xref: /linux/drivers/md/dm-vdo/packer.c (revision 79790b6818e96c58fe2bffee1b418c16e64e7b80)
1 // SPDX-License-Identifier: GPL-2.0-only
2 /*
3  * Copyright 2023 Red Hat
4  */
5 
6 #include "packer.h"
7 
8 #include <linux/atomic.h>
9 #include <linux/blkdev.h>
10 
11 #include "logger.h"
12 #include "memory-alloc.h"
13 #include "permassert.h"
14 #include "string-utils.h"
15 
16 #include "admin-state.h"
17 #include "completion.h"
18 #include "constants.h"
19 #include "data-vio.h"
20 #include "dedupe.h"
21 #include "encodings.h"
22 #include "io-submitter.h"
23 #include "physical-zone.h"
24 #include "status-codes.h"
25 #include "vdo.h"
26 #include "vio.h"
27 
28 static const struct version_number COMPRESSED_BLOCK_1_0 = {
29 	.major_version = 1,
30 	.minor_version = 0,
31 };
32 
33 #define COMPRESSED_BLOCK_1_0_SIZE (4 + 4 + (2 * VDO_MAX_COMPRESSION_SLOTS))
34 
35 /**
36  * vdo_get_compressed_block_fragment() - Get a reference to a compressed fragment from a compressed
37  *                                       block.
38  * @mapping_state [in] The mapping state for the look up.
39  * @compressed_block [in] The compressed block that was read from disk.
40  * @fragment_offset [out] The offset of the fragment within a compressed block.
41  * @fragment_size [out] The size of the fragment.
42  *
43  * Return: If a valid compressed fragment is found, VDO_SUCCESS; otherwise, VDO_INVALID_FRAGMENT if
44  *         the fragment is invalid.
45  */
vdo_get_compressed_block_fragment(enum block_mapping_state mapping_state,struct compressed_block * block,u16 * fragment_offset,u16 * fragment_size)46 int vdo_get_compressed_block_fragment(enum block_mapping_state mapping_state,
47 				      struct compressed_block *block,
48 				      u16 *fragment_offset, u16 *fragment_size)
49 {
50 	u16 compressed_size;
51 	u16 offset = 0;
52 	unsigned int i;
53 	u8 slot;
54 	struct version_number version;
55 
56 	if (!vdo_is_state_compressed(mapping_state))
57 		return VDO_INVALID_FRAGMENT;
58 
59 	version = vdo_unpack_version_number(block->header.version);
60 	if (!vdo_are_same_version(version, COMPRESSED_BLOCK_1_0))
61 		return VDO_INVALID_FRAGMENT;
62 
63 	slot = mapping_state - VDO_MAPPING_STATE_COMPRESSED_BASE;
64 	if (slot >= VDO_MAX_COMPRESSION_SLOTS)
65 		return VDO_INVALID_FRAGMENT;
66 
67 	compressed_size = __le16_to_cpu(block->header.sizes[slot]);
68 	for (i = 0; i < slot; i++) {
69 		offset += __le16_to_cpu(block->header.sizes[i]);
70 		if (offset >= VDO_COMPRESSED_BLOCK_DATA_SIZE)
71 			return VDO_INVALID_FRAGMENT;
72 	}
73 
74 	if ((offset + compressed_size) > VDO_COMPRESSED_BLOCK_DATA_SIZE)
75 		return VDO_INVALID_FRAGMENT;
76 
77 	*fragment_offset = offset;
78 	*fragment_size = compressed_size;
79 	return VDO_SUCCESS;
80 }
81 
82 /**
83  * assert_on_packer_thread() - Check that we are on the packer thread.
84  * @packer: The packer.
85  * @caller: The function which is asserting.
86  */
assert_on_packer_thread(struct packer * packer,const char * caller)87 static inline void assert_on_packer_thread(struct packer *packer, const char *caller)
88 {
89 	VDO_ASSERT_LOG_ONLY((vdo_get_callback_thread_id() == packer->thread_id),
90 			    "%s() called from packer thread", caller);
91 }
92 
93 /**
94  * insert_in_sorted_list() - Insert a bin to the list.
95  * @packer: The packer.
96  * @bin: The bin to move to its sorted position.
97  *
98  * The list is in ascending order of free space. Since all bins are already in the list, this
99  * actually moves the bin to the correct position in the list.
100  */
insert_in_sorted_list(struct packer * packer,struct packer_bin * bin)101 static void insert_in_sorted_list(struct packer *packer, struct packer_bin *bin)
102 {
103 	struct packer_bin *active_bin;
104 
105 	list_for_each_entry(active_bin, &packer->bins, list)
106 		if (active_bin->free_space > bin->free_space) {
107 			list_move_tail(&bin->list, &active_bin->list);
108 			return;
109 		}
110 
111 	list_move_tail(&bin->list, &packer->bins);
112 }
113 
114 /**
115  * make_bin() - Allocate a bin and put it into the packer's list.
116  * @packer: The packer.
117  */
make_bin(struct packer * packer)118 static int __must_check make_bin(struct packer *packer)
119 {
120 	struct packer_bin *bin;
121 	int result;
122 
123 	result = vdo_allocate_extended(struct packer_bin, VDO_MAX_COMPRESSION_SLOTS,
124 				       struct vio *, __func__, &bin);
125 	if (result != VDO_SUCCESS)
126 		return result;
127 
128 	bin->free_space = VDO_COMPRESSED_BLOCK_DATA_SIZE;
129 	INIT_LIST_HEAD(&bin->list);
130 	list_add_tail(&bin->list, &packer->bins);
131 	return VDO_SUCCESS;
132 }
133 
134 /**
135  * vdo_make_packer() - Make a new block packer.
136  *
137  * @vdo: The vdo to which this packer belongs.
138  * @bin_count: The number of partial bins to keep in memory.
139  * @packer_ptr: A pointer to hold the new packer.
140  *
141  * Return: VDO_SUCCESS or an error
142  */
vdo_make_packer(struct vdo * vdo,block_count_t bin_count,struct packer ** packer_ptr)143 int vdo_make_packer(struct vdo *vdo, block_count_t bin_count, struct packer **packer_ptr)
144 {
145 	struct packer *packer;
146 	block_count_t i;
147 	int result;
148 
149 	result = vdo_allocate(1, struct packer, __func__, &packer);
150 	if (result != VDO_SUCCESS)
151 		return result;
152 
153 	packer->thread_id = vdo->thread_config.packer_thread;
154 	packer->size = bin_count;
155 	INIT_LIST_HEAD(&packer->bins);
156 	vdo_set_admin_state_code(&packer->state, VDO_ADMIN_STATE_NORMAL_OPERATION);
157 
158 	for (i = 0; i < bin_count; i++) {
159 		result = make_bin(packer);
160 		if (result != VDO_SUCCESS) {
161 			vdo_free_packer(packer);
162 			return result;
163 		}
164 	}
165 
166 	/*
167 	 * The canceled bin can hold up to half the number of user vios. Every canceled vio in the
168 	 * bin must have a canceler for which it is waiting, and any canceler will only have
169 	 * canceled one lock holder at a time.
170 	 */
171 	result = vdo_allocate_extended(struct packer_bin, MAXIMUM_VDO_USER_VIOS / 2,
172 				       struct vio *, __func__, &packer->canceled_bin);
173 	if (result != VDO_SUCCESS) {
174 		vdo_free_packer(packer);
175 		return result;
176 	}
177 
178 	result = vdo_make_default_thread(vdo, packer->thread_id);
179 	if (result != VDO_SUCCESS) {
180 		vdo_free_packer(packer);
181 		return result;
182 	}
183 
184 	*packer_ptr = packer;
185 	return VDO_SUCCESS;
186 }
187 
188 /**
189  * vdo_free_packer() - Free a block packer.
190  * @packer: The packer to free.
191  */
vdo_free_packer(struct packer * packer)192 void vdo_free_packer(struct packer *packer)
193 {
194 	struct packer_bin *bin, *tmp;
195 
196 	if (packer == NULL)
197 		return;
198 
199 	list_for_each_entry_safe(bin, tmp, &packer->bins, list) {
200 		list_del_init(&bin->list);
201 		vdo_free(bin);
202 	}
203 
204 	vdo_free(vdo_forget(packer->canceled_bin));
205 	vdo_free(packer);
206 }
207 
208 /**
209  * get_packer_from_data_vio() - Get the packer from a data_vio.
210  * @data_vio: The data_vio.
211  *
212  * Return: The packer from the VDO to which the data_vio belongs.
213  */
get_packer_from_data_vio(struct data_vio * data_vio)214 static inline struct packer *get_packer_from_data_vio(struct data_vio *data_vio)
215 {
216 	return vdo_from_data_vio(data_vio)->packer;
217 }
218 
219 /**
220  * vdo_get_packer_statistics() - Get the current statistics from the packer.
221  * @packer: The packer to query.
222  *
223  * Return: a copy of the current statistics for the packer.
224  */
vdo_get_packer_statistics(const struct packer * packer)225 struct packer_statistics vdo_get_packer_statistics(const struct packer *packer)
226 {
227 	const struct packer_statistics *stats = &packer->statistics;
228 
229 	return (struct packer_statistics) {
230 		.compressed_fragments_written = READ_ONCE(stats->compressed_fragments_written),
231 		.compressed_blocks_written = READ_ONCE(stats->compressed_blocks_written),
232 		.compressed_fragments_in_packer = READ_ONCE(stats->compressed_fragments_in_packer),
233 	};
234 }
235 
236 /**
237  * abort_packing() - Abort packing a data_vio.
238  * @data_vio: The data_vio to abort.
239  */
abort_packing(struct data_vio * data_vio)240 static void abort_packing(struct data_vio *data_vio)
241 {
242 	struct packer *packer = get_packer_from_data_vio(data_vio);
243 
244 	WRITE_ONCE(packer->statistics.compressed_fragments_in_packer,
245 		   packer->statistics.compressed_fragments_in_packer - 1);
246 
247 	write_data_vio(data_vio);
248 }
249 
250 /**
251  * release_compressed_write_waiter() - Update a data_vio for which a successful compressed write
252  *                                     has completed and send it on its way.
253 
254  * @data_vio: The data_vio to release.
255  * @allocation: The allocation to which the compressed block was written.
256  */
release_compressed_write_waiter(struct data_vio * data_vio,struct allocation * allocation)257 static void release_compressed_write_waiter(struct data_vio *data_vio,
258 					    struct allocation *allocation)
259 {
260 	data_vio->new_mapped = (struct zoned_pbn) {
261 		.pbn = allocation->pbn,
262 		.zone = allocation->zone,
263 		.state = data_vio->compression.slot + VDO_MAPPING_STATE_COMPRESSED_BASE,
264 	};
265 
266 	vdo_share_compressed_write_lock(data_vio, allocation->lock);
267 	update_metadata_for_data_vio_write(data_vio, allocation->lock);
268 }
269 
270 /**
271  * finish_compressed_write() - Finish a compressed block write.
272  * @completion: The compressed write completion.
273  *
274  * This callback is registered in continue_after_allocation().
275  */
finish_compressed_write(struct vdo_completion * completion)276 static void finish_compressed_write(struct vdo_completion *completion)
277 {
278 	struct data_vio *agent = as_data_vio(completion);
279 	struct data_vio *client, *next;
280 
281 	assert_data_vio_in_allocated_zone(agent);
282 
283 	/*
284 	 * Process all the non-agent waiters first to ensure that the pbn lock can not be released
285 	 * until all of them have had a chance to journal their increfs.
286 	 */
287 	for (client = agent->compression.next_in_batch; client != NULL; client = next) {
288 		next = client->compression.next_in_batch;
289 		release_compressed_write_waiter(client, &agent->allocation);
290 	}
291 
292 	completion->error_handler = handle_data_vio_error;
293 	release_compressed_write_waiter(agent, &agent->allocation);
294 }
295 
handle_compressed_write_error(struct vdo_completion * completion)296 static void handle_compressed_write_error(struct vdo_completion *completion)
297 {
298 	struct data_vio *agent = as_data_vio(completion);
299 	struct allocation *allocation = &agent->allocation;
300 	struct data_vio *client, *next;
301 
302 	if (vdo_requeue_completion_if_needed(completion, allocation->zone->thread_id))
303 		return;
304 
305 	update_vio_error_stats(as_vio(completion),
306 			       "Completing compressed write vio for physical block %llu with error",
307 			       (unsigned long long) allocation->pbn);
308 
309 	for (client = agent->compression.next_in_batch; client != NULL; client = next) {
310 		next = client->compression.next_in_batch;
311 		write_data_vio(client);
312 	}
313 
314 	/* Now that we've released the batch from the packer, forget the error and continue on. */
315 	vdo_reset_completion(completion);
316 	completion->error_handler = handle_data_vio_error;
317 	write_data_vio(agent);
318 }
319 
320 /**
321  * add_to_bin() - Put a data_vio in a specific packer_bin in which it will definitely fit.
322  * @bin: The bin in which to put the data_vio.
323  * @data_vio: The data_vio to add.
324  */
add_to_bin(struct packer_bin * bin,struct data_vio * data_vio)325 static void add_to_bin(struct packer_bin *bin, struct data_vio *data_vio)
326 {
327 	data_vio->compression.bin = bin;
328 	data_vio->compression.slot = bin->slots_used;
329 	bin->incoming[bin->slots_used++] = data_vio;
330 }
331 
332 /**
333  * remove_from_bin() - Get the next data_vio whose compression has not been canceled from a bin.
334  * @packer: The packer.
335  * @bin: The bin from which to get a data_vio.
336  *
337  * Any canceled data_vios will be moved to the canceled bin.
338  * Return: An uncanceled data_vio from the bin or NULL if there are none.
339  */
remove_from_bin(struct packer * packer,struct packer_bin * bin)340 static struct data_vio *remove_from_bin(struct packer *packer, struct packer_bin *bin)
341 {
342 	while (bin->slots_used > 0) {
343 		struct data_vio *data_vio = bin->incoming[--bin->slots_used];
344 
345 		if (!advance_data_vio_compression_stage(data_vio).may_not_compress) {
346 			data_vio->compression.bin = NULL;
347 			return data_vio;
348 		}
349 
350 		add_to_bin(packer->canceled_bin, data_vio);
351 	}
352 
353 	/* The bin is now empty. */
354 	bin->free_space = VDO_COMPRESSED_BLOCK_DATA_SIZE;
355 	return NULL;
356 }
357 
358 /**
359  * initialize_compressed_block() - Initialize a compressed block.
360  * @block: The compressed block to initialize.
361  * @size: The size of the agent's fragment.
362  *
363  * This method initializes the compressed block in the compressed write agent. Because the
364  * compressor already put the agent's compressed fragment at the start of the compressed block's
365  * data field, it needn't be copied. So all we need do is initialize the header and set the size of
366  * the agent's fragment.
367  */
initialize_compressed_block(struct compressed_block * block,u16 size)368 static void initialize_compressed_block(struct compressed_block *block, u16 size)
369 {
370 	/*
371 	 * Make sure the block layout isn't accidentally changed by changing the length of the
372 	 * block header.
373 	 */
374 	BUILD_BUG_ON(sizeof(struct compressed_block_header) != COMPRESSED_BLOCK_1_0_SIZE);
375 
376 	block->header.version = vdo_pack_version_number(COMPRESSED_BLOCK_1_0);
377 	block->header.sizes[0] = __cpu_to_le16(size);
378 }
379 
380 /**
381  * pack_fragment() - Pack a data_vio's fragment into the compressed block in which it is already
382  *                   known to fit.
383  * @compression: The agent's compression_state to pack in to.
384  * @data_vio: The data_vio to pack.
385  * @offset: The offset into the compressed block at which to pack the fragment.
386  * @compressed_block: The compressed block which will be written out when batch is fully packed.
387  *
388  * Return: The new amount of space used.
389  */
pack_fragment(struct compression_state * compression,struct data_vio * data_vio,block_size_t offset,slot_number_t slot,struct compressed_block * block)390 static block_size_t __must_check pack_fragment(struct compression_state *compression,
391 					       struct data_vio *data_vio,
392 					       block_size_t offset, slot_number_t slot,
393 					       struct compressed_block *block)
394 {
395 	struct compression_state *to_pack = &data_vio->compression;
396 	char *fragment = to_pack->block->data;
397 
398 	to_pack->next_in_batch = compression->next_in_batch;
399 	compression->next_in_batch = data_vio;
400 	to_pack->slot = slot;
401 	block->header.sizes[slot] = __cpu_to_le16(to_pack->size);
402 	memcpy(&block->data[offset], fragment, to_pack->size);
403 	return (offset + to_pack->size);
404 }
405 
406 /**
407  * compressed_write_end_io() - The bio_end_io for a compressed block write.
408  * @bio: The bio for the compressed write.
409  */
compressed_write_end_io(struct bio * bio)410 static void compressed_write_end_io(struct bio *bio)
411 {
412 	struct data_vio *data_vio = vio_as_data_vio(bio->bi_private);
413 
414 	vdo_count_completed_bios(bio);
415 	set_data_vio_allocated_zone_callback(data_vio, finish_compressed_write);
416 	continue_data_vio_with_error(data_vio, blk_status_to_errno(bio->bi_status));
417 }
418 
419 /**
420  * write_bin() - Write out a bin.
421  * @packer: The packer.
422  * @bin: The bin to write.
423  */
write_bin(struct packer * packer,struct packer_bin * bin)424 static void write_bin(struct packer *packer, struct packer_bin *bin)
425 {
426 	int result;
427 	block_size_t offset;
428 	slot_number_t slot = 1;
429 	struct compression_state *compression;
430 	struct compressed_block *block;
431 	struct data_vio *agent = remove_from_bin(packer, bin);
432 	struct data_vio *client;
433 	struct packer_statistics *stats;
434 
435 	if (agent == NULL)
436 		return;
437 
438 	compression = &agent->compression;
439 	compression->slot = 0;
440 	block = compression->block;
441 	initialize_compressed_block(block, compression->size);
442 	offset = compression->size;
443 
444 	while ((client = remove_from_bin(packer, bin)) != NULL)
445 		offset = pack_fragment(compression, client, offset, slot++, block);
446 
447 	/*
448 	 * If the batch contains only a single vio, then we save nothing by saving the compressed
449 	 * form. Continue processing the single vio in the batch.
450 	 */
451 	if (slot == 1) {
452 		abort_packing(agent);
453 		return;
454 	}
455 
456 	if (slot < VDO_MAX_COMPRESSION_SLOTS) {
457 		/* Clear out the sizes of the unused slots */
458 		memset(&block->header.sizes[slot], 0,
459 		       (VDO_MAX_COMPRESSION_SLOTS - slot) * sizeof(__le16));
460 	}
461 
462 	agent->vio.completion.error_handler = handle_compressed_write_error;
463 	if (vdo_is_read_only(vdo_from_data_vio(agent))) {
464 		continue_data_vio_with_error(agent, VDO_READ_ONLY);
465 		return;
466 	}
467 
468 	result = vio_reset_bio(&agent->vio, (char *) block, compressed_write_end_io,
469 			       REQ_OP_WRITE, agent->allocation.pbn);
470 	if (result != VDO_SUCCESS) {
471 		continue_data_vio_with_error(agent, result);
472 		return;
473 	}
474 
475 	/*
476 	 * Once the compressed write is submitted, the fragments are no longer in the packer, so
477 	 * update stats now.
478 	 */
479 	stats = &packer->statistics;
480 	WRITE_ONCE(stats->compressed_fragments_in_packer,
481 		   (stats->compressed_fragments_in_packer - slot));
482 	WRITE_ONCE(stats->compressed_fragments_written,
483 		   (stats->compressed_fragments_written + slot));
484 	WRITE_ONCE(stats->compressed_blocks_written,
485 		   stats->compressed_blocks_written + 1);
486 
487 	vdo_submit_data_vio(agent);
488 }
489 
490 /**
491  * add_data_vio_to_packer_bin() - Add a data_vio to a bin's incoming queue
492  * @packer: The packer.
493  * @bin: The bin to which to add the data_vio.
494  * @data_vio: The data_vio to add to the bin's queue.
495  *
496  * Adds a data_vio to a bin's incoming queue, handles logical space change, and calls physical
497  * space processor.
498  */
add_data_vio_to_packer_bin(struct packer * packer,struct packer_bin * bin,struct data_vio * data_vio)499 static void add_data_vio_to_packer_bin(struct packer *packer, struct packer_bin *bin,
500 				       struct data_vio *data_vio)
501 {
502 	/* If the selected bin doesn't have room, start a new batch to make room. */
503 	if (bin->free_space < data_vio->compression.size)
504 		write_bin(packer, bin);
505 
506 	add_to_bin(bin, data_vio);
507 	bin->free_space -= data_vio->compression.size;
508 
509 	/* If we happen to exactly fill the bin, start a new batch. */
510 	if ((bin->slots_used == VDO_MAX_COMPRESSION_SLOTS) ||
511 	    (bin->free_space == 0))
512 		write_bin(packer, bin);
513 
514 	/* Now that we've finished changing the free space, restore the sort order. */
515 	insert_in_sorted_list(packer, bin);
516 }
517 
518 /**
519  * select_bin() - Select the bin that should be used to pack the compressed data in a data_vio with
520  *                other data_vios.
521  * @packer: The packer.
522  * @data_vio: The data_vio.
523  */
select_bin(struct packer * packer,struct data_vio * data_vio)524 static struct packer_bin * __must_check select_bin(struct packer *packer,
525 						   struct data_vio *data_vio)
526 {
527 	/*
528 	 * First best fit: select the bin with the least free space that has enough room for the
529 	 * compressed data in the data_vio.
530 	 */
531 	struct packer_bin *bin, *fullest_bin;
532 
533 	list_for_each_entry(bin, &packer->bins, list) {
534 		if (bin->free_space >= data_vio->compression.size)
535 			return bin;
536 	}
537 
538 	/*
539 	 * None of the bins have enough space for the data_vio. We're not allowed to create new
540 	 * bins, so we have to overflow one of the existing bins. It's pretty intuitive to select
541 	 * the fullest bin, since that "wastes" the least amount of free space in the compressed
542 	 * block. But if the space currently used in the fullest bin is smaller than the compressed
543 	 * size of the incoming block, it seems wrong to force that bin to write when giving up on
544 	 * compressing the incoming data_vio would likewise "waste" the least amount of free space.
545 	 */
546 	fullest_bin = list_first_entry(&packer->bins, struct packer_bin, list);
547 	if (data_vio->compression.size >=
548 	    (VDO_COMPRESSED_BLOCK_DATA_SIZE - fullest_bin->free_space))
549 		return NULL;
550 
551 	/*
552 	 * The fullest bin doesn't have room, but writing it out and starting a new batch with the
553 	 * incoming data_vio will increase the packer's free space.
554 	 */
555 	return fullest_bin;
556 }
557 
558 /**
559  * vdo_attempt_packing() - Attempt to rewrite the data in this data_vio as part of a compressed
560  *                         block.
561  * @data_vio: The data_vio to pack.
562  */
vdo_attempt_packing(struct data_vio * data_vio)563 void vdo_attempt_packing(struct data_vio *data_vio)
564 {
565 	int result;
566 	struct packer_bin *bin;
567 	struct data_vio_compression_status status = get_data_vio_compression_status(data_vio);
568 	struct packer *packer = get_packer_from_data_vio(data_vio);
569 
570 	assert_on_packer_thread(packer, __func__);
571 
572 	result = VDO_ASSERT((status.stage == DATA_VIO_COMPRESSING),
573 			    "attempt to pack data_vio not ready for packing, stage: %u",
574 			    status.stage);
575 	if (result != VDO_SUCCESS)
576 		return;
577 
578 	/*
579 	 * Increment whether or not this data_vio will be packed or not since abort_packing()
580 	 * always decrements the counter.
581 	 */
582 	WRITE_ONCE(packer->statistics.compressed_fragments_in_packer,
583 		   packer->statistics.compressed_fragments_in_packer + 1);
584 
585 	/*
586 	 * If packing of this data_vio is disallowed for administrative reasons, give up before
587 	 * making any state changes.
588 	 */
589 	if (!vdo_is_state_normal(&packer->state) ||
590 	    (data_vio->flush_generation < packer->flush_generation)) {
591 		abort_packing(data_vio);
592 		return;
593 	}
594 
595 	/*
596 	 * The advance_data_vio_compression_stage() check here verifies that the data_vio is
597 	 * allowed to be compressed (if it has already been canceled, we'll fall out here). Once
598 	 * the data_vio is in the DATA_VIO_PACKING state, it must be guaranteed to be put in a bin
599 	 * before any more requests can be processed by the packer thread. Otherwise, a canceling
600 	 * data_vio could attempt to remove the canceled data_vio from the packer and fail to
601 	 * rendezvous with it. Thus, we must call select_bin() first to ensure that we will
602 	 * actually add the data_vio to a bin before advancing to the DATA_VIO_PACKING stage.
603 	 */
604 	bin = select_bin(packer, data_vio);
605 	if ((bin == NULL) ||
606 	    (advance_data_vio_compression_stage(data_vio).stage != DATA_VIO_PACKING)) {
607 		abort_packing(data_vio);
608 		return;
609 	}
610 
611 	add_data_vio_to_packer_bin(packer, bin, data_vio);
612 }
613 
614 /**
615  * check_for_drain_complete() - Check whether the packer has drained.
616  * @packer: The packer.
617  */
check_for_drain_complete(struct packer * packer)618 static void check_for_drain_complete(struct packer *packer)
619 {
620 	if (vdo_is_state_draining(&packer->state) && (packer->canceled_bin->slots_used == 0))
621 		vdo_finish_draining(&packer->state);
622 }
623 
624 /**
625  * write_all_non_empty_bins() - Write out all non-empty bins on behalf of a flush or suspend.
626  * @packer: The packer being flushed.
627  */
write_all_non_empty_bins(struct packer * packer)628 static void write_all_non_empty_bins(struct packer *packer)
629 {
630 	struct packer_bin *bin;
631 
632 	list_for_each_entry(bin, &packer->bins, list)
633 		write_bin(packer, bin);
634 		/*
635 		 * We don't need to re-sort the bin here since this loop will make every bin have
636 		 * the same amount of free space, so every ordering is sorted.
637 		 */
638 
639 	check_for_drain_complete(packer);
640 }
641 
642 /**
643  * vdo_flush_packer() - Request that the packer flush asynchronously.
644  * @packer: The packer to flush.
645  *
646  * All bins with at least two compressed data blocks will be written out, and any solitary pending
647  * VIOs will be released from the packer. While flushing is in progress, any VIOs submitted to
648  * vdo_attempt_packing() will be continued immediately without attempting to pack them.
649  */
vdo_flush_packer(struct packer * packer)650 void vdo_flush_packer(struct packer *packer)
651 {
652 	assert_on_packer_thread(packer, __func__);
653 	if (vdo_is_state_normal(&packer->state))
654 		write_all_non_empty_bins(packer);
655 }
656 
657 /**
658  * vdo_remove_lock_holder_from_packer() - Remove a lock holder from the packer.
659  * @completion: The data_vio which needs a lock held by a data_vio in the packer. The data_vio's
660  *              compression.lock_holder field will point to the data_vio to remove.
661  */
vdo_remove_lock_holder_from_packer(struct vdo_completion * completion)662 void vdo_remove_lock_holder_from_packer(struct vdo_completion *completion)
663 {
664 	struct data_vio *data_vio = as_data_vio(completion);
665 	struct packer *packer = get_packer_from_data_vio(data_vio);
666 	struct data_vio *lock_holder;
667 	struct packer_bin *bin;
668 	slot_number_t slot;
669 
670 	assert_data_vio_in_packer_zone(data_vio);
671 
672 	lock_holder = vdo_forget(data_vio->compression.lock_holder);
673 	bin = lock_holder->compression.bin;
674 	VDO_ASSERT_LOG_ONLY((bin != NULL), "data_vio in packer has a bin");
675 
676 	slot = lock_holder->compression.slot;
677 	bin->slots_used--;
678 	if (slot < bin->slots_used) {
679 		bin->incoming[slot] = bin->incoming[bin->slots_used];
680 		bin->incoming[slot]->compression.slot = slot;
681 	}
682 
683 	lock_holder->compression.bin = NULL;
684 	lock_holder->compression.slot = 0;
685 
686 	if (bin != packer->canceled_bin) {
687 		bin->free_space += lock_holder->compression.size;
688 		insert_in_sorted_list(packer, bin);
689 	}
690 
691 	abort_packing(lock_holder);
692 	check_for_drain_complete(packer);
693 }
694 
695 /**
696  * vdo_increment_packer_flush_generation() - Increment the flush generation in the packer.
697  * @packer: The packer.
698  *
699  * This will also cause the packer to flush so that any VIOs from previous generations will exit
700  * the packer.
701  */
vdo_increment_packer_flush_generation(struct packer * packer)702 void vdo_increment_packer_flush_generation(struct packer *packer)
703 {
704 	assert_on_packer_thread(packer, __func__);
705 	packer->flush_generation++;
706 	vdo_flush_packer(packer);
707 }
708 
709 /**
710  * initiate_drain() - Initiate a drain.
711  *
712  * Implements vdo_admin_initiator_fn.
713  */
initiate_drain(struct admin_state * state)714 static void initiate_drain(struct admin_state *state)
715 {
716 	struct packer *packer = container_of(state, struct packer, state);
717 
718 	write_all_non_empty_bins(packer);
719 }
720 
721 /**
722  * vdo_drain_packer() - Drain the packer by preventing any more VIOs from entering the packer and
723  *                      then flushing.
724  * @packer: The packer to drain.
725  * @completion: The completion to finish when the packer has drained.
726  */
vdo_drain_packer(struct packer * packer,struct vdo_completion * completion)727 void vdo_drain_packer(struct packer *packer, struct vdo_completion *completion)
728 {
729 	assert_on_packer_thread(packer, __func__);
730 	vdo_start_draining(&packer->state, VDO_ADMIN_STATE_SUSPENDING, completion,
731 			   initiate_drain);
732 }
733 
734 /**
735  * vdo_resume_packer() - Resume a packer which has been suspended.
736  * @packer: The packer to resume.
737  * @parent: The completion to finish when the packer has resumed.
738  */
vdo_resume_packer(struct packer * packer,struct vdo_completion * parent)739 void vdo_resume_packer(struct packer *packer, struct vdo_completion *parent)
740 {
741 	assert_on_packer_thread(packer, __func__);
742 	vdo_continue_completion(parent, vdo_resume_if_quiescent(&packer->state));
743 }
744 
dump_packer_bin(const struct packer_bin * bin,bool canceled)745 static void dump_packer_bin(const struct packer_bin *bin, bool canceled)
746 {
747 	if (bin->slots_used == 0)
748 		/* Don't dump empty bins. */
749 		return;
750 
751 	vdo_log_info("	  %sBin slots_used=%u free_space=%zu",
752 		     (canceled ? "Canceled" : ""), bin->slots_used, bin->free_space);
753 
754 	/*
755 	 * FIXME: dump vios in bin->incoming? The vios should have been dumped from the vio pool.
756 	 * Maybe just dump their addresses so it's clear they're here?
757 	 */
758 }
759 
760 /**
761  * vdo_dump_packer() - Dump the packer.
762  * @packer: The packer.
763  *
764  * Context: dumps in a thread-unsafe fashion.
765  */
vdo_dump_packer(const struct packer * packer)766 void vdo_dump_packer(const struct packer *packer)
767 {
768 	struct packer_bin *bin;
769 
770 	vdo_log_info("packer");
771 	vdo_log_info("	flushGeneration=%llu state %s  packer_bin_count=%llu",
772 		     (unsigned long long) packer->flush_generation,
773 		     vdo_get_admin_state_code(&packer->state)->name,
774 		     (unsigned long long) packer->size);
775 
776 	list_for_each_entry(bin, &packer->bins, list)
777 		dump_packer_bin(bin, false);
778 
779 	dump_packer_bin(packer->canceled_bin, true);
780 }
781