 */
vdo_get_admin_state(const struct vdo * vdo)889 const struct admin_state_code *vdo_get_admin_state(const struct vdo *vdo)
890 {
891 return vdo_get_admin_state_code(&vdo->admin.state);
892 }
893
894 /**
895 * record_vdo() - Record the state of the VDO for encoding in the super block.
896 */
record_vdo(struct vdo * vdo)897 static void record_vdo(struct vdo *vdo)
898 {
899 /* This is for backwards compatibility. */
900 vdo->states.unused = vdo->geometry.unused;
901 vdo->states.vdo.state = vdo_get_state(vdo);
902 vdo->states.block_map = vdo_record_block_map(vdo->block_map);
903 vdo->states.recovery_journal = vdo_record_recovery_journal(vdo->recovery_journal);
904 vdo->states.slab_depot = vdo_record_slab_depot(vdo->depot);
905 vdo->states.layout = vdo->layout;
906 }
907
908 /**
909 * continue_super_block_parent() - Continue the parent of a super block save operation.
910 * @completion: The super block vio.
911 *
912 * This callback is registered in vdo_save_components().
913 */
continue_super_block_parent(struct vdo_completion * completion)914 static void continue_super_block_parent(struct vdo_completion *completion)
915 {
916 vdo_continue_completion(vdo_forget(completion->parent), completion->result);
917 }
918
919 /**
920 * handle_save_error() - Log a super block save error.
921 * @completion: The super block vio.
922 *
923 * This error handler is registered in vdo_save_components().
924 */
handle_save_error(struct vdo_completion * completion)925 static void handle_save_error(struct vdo_completion *completion)
926 {
927 struct vdo_super_block *super_block =
928 container_of(as_vio(completion), struct vdo_super_block, vio);
929
930 vio_record_metadata_io_error(&super_block->vio);
931 vdo_log_error_strerror(completion->result, "super block save failed");
932 /*
933 * Mark the super block as unwritable so that we won't attempt to write it again. This
934 * avoids the case where a growth attempt fails writing the super block with the new size,
935 * but the subsequent attempt to write out the read-only state succeeds. In this case,
936 * writes which happened just before the suspend would not be visible if the VDO is
937 * restarted without rebuilding, but, after a read-only rebuild, the effects of those
938 * writes would reappear.
939 */
940 super_block->unwritable = true;
941 completion->callback(completion);
942 }
943
super_block_write_endio(struct bio * bio)944 static void super_block_write_endio(struct bio *bio)
945 {
946 struct vio *vio = bio->bi_private;
947 struct vdo_completion *parent = vio->completion.parent;
948
949 continue_vio_after_io(vio, continue_super_block_parent,
950 parent->callback_thread_id);
951 }
952
953 /**
954 * vdo_save_components() - Encode the vdo and save the super block asynchronously.
955 * @vdo: The vdo whose state is being saved.
956 * @parent: The completion to notify when the save is complete.
957 */
vdo_save_components(struct vdo * vdo,struct vdo_completion * parent)958 void vdo_save_components(struct vdo *vdo, struct vdo_completion *parent)
959 {
960 struct vdo_super_block *super_block = &vdo->super_block;
961
962 if (super_block->unwritable) {
963 vdo_continue_completion(parent, VDO_READ_ONLY);
964 return;
965 }
966
967 if (super_block->vio.completion.parent != NULL) {
968 vdo_continue_completion(parent, VDO_COMPONENT_BUSY);
969 return;
970 }
971
972 record_vdo(vdo);
973
974 vdo_encode_super_block(super_block->buffer, &vdo->states);
975 super_block->vio.completion.parent = parent;
976 super_block->vio.completion.callback_thread_id = parent->callback_thread_id;
977 vdo_submit_metadata_vio(&super_block->vio,
978 vdo_get_data_region_start(vdo->geometry),
979 super_block_write_endio, handle_save_error,
980 REQ_OP_WRITE | REQ_PREFLUSH | REQ_FUA);
981 }
982
983 /**
984 * vdo_register_read_only_listener() - Register a listener to be notified when the VDO goes
985 * read-only.
986 * @vdo: The vdo to register with.
987 * @listener: The object to notify.
988 * @notification: The function to call to send the notification.
989 * @thread_id: The id of the thread on which to send the notification.
990 *
991 * Return: VDO_SUCCESS or an error.
992 */
vdo_register_read_only_listener(struct vdo * vdo,void * listener,vdo_read_only_notification_fn notification,thread_id_t thread_id)993 int vdo_register_read_only_listener(struct vdo *vdo, void *listener,
994 vdo_read_only_notification_fn notification,
995 thread_id_t thread_id)
996 {
997 struct vdo_thread *thread = &vdo->threads[thread_id];
998 struct read_only_listener *read_only_listener;
999 int result;
1000
1001 result = VDO_ASSERT(thread_id != vdo->thread_config.dedupe_thread,
1002 "read only listener not registered on dedupe thread");
1003 if (result != VDO_SUCCESS)
1004 return result;
1005
1006 result = vdo_allocate(1, struct read_only_listener, __func__,
1007 &read_only_listener);
1008 if (result != VDO_SUCCESS)
1009 return result;
1010
1011 *read_only_listener = (struct read_only_listener) {
1012 .listener = listener,
1013 .notify = notification,
1014 .next = thread->listeners,
1015 };
1016
1017 thread->listeners = read_only_listener;
1018 return VDO_SUCCESS;
1019 }
1020
1021 /**
1022 * notify_vdo_of_read_only_mode() - Notify a vdo that it is going read-only.
1023 * @listener: The vdo.
1024 * @parent: The completion to notify in order to acknowledge the notification.
1025 *
1026 * This will save the read-only state to the super block.
1027 *
1028 * Implements vdo_read_only_notification_fn.
1029 */
notify_vdo_of_read_only_mode(void * listener,struct vdo_completion * parent)1030 static void notify_vdo_of_read_only_mode(void *listener, struct vdo_completion *parent)
1031 {
1032 struct vdo *vdo = listener;
1033
1034 if (vdo_in_read_only_mode(vdo))
1035 vdo_finish_completion(parent);
1036
1037 vdo_set_state(vdo, VDO_READ_ONLY_MODE);
1038 vdo_save_components(vdo, parent);
1039 }
1040
1041 /**
1042 * vdo_enable_read_only_entry() - Enable a vdo to enter read-only mode on errors.
1043 * @vdo: The vdo to enable.
1044 *
1045 * Return: VDO_SUCCESS or an error.
1046 */
vdo_enable_read_only_entry(struct vdo * vdo)1047 int vdo_enable_read_only_entry(struct vdo *vdo)
1048 {
1049 thread_id_t id;
1050 bool is_read_only = vdo_in_read_only_mode(vdo);
1051 struct read_only_notifier *notifier = &vdo->read_only_notifier;
1052
1053 if (is_read_only) {
1054 notifier->read_only_error = VDO_READ_ONLY;
1055 notifier->state = NOTIFIED;
1056 } else {
1057 notifier->state = MAY_NOT_NOTIFY;
1058 }
1059
1060 spin_lock_init(¬ifier->lock);
1061 vdo_initialize_completion(¬ifier->completion, vdo,
1062 VDO_READ_ONLY_MODE_COMPLETION);
1063
1064 for (id = 0; id < vdo->thread_config.thread_count; id++)
1065 vdo->threads[id].is_read_only = is_read_only;
1066
1067 return vdo_register_read_only_listener(vdo, vdo, notify_vdo_of_read_only_mode,
1068 vdo->thread_config.admin_thread);
1069 }
1070
1071 /**
1072 * vdo_wait_until_not_entering_read_only_mode() - Wait until no read-only notifications are in
1073 * progress and prevent any subsequent
1074 * notifications.
1075 * @parent: The completion to notify when no threads are entering read-only mode.
1076 *
1077 * Notifications may be re-enabled by calling vdo_allow_read_only_mode_entry().
1078 */
vdo_wait_until_not_entering_read_only_mode(struct vdo_completion * parent)1079 void vdo_wait_until_not_entering_read_only_mode(struct vdo_completion *parent)
1080 {
1081 struct vdo *vdo = parent->vdo;
1082 struct read_only_notifier *notifier = &vdo->read_only_notifier;
1083
1084 vdo_assert_on_admin_thread(vdo, __func__);
1085
1086 if (notifier->waiter != NULL) {
1087 vdo_continue_completion(parent, VDO_COMPONENT_BUSY);
1088 return;
1089 }
1090
1091 spin_lock(¬ifier->lock);
1092 if (notifier->state == NOTIFYING)
1093 notifier->waiter = parent;
1094 else if (notifier->state == MAY_NOTIFY)
1095 notifier->state = MAY_NOT_NOTIFY;
1096 spin_unlock(¬ifier->lock);
1097
1098 if (notifier->waiter == NULL) {
1099 /*
1100 * A notification was not in progress, and now they are
1101 * disallowed.
1102 */
1103 vdo_launch_completion(parent);
1104 return;
1105 }
1106 }
1107
1108 /**
1109 * as_notifier() - Convert a generic vdo_completion to a read_only_notifier.
1110 * @completion: The completion to convert.
1111 *
1112 * Return: The completion as a read_only_notifier.
1113 */
as_notifier(struct vdo_completion * completion)1114 static inline struct read_only_notifier *as_notifier(struct vdo_completion *completion)
1115 {
1116 vdo_assert_completion_type(completion, VDO_READ_ONLY_MODE_COMPLETION);
1117 return container_of(completion, struct read_only_notifier, completion);
1118 }
1119
1120 /**
1121 * finish_entering_read_only_mode() - Complete the process of entering read only mode.
1122 * @completion: The read-only mode completion.
1123 */
finish_entering_read_only_mode(struct vdo_completion * completion)1124 static void finish_entering_read_only_mode(struct vdo_completion *completion)
1125 {
1126 struct read_only_notifier *notifier = as_notifier(completion);
1127
1128 vdo_assert_on_admin_thread(completion->vdo, __func__);
1129
1130 spin_lock(¬ifier->lock);
1131 notifier->state = NOTIFIED;
1132 spin_unlock(¬ifier->lock);
1133
1134 if (notifier->waiter != NULL)
1135 vdo_continue_completion(vdo_forget(notifier->waiter),
1136 completion->result);
1137 }
1138
1139 /**
1140 * make_thread_read_only() - Inform each thread that the VDO is in read-only mode.
1141 * @completion: The read-only mode completion.
1142 */
make_thread_read_only(struct vdo_completion * completion)1143 static void make_thread_read_only(struct vdo_completion *completion)
1144 {
1145 struct vdo *vdo = completion->vdo;
1146 thread_id_t thread_id = completion->callback_thread_id;
1147 struct read_only_notifier *notifier = as_notifier(completion);
1148 struct read_only_listener *listener = completion->parent;
1149
1150 if (listener == NULL) {
1151 /* This is the first call on this thread */
1152 struct vdo_thread *thread = &vdo->threads[thread_id];
1153
1154 thread->is_read_only = true;
1155 listener = thread->listeners;
1156 if (thread_id == 0)
1157 vdo_log_error_strerror(READ_ONCE(notifier->read_only_error),
1158 "Unrecoverable error, entering read-only mode");
1159 } else {
1160 /* We've just finished notifying a listener */
1161 listener = listener->next;
1162 }
1163
1164 if (listener != NULL) {
1165 /* We have a listener to notify */
1166 vdo_prepare_completion(completion, make_thread_read_only,
1167 make_thread_read_only, thread_id,
1168 listener);
1169 listener->notify(listener->listener, completion);
1170 return;
1171 }
1172
1173 /* We're done with this thread */
1174 if (++thread_id == vdo->thread_config.dedupe_thread) {
1175 /*
1176 * We don't want to notify the dedupe thread since it may be
1177 * blocked rebuilding the index.
1178 */
1179 thread_id++;
1180 }
1181
1182 if (thread_id >= vdo->thread_config.thread_count) {
1183 /* There are no more threads */
1184 vdo_prepare_completion(completion, finish_entering_read_only_mode,
1185 finish_entering_read_only_mode,
1186 vdo->thread_config.admin_thread, NULL);
1187 } else {
1188 vdo_prepare_completion(completion, make_thread_read_only,
1189 make_thread_read_only, thread_id, NULL);
1190 }
1191
1192 vdo_launch_completion(completion);
1193 }
1194
1195 /**
1196 * vdo_allow_read_only_mode_entry() - Allow the notifier to put the VDO into read-only mode,
1197 * reversing the effects of
1198 * vdo_wait_until_not_entering_read_only_mode().
1199 * @parent: The object to notify once the operation is complete.
1200 *
1201 * If some thread tried to put the vdo into read-only mode while notifications were disallowed, it
1202 * will be done when this method is called. If that happens, the parent will not be notified until
1203 * the vdo has actually entered read-only mode and attempted to save the super block.
1204 *
1205 * Context: This method may only be called from the admin thread.
1206 */
vdo_allow_read_only_mode_entry(struct vdo_completion * parent)1207 void vdo_allow_read_only_mode_entry(struct vdo_completion *parent)
1208 {
1209 struct vdo *vdo = parent->vdo;
1210 struct read_only_notifier *notifier = &vdo->read_only_notifier;
1211
1212 vdo_assert_on_admin_thread(vdo, __func__);
1213
1214 if (notifier->waiter != NULL) {
1215 vdo_continue_completion(parent, VDO_COMPONENT_BUSY);
1216 return;
1217 }
1218
1219 spin_lock(¬ifier->lock);
1220 if (notifier->state == MAY_NOT_NOTIFY) {
1221 if (notifier->read_only_error == VDO_SUCCESS) {
1222 notifier->state = MAY_NOTIFY;
1223 } else {
1224 notifier->state = NOTIFYING;
1225 notifier->waiter = parent;
1226 }
1227 }
1228 spin_unlock(¬ifier->lock);
1229
1230 if (notifier->waiter == NULL) {
1231 /* We're done */
1232 vdo_launch_completion(parent);
1233 return;
1234 }
1235
1236 /* Do the pending notification. */
1237 make_thread_read_only(¬ifier->completion);
1238 }
1239
1240 /**
1241 * vdo_enter_read_only_mode() - Put a VDO into read-only mode and save the read-only state in the
1242 * super block.
1243 * @vdo: The vdo.
1244 * @error_code: The error which caused the VDO to enter read-only mode.
1245 *
1246 * This method is a no-op if the VDO is already read-only.
1247 */
vdo_enter_read_only_mode(struct vdo * vdo,int error_code)1248 void vdo_enter_read_only_mode(struct vdo *vdo, int error_code)
1249 {
1250 bool notify = false;
1251 thread_id_t thread_id = vdo_get_callback_thread_id();
1252 struct read_only_notifier *notifier = &vdo->read_only_notifier;
1253 struct vdo_thread *thread;
1254
1255 if (thread_id != VDO_INVALID_THREAD_ID) {
1256 thread = &vdo->threads[thread_id];
1257 if (thread->is_read_only) {
1258 /* This thread has already gone read-only. */
1259 return;
1260 }
1261
1262 /* Record for this thread that the VDO is read-only. */
1263 thread->is_read_only = true;
1264 }
1265
1266 spin_lock(¬ifier->lock);
1267 if (notifier->read_only_error == VDO_SUCCESS) {
1268 WRITE_ONCE(notifier->read_only_error, error_code);
1269 if (notifier->state == MAY_NOTIFY) {
1270 notifier->state = NOTIFYING;
1271 notify = true;
1272 }
1273 }
1274 spin_unlock(¬ifier->lock);
1275
1276 if (!notify) {
1277 /* The notifier is already aware of a read-only error */
1278 return;
1279 }
1280
1281 /* Initiate a notification starting on the lowest numbered thread. */
1282 vdo_launch_completion_callback(¬ifier->completion, make_thread_read_only, 0);
1283 }
1284
1285 /**
1286 * vdo_is_read_only() - Check whether the VDO is read-only.
1287 * @vdo: The vdo.
1288 *
1289 * Return: true if the vdo is read-only.
1290 *
1291 * This method may be called from any thread, as opposed to examining the VDO's state field which
1292 * is only safe to check from the admin thread.
1293 */
vdo_is_read_only(struct vdo * vdo)1294 bool vdo_is_read_only(struct vdo *vdo)
1295 {
1296 return vdo->threads[vdo_get_callback_thread_id()].is_read_only;
1297 }
1298
1299 /**
1300 * vdo_in_read_only_mode() - Check whether a vdo is in read-only mode.
1301 * @vdo: The vdo to query.
1302 *
1303 * Return: true if the vdo is in read-only mode.
1304 */
vdo_in_read_only_mode(const struct vdo * vdo)1305 bool vdo_in_read_only_mode(const struct vdo *vdo)
1306 {
1307 return (vdo_get_state(vdo) == VDO_READ_ONLY_MODE);
1308 }
1309
1310 /**
1311 * vdo_in_recovery_mode() - Check whether the vdo is in recovery mode.
1312 * @vdo: The vdo to query.
1313 *
1314 * Return: true if the vdo is in recovery mode.
1315 */
vdo_in_recovery_mode(const struct vdo * vdo)1316 bool vdo_in_recovery_mode(const struct vdo *vdo)
1317 {
1318 return (vdo_get_state(vdo) == VDO_RECOVERING);
1319 }
1320
1321 /**
1322 * vdo_enter_recovery_mode() - Put the vdo into recovery mode.
1323 * @vdo: The vdo.
1324 */
vdo_enter_recovery_mode(struct vdo * vdo)1325 void vdo_enter_recovery_mode(struct vdo *vdo)
1326 {
1327 vdo_assert_on_admin_thread(vdo, __func__);
1328
1329 if (vdo_in_read_only_mode(vdo))
1330 return;
1331
1332 vdo_log_info("Entering recovery mode");
1333 vdo_set_state(vdo, VDO_RECOVERING);
1334 }
1335
1336 /**
1337 * complete_synchronous_action() - Signal the waiting thread that a synchronous action is complete.
1338 * @completion: The sync completion.
1339 */
complete_synchronous_action(struct vdo_completion * completion)1340 static void complete_synchronous_action(struct vdo_completion *completion)
1341 {
1342 vdo_assert_completion_type(completion, VDO_SYNC_COMPLETION);
1343 complete(&(container_of(completion, struct sync_completion,
1344 vdo_completion)->completion));
1345 }
1346
1347 /**
1348 * perform_synchronous_action() - Launch an action on a VDO thread and wait for it to complete.
1349 * @vdo: The vdo.
1350 * @action: The callback to launch.
1351 * @thread_id: The thread on which to run the action.
1352 * @parent: The parent of the sync completion (may be NULL).
1353 */
perform_synchronous_action(struct vdo * vdo,vdo_action_fn action,thread_id_t thread_id,void * parent)1354 static int perform_synchronous_action(struct vdo *vdo, vdo_action_fn action,
1355 thread_id_t thread_id, void *parent)
1356 {
1357 struct sync_completion sync;
1358
1359 vdo_initialize_completion(&sync.vdo_completion, vdo, VDO_SYNC_COMPLETION);
1360 init_completion(&sync.completion);
1361 sync.vdo_completion.parent = parent;
1362 vdo_launch_completion_callback(&sync.vdo_completion, action, thread_id);
1363 wait_for_completion(&sync.completion);
1364 return sync.vdo_completion.result;
1365 }
1366
1367 /**
1368 * set_compression_callback() - Callback to turn compression on or off.
1369 * @completion: The completion.
1370 */
set_compression_callback(struct vdo_completion * completion)1371 static void set_compression_callback(struct vdo_completion *completion)
1372 {
1373 struct vdo *vdo = completion->vdo;
1374 bool *enable = completion->parent;
1375 bool was_enabled = vdo_get_compressing(vdo);
1376
1377 if (*enable != was_enabled) {
1378 WRITE_ONCE(vdo->compressing, *enable);
1379 if (was_enabled) {
1380 /* Signal the packer to flush since compression has been disabled. */
1381 vdo_flush_packer(vdo->packer);
1382 }
1383 }
1384
1385 vdo_log_info("compression is %s", (*enable ? "enabled" : "disabled"));
1386 *enable = was_enabled;
1387 complete_synchronous_action(completion);
1388 }
1389
1390 /**
1391 * vdo_set_compressing() - Turn compression on or off.
1392 * @vdo: The vdo.
1393 * @enable: Whether to enable or disable compression.
1394 *
1395 * Return: Whether compression was previously on or off.
1396 */
vdo_set_compressing(struct vdo * vdo,bool enable)1397 bool vdo_set_compressing(struct vdo *vdo, bool enable)
1398 {
1399 perform_synchronous_action(vdo, set_compression_callback,
1400 vdo->thread_config.packer_thread,
1401 &enable);
1402 return enable;
1403 }
1404
1405 /**
1406 * vdo_get_compressing() - Get whether compression is enabled in a vdo.
1407 * @vdo: The vdo.
1408 *
1409 * Return: State of compression.
1410 */
vdo_get_compressing(struct vdo * vdo)1411 bool vdo_get_compressing(struct vdo *vdo)
1412 {
1413 return READ_ONCE(vdo->compressing);
1414 }
1415
get_block_map_cache_size(const struct vdo * vdo)1416 static size_t get_block_map_cache_size(const struct vdo *vdo)
1417 {
1418 return ((size_t) vdo->device_config->cache_size) * VDO_BLOCK_SIZE;
1419 }
1420
get_vdo_error_statistics(const struct vdo * vdo)1421 static struct error_statistics __must_check get_vdo_error_statistics(const struct vdo *vdo)
1422 {
1423 /*
1424 * The error counts can be incremented from arbitrary threads and so must be incremented
1425 * atomically, but they are just statistics with no semantics that could rely on memory
1426 * order, so unfenced reads are sufficient.
1427 */
1428 const struct atomic_statistics *atoms = &vdo->stats;
1429
1430 return (struct error_statistics) {
1431 .invalid_advice_pbn_count = atomic64_read(&atoms->invalid_advice_pbn_count),
1432 .no_space_error_count = atomic64_read(&atoms->no_space_error_count),
1433 .read_only_error_count = atomic64_read(&atoms->read_only_error_count),
1434 };
1435 }
1436
copy_bio_stat(struct bio_stats * b,const struct atomic_bio_stats * a)1437 static void copy_bio_stat(struct bio_stats *b, const struct atomic_bio_stats *a)
1438 {
1439 b->read = atomic64_read(&a->read);
1440 b->write = atomic64_read(&a->write);
1441 b->discard = atomic64_read(&a->discard);
1442 b->flush = atomic64_read(&a->flush);
1443 b->empty_flush = atomic64_read(&a->empty_flush);
1444 b->fua = atomic64_read(&a->fua);
1445 }
1446
subtract_bio_stats(struct bio_stats minuend,struct bio_stats subtrahend)1447 static struct bio_stats subtract_bio_stats(struct bio_stats minuend,
1448 struct bio_stats subtrahend)
1449 {
1450 return (struct bio_stats) {
1451 .read = minuend.read - subtrahend.read,
1452 .write = minuend.write - subtrahend.write,
1453 .discard = minuend.discard - subtrahend.discard,
1454 .flush = minuend.flush - subtrahend.flush,
1455 .empty_flush = minuend.empty_flush - subtrahend.empty_flush,
1456 .fua = minuend.fua - subtrahend.fua,
1457 };
1458 }
1459
1460 /**
1461 * vdo_get_physical_blocks_allocated() - Get the number of physical blocks in use by user data.
1462 * @vdo: The vdo.
1463 *
1464 * Return: The number of blocks allocated for user data.
1465 */
vdo_get_physical_blocks_allocated(const struct vdo * vdo)1466 static block_count_t __must_check vdo_get_physical_blocks_allocated(const struct vdo *vdo)
1467 {
1468 return (vdo_get_slab_depot_allocated_blocks(vdo->depot) -
1469 vdo_get_journal_block_map_data_blocks_used(vdo->recovery_journal));
1470 }
1471
1472 /**
1473 * vdo_get_physical_blocks_overhead() - Get the number of physical blocks used by vdo metadata.
1474 * @vdo: The vdo.
1475 *
1476 * Return: The number of overhead blocks.
1477 */
vdo_get_physical_blocks_overhead(const struct vdo * vdo)1478 static block_count_t __must_check vdo_get_physical_blocks_overhead(const struct vdo *vdo)
1479 {
1480 /*
1481 * config.physical_blocks is mutated during resize and is in a packed structure,
1482 * but resize runs on admin thread.
1483 * TODO: Verify that this is always safe.
1484 */
1485 return (vdo->states.vdo.config.physical_blocks -
1486 vdo_get_slab_depot_data_blocks(vdo->depot) +
1487 vdo_get_journal_block_map_data_blocks_used(vdo->recovery_journal));
1488 }
1489
vdo_describe_state(enum vdo_state state)1490 static const char *vdo_describe_state(enum vdo_state state)
1491 {
1492 /* These strings should all fit in the 15 chars of VDOStatistics.mode. */
1493 switch (state) {
1494 case VDO_RECOVERING:
1495 return "recovering";
1496
1497 case VDO_READ_ONLY_MODE:
1498 return "read-only";
1499
1500 default:
1501 return "normal";
1502 }
1503 }
1504
1505 /**
1506 * get_vdo_statistics() - Populate a vdo_statistics structure on the admin thread.
1507 * @vdo: The vdo.
1508 * @stats: The statistics structure to populate.
1509 */
get_vdo_statistics(const struct vdo * vdo,struct vdo_statistics * stats)1510 static void get_vdo_statistics(const struct vdo *vdo, struct vdo_statistics *stats)
1511 {
1512 struct recovery_journal *journal = vdo->recovery_journal;
1513 enum vdo_state state = vdo_get_state(vdo);
1514
1515 vdo_assert_on_admin_thread(vdo, __func__);
1516
1517 /* start with a clean slate */
1518 memset(stats, 0, sizeof(struct vdo_statistics));
1519
1520 /*
1521 * These are immutable properties of the vdo object, so it is safe to query them from any
1522 * thread.
1523 */
1524 stats->version = STATISTICS_VERSION;
1525 stats->logical_blocks = vdo->states.vdo.config.logical_blocks;
1526 /*
1527 * config.physical_blocks is mutated during resize and is in a packed structure, but resize
1528 * runs on the admin thread.
1529 * TODO: verify that this is always safe
1530 */
1531 stats->physical_blocks = vdo->states.vdo.config.physical_blocks;
1532 stats->block_size = VDO_BLOCK_SIZE;
1533 stats->complete_recoveries = vdo->states.vdo.complete_recoveries;
1534 stats->read_only_recoveries = vdo->states.vdo.read_only_recoveries;
1535 stats->block_map_cache_size = get_block_map_cache_size(vdo);
1536
1537 /* The callees are responsible for thread-safety. */
1538 stats->data_blocks_used = vdo_get_physical_blocks_allocated(vdo);
1539 stats->overhead_blocks_used = vdo_get_physical_blocks_overhead(vdo);
1540 stats->logical_blocks_used = vdo_get_recovery_journal_logical_blocks_used(journal);
1541 vdo_get_slab_depot_statistics(vdo->depot, stats);
1542 stats->journal = vdo_get_recovery_journal_statistics(journal);
1543 stats->packer = vdo_get_packer_statistics(vdo->packer);
1544 stats->block_map = vdo_get_block_map_statistics(vdo->block_map);
1545 vdo_get_dedupe_statistics(vdo->hash_zones, stats);
1546 stats->errors = get_vdo_error_statistics(vdo);
1547 stats->in_recovery_mode = (state == VDO_RECOVERING);
1548 snprintf(stats->mode, sizeof(stats->mode), "%s", vdo_describe_state(state));
1549
1550 stats->instance = vdo->instance;
1551 stats->current_vios_in_progress = get_data_vio_pool_active_requests(vdo->data_vio_pool);
1552 stats->max_vios = get_data_vio_pool_maximum_requests(vdo->data_vio_pool);
1553
1554 stats->flush_out = atomic64_read(&vdo->stats.flush_out);
1555 stats->logical_block_size = vdo->device_config->logical_block_size;
1556 copy_bio_stat(&stats->bios_in, &vdo->stats.bios_in);
1557 copy_bio_stat(&stats->bios_in_partial, &vdo->stats.bios_in_partial);
1558 copy_bio_stat(&stats->bios_out, &vdo->stats.bios_out);
1559 copy_bio_stat(&stats->bios_meta, &vdo->stats.bios_meta);
1560 copy_bio_stat(&stats->bios_journal, &vdo->stats.bios_journal);
1561 copy_bio_stat(&stats->bios_page_cache, &vdo->stats.bios_page_cache);
1562 copy_bio_stat(&stats->bios_out_completed, &vdo->stats.bios_out_completed);
1563 copy_bio_stat(&stats->bios_meta_completed, &vdo->stats.bios_meta_completed);
1564 copy_bio_stat(&stats->bios_journal_completed,
1565 &vdo->stats.bios_journal_completed);
1566 copy_bio_stat(&stats->bios_page_cache_completed,
1567 &vdo->stats.bios_page_cache_completed);
1568 copy_bio_stat(&stats->bios_acknowledged, &vdo->stats.bios_acknowledged);
1569 copy_bio_stat(&stats->bios_acknowledged_partial, &vdo->stats.bios_acknowledged_partial);
1570 stats->bios_in_progress =
1571 subtract_bio_stats(stats->bios_in, stats->bios_acknowledged);
1572 vdo_get_memory_stats(&stats->memory_usage.bytes_used,
1573 &stats->memory_usage.peak_bytes_used);
1574 }
1575
1576 /**
1577 * vdo_fetch_statistics_callback() - Action to populate a vdo_statistics
1578 * structure on the admin thread.
1579 * @completion: The completion.
1580 *
1581 * This callback is registered in vdo_fetch_statistics().
1582 */
vdo_fetch_statistics_callback(struct vdo_completion * completion)1583 static void vdo_fetch_statistics_callback(struct vdo_completion *completion)
1584 {
1585 get_vdo_statistics(completion->vdo, completion->parent);
1586 complete_synchronous_action(completion);
1587 }
1588
1589 /**
1590 * vdo_fetch_statistics() - Fetch statistics on the correct thread.
1591 * @vdo: The vdo.
1592 * @stats: The vdo statistics are returned here.
1593 */
vdo_fetch_statistics(struct vdo * vdo,struct vdo_statistics * stats)1594 void vdo_fetch_statistics(struct vdo *vdo, struct vdo_statistics *stats)
1595 {
1596 perform_synchronous_action(vdo, vdo_fetch_statistics_callback,
1597 vdo->thread_config.admin_thread, stats);
1598 }
1599
1600 /**
1601 * vdo_get_callback_thread_id() - Get the id of the callback thread on which a completion is
1602 * currently running.
1603 *
1604 * Return: The current thread ID, or -1 if no such thread.
1605 */
vdo_get_callback_thread_id(void)1606 thread_id_t vdo_get_callback_thread_id(void)
1607 {
1608 struct vdo_work_queue *queue = vdo_get_current_work_queue();
1609 struct vdo_thread *thread;
1610 thread_id_t thread_id;
1611
1612 if (queue == NULL)
1613 return VDO_INVALID_THREAD_ID;
1614
1615 thread = vdo_get_work_queue_owner(queue);
1616 thread_id = thread->thread_id;
1617
1618 if (PARANOID_THREAD_CONSISTENCY_CHECKS) {
1619 BUG_ON(thread_id >= thread->vdo->thread_config.thread_count);
1620 BUG_ON(thread != &thread->vdo->threads[thread_id]);
1621 }
1622
1623 return thread_id;
1624 }
1625
1626 /**
1627 * vdo_dump_status() - Dump status information about a vdo to the log for debugging.
1628 * @vdo: The vdo to dump.
1629 */
vdo_dump_status(const struct vdo * vdo)1630 void vdo_dump_status(const struct vdo *vdo)
1631 {
1632 zone_count_t zone;
1633
1634 vdo_dump_flusher(vdo->flusher);
1635 vdo_dump_recovery_journal_statistics(vdo->recovery_journal);
1636 vdo_dump_packer(vdo->packer);
1637 vdo_dump_slab_depot(vdo->depot);
1638
1639 for (zone = 0; zone < vdo->thread_config.logical_zone_count; zone++)
1640 vdo_dump_logical_zone(&vdo->logical_zones->zones[zone]);
1641
1642 for (zone = 0; zone < vdo->thread_config.physical_zone_count; zone++)
1643 vdo_dump_physical_zone(&vdo->physical_zones->zones[zone]);
1644
1645 vdo_dump_hash_zones(vdo->hash_zones);
1646 }
1647
1648 /**
1649 * vdo_assert_on_admin_thread() - Assert that we are running on the admin thread.
1650 * @vdo: The vdo.
1651 * @name: The name of the function which should be running on the admin thread (for logging).
1652 */
vdo_assert_on_admin_thread(const struct vdo * vdo,const char * name)1653 void vdo_assert_on_admin_thread(const struct vdo *vdo, const char *name)
1654 {
1655 VDO_ASSERT_LOG_ONLY((vdo_get_callback_thread_id() == vdo->thread_config.admin_thread),
1656 "%s called on admin thread", name);
1657 }
1658
1659 /**
1660 * vdo_assert_on_logical_zone_thread() - Assert that this function was called on the specified
1661 * logical zone thread.
1662 * @vdo: The vdo.
1663 * @logical_zone: The number of the logical zone.
1664 * @name: The name of the calling function.
1665 */
vdo_assert_on_logical_zone_thread(const struct vdo * vdo,zone_count_t logical_zone,const char * name)1666 void vdo_assert_on_logical_zone_thread(const struct vdo *vdo, zone_count_t logical_zone,
1667 const char *name)
1668 {
1669 VDO_ASSERT_LOG_ONLY((vdo_get_callback_thread_id() ==
1670 vdo->thread_config.logical_threads[logical_zone]),
1671 "%s called on logical thread", name);
1672 }
1673
1674 /**
1675 * vdo_assert_on_physical_zone_thread() - Assert that this function was called on the specified
1676 * physical zone thread.
1677 * @vdo: The vdo.
1678 * @physical_zone: The number of the physical zone.
1679 * @name: The name of the calling function.
1680 */
vdo_assert_on_physical_zone_thread(const struct vdo * vdo,zone_count_t physical_zone,const char * name)1681 void vdo_assert_on_physical_zone_thread(const struct vdo *vdo,
1682 zone_count_t physical_zone, const char *name)
1683 {
1684 VDO_ASSERT_LOG_ONLY((vdo_get_callback_thread_id() ==
1685 vdo->thread_config.physical_threads[physical_zone]),
1686 "%s called on physical thread", name);
1687 }
1688
1689 /**
1690 * vdo_get_physical_zone() - Get the physical zone responsible for a given physical block number.
1691 * @vdo: The vdo containing the physical zones.
1692 * @pbn: The PBN of the data block.
1693 * @zone_ptr: A pointer to return the physical zone.
1694 *
1695 * Gets the physical zone responsible for a given physical block number of a data block in this vdo
1696 * instance, or of the zero block (for which a NULL zone is returned). For any other block number
1697 * that is not in the range of valid data block numbers in any slab, an error will be returned.
1698 * This function is safe to call on invalid block numbers; it will not put the vdo into read-only
1699 * mode.
1700 *
1701 * Return: VDO_SUCCESS or VDO_OUT_OF_RANGE if the block number is invalid or an error code for any
1702 * other failure.
1703 */
vdo_get_physical_zone(const struct vdo * vdo,physical_block_number_t pbn,struct physical_zone ** zone_ptr)1704 int vdo_get_physical_zone(const struct vdo *vdo, physical_block_number_t pbn,
1705 struct physical_zone **zone_ptr)
1706 {
1707 struct vdo_slab *slab;
1708 int result;
1709
1710 if (pbn == VDO_ZERO_BLOCK) {
1711 *zone_ptr = NULL;
1712 return VDO_SUCCESS;
1713 }
1714
1715 /*
1716 * Used because it does a more restrictive bounds check than vdo_get_slab(), and done first
1717 * because it won't trigger read-only mode on an invalid PBN.
1718 */
1719 if (!vdo_is_physical_data_block(vdo->depot, pbn))
1720 return VDO_OUT_OF_RANGE;
1721
1722 /* With the PBN already checked, we should always succeed in finding a slab. */
1723 slab = vdo_get_slab(vdo->depot, pbn);
1724 result = VDO_ASSERT(slab != NULL, "vdo_get_slab must succeed on all valid PBNs");
1725 if (result != VDO_SUCCESS)
1726 return result;
1727
1728 *zone_ptr = &vdo->physical_zones->zones[slab->allocator->zone_number];
1729 return VDO_SUCCESS;
1730 }
1731