2849 lines · 68.0 KB · C
RawDownload
/* * Copyright (C) 2011 Red Hat UK. * * This file is released under the GPL. */ #include "dm-thin-metadata.h" #include <linux/device-mapper.h>#include <linux/dm-io.h>#include <linux/dm-kcopyd.h>#include <linux/list.h>#include <linux/init.h>#include <linux/module.h>#include <linux/slab.h> #define	DM_MSG_PREFIX	"thin" /* * Tunable constants */#define ENDIO_HOOK_POOL_SIZE 10240#define DEFERRED_SET_SIZE 64#define MAPPING_POOL_SIZE 1024#define PRISON_CELLS 1024#define COMMIT_PERIOD HZ /* * The block size of the device holding pool data must be * between 64KB and 1GB. */#define DATA_DEV_BLOCK_SIZE_MIN_SECTORS (64 * 1024 >> SECTOR_SHIFT)#define DATA_DEV_BLOCK_SIZE_MAX_SECTORS (1024 * 1024 * 1024 >> SECTOR_SHIFT) /* * Device id is restricted to 24 bits. */#define MAX_DEV_ID ((1 << 24) - 1) /* * How do we handle breaking sharing of data blocks? * ================================================= * * We use a standard copy-on-write btree to store the mappings for the * devices (note I'm talking about copy-on-write of the metadata here, not * the data).  When you take an internal snapshot you clone the root node * of the origin btree.  After this there is no concept of an origin or a * snapshot.  They are just two device trees that happen to point to the * same data blocks. * * When we get a write in we decide if it's to a shared data block using * some timestamp magic.  If it is, we have to break sharing. * * Let's say we write to a shared block in what was the origin.  The * steps are: * * i) plug io further to this physical block. (see bio_prison code). * * ii) quiesce any read io to that shared data block.  Obviously * including all devices that share this block.  (see deferred_set code) * * iii) copy the data block to a newly allocate block.  This step can be * missed out if the io covers the block. (schedule_copy). * * iv) insert the new mapping into the origin's btree * (process_prepared_mapping).  This act of inserting breaks some * sharing of btree nodes between the two devices.  Breaking sharing only * effects the btree of that specific device.  Btrees for the other * devices that share the block never change.  The btree for the origin * device as it was after the last commit is untouched, ie. we're using * persistent data structures in the functional programming sense. * * v) unplug io to this physical block, including the io that triggered * the breaking of sharing. * * Steps (ii) and (iii) occur in parallel. * * The metadata _doesn't_ need to be committed before the io continues.  We * get away with this because the io is always written to a _new_ block. * If there's a crash, then: * * - The origin mapping will point to the old origin block (the shared * one).  This will contain the data as it was before the io that triggered * the breaking of sharing came in. * * - The snap mapping still points to the old block.  As it would after * the commit. * * The downside of this scheme is the timestamp magic isn't perfect, and * will continue to think that data block in the snapshot device is shared * even after the write to the origin has broken sharing.  I suspect data * blocks will typically be shared by many different devices, so we're * breaking sharing n + 1 times, rather than n, where n is the number of * devices that reference this data block.  At the moment I think the * benefits far, far outweigh the disadvantages. */ /*----------------------------------------------------------------*/ /* * Sometimes we can't deal with a bio straight away.  We put them in prison * where they can't cause any mischief.  Bios are put in a cell identified * by a key, multiple bios can be in the same cell.  When the cell is * subsequently unlocked the bios become available. */struct bio_prison; struct cell_key {	int virtual;	dm_thin_id dev;	dm_block_t block;}; struct dm_bio_prison_cell {	struct hlist_node list;	struct bio_prison *prison;	struct cell_key key;	struct bio *holder;	struct bio_list bios;}; struct bio_prison {	spinlock_t lock;	mempool_t *cell_pool; 	unsigned nr_buckets;	unsigned hash_mask;	struct hlist_head *cells;}; static uint32_t calc_nr_buckets(unsigned nr_cells){	uint32_t n = 128; 	nr_cells /= 4;	nr_cells = min(nr_cells, 8192u); 	while (n < nr_cells)		n <<= 1; 	return n;} static struct kmem_cache *_cell_cache; /* * @nr_cells should be the number of cells you want in use _concurrently_. * Don't confuse it with the number of distinct keys. */static struct bio_prison *prison_create(unsigned nr_cells){	unsigned i;	uint32_t nr_buckets = calc_nr_buckets(nr_cells);	size_t len = sizeof(struct bio_prison) +		(sizeof(struct hlist_head) * nr_buckets);	struct bio_prison *prison = kmalloc(len, GFP_KERNEL); 	if (!prison)		return NULL; 	spin_lock_init(&prison->lock);	prison->cell_pool = mempool_create_slab_pool(nr_cells, _cell_cache);	if (!prison->cell_pool) {		kfree(prison);		return NULL;	} 	prison->nr_buckets = nr_buckets;	prison->hash_mask = nr_buckets - 1;	prison->cells = (struct hlist_head *) (prison + 1);	for (i = 0; i < nr_buckets; i++)		INIT_HLIST_HEAD(prison->cells + i); 	return prison;} static void prison_destroy(struct bio_prison *prison){	mempool_destroy(prison->cell_pool);	kfree(prison);} static uint32_t hash_key(struct bio_prison *prison, struct cell_key *key){	const unsigned long BIG_PRIME = 4294967291UL;	uint64_t hash = key->block * BIG_PRIME; 	return (uint32_t) (hash & prison->hash_mask);} static int keys_equal(struct cell_key *lhs, struct cell_key *rhs){	       return (lhs->virtual == rhs->virtual) &&		       (lhs->dev == rhs->dev) &&		       (lhs->block == rhs->block);} static struct dm_bio_prison_cell *__search_bucket(struct hlist_head *bucket,						  struct cell_key *key){	struct dm_bio_prison_cell *cell;	struct hlist_node *tmp; 	hlist_for_each_entry(cell, tmp, bucket, list)		if (keys_equal(&cell->key, key))			return cell; 	return NULL;} /* * This may block if a new cell needs allocating.  You must ensure that * cells will be unlocked even if the calling thread is blocked. * * Returns 1 if the cell was already held, 0 if @inmate is the new holder. */static int bio_detain(struct bio_prison *prison, struct cell_key *key,		      struct bio *inmate, struct dm_bio_prison_cell **ref){	int r = 1;	unsigned long flags;	uint32_t hash = hash_key(prison, key);	struct dm_bio_prison_cell *cell, *cell2; 	BUG_ON(hash > prison->nr_buckets); 	spin_lock_irqsave(&prison->lock, flags); 	cell = __search_bucket(prison->cells + hash, key);	if (cell) {		bio_list_add(&cell->bios, inmate);		goto out;	} 	/*	 * Allocate a new cell	 */	spin_unlock_irqrestore(&prison->lock, flags);	cell2 = mempool_alloc(prison->cell_pool, GFP_NOIO);	spin_lock_irqsave(&prison->lock, flags); 	/*	 * We've been unlocked, so we have to double check that	 * nobody else has inserted this cell in the meantime.	 */	cell = __search_bucket(prison->cells + hash, key);	if (cell) {		mempool_free(cell2, prison->cell_pool);		bio_list_add(&cell->bios, inmate);		goto out;	} 	/*	 * Use new cell.	 */	cell = cell2; 	cell->prison = prison;	memcpy(&cell->key, key, sizeof(cell->key));	cell->holder = inmate;	bio_list_init(&cell->bios);	hlist_add_head(&cell->list, prison->cells + hash); 	r = 0; out:	spin_unlock_irqrestore(&prison->lock, flags); 	*ref = cell; 	return r;} /* * @inmates must have been initialised prior to this call */static void __cell_release(struct dm_bio_prison_cell *cell, struct bio_list *inmates){	struct bio_prison *prison = cell->prison; 	hlist_del(&cell->list); 	if (inmates) {		bio_list_add(inmates, cell->holder);		bio_list_merge(inmates, &cell->bios);	} 	mempool_free(cell, prison->cell_pool);} static void cell_release(struct dm_bio_prison_cell *cell, struct bio_list *bios){	unsigned long flags;	struct bio_prison *prison = cell->prison; 	spin_lock_irqsave(&prison->lock, flags);	__cell_release(cell, bios);	spin_unlock_irqrestore(&prison->lock, flags);} /* * There are a couple of places where we put a bio into a cell briefly * before taking it out again.  In these situations we know that no other * bio may be in the cell.  This function releases the cell, and also does * a sanity check. */static void __cell_release_singleton(struct dm_bio_prison_cell *cell, struct bio *bio){	BUG_ON(cell->holder != bio);	BUG_ON(!bio_list_empty(&cell->bios)); 	__cell_release(cell, NULL);} static void cell_release_singleton(struct dm_bio_prison_cell *cell, struct bio *bio){	unsigned long flags;	struct bio_prison *prison = cell->prison; 	spin_lock_irqsave(&prison->lock, flags);	__cell_release_singleton(cell, bio);	spin_unlock_irqrestore(&prison->lock, flags);} /* * Sometimes we don't want the holder, just the additional bios. */static void __cell_release_no_holder(struct dm_bio_prison_cell *cell,				     struct bio_list *inmates){	struct bio_prison *prison = cell->prison; 	hlist_del(&cell->list);	bio_list_merge(inmates, &cell->bios); 	mempool_free(cell, prison->cell_pool);} static void cell_release_no_holder(struct dm_bio_prison_cell *cell,				   struct bio_list *inmates){	unsigned long flags;	struct bio_prison *prison = cell->prison; 	spin_lock_irqsave(&prison->lock, flags);	__cell_release_no_holder(cell, inmates);	spin_unlock_irqrestore(&prison->lock, flags);} static void cell_error(struct dm_bio_prison_cell *cell){	struct bio_prison *prison = cell->prison;	struct bio_list bios;	struct bio *bio;	unsigned long flags; 	bio_list_init(&bios); 	spin_lock_irqsave(&prison->lock, flags);	__cell_release(cell, &bios);	spin_unlock_irqrestore(&prison->lock, flags); 	while ((bio = bio_list_pop(&bios)))		bio_io_error(bio);} /*----------------------------------------------------------------*/ /* * We use the deferred set to keep track of pending reads to shared blocks. * We do this to ensure the new mapping caused by a write isn't performed * until these prior reads have completed.  Otherwise the insertion of the * new mapping could free the old block that the read bios are mapped to. */ struct deferred_set;struct deferred_entry {	struct deferred_set *ds;	unsigned count;	struct list_head work_items;}; struct deferred_set {	spinlock_t lock;	unsigned current_entry;	unsigned sweeper;	struct deferred_entry entries[DEFERRED_SET_SIZE];}; static void ds_init(struct deferred_set *ds){	int i; 	spin_lock_init(&ds->lock);	ds->current_entry = 0;	ds->sweeper = 0;	for (i = 0; i < DEFERRED_SET_SIZE; i++) {		ds->entries[i].ds = ds;		ds->entries[i].count = 0;		INIT_LIST_HEAD(&ds->entries[i].work_items);	}} static struct deferred_entry *ds_inc(struct deferred_set *ds){	unsigned long flags;	struct deferred_entry *entry; 	spin_lock_irqsave(&ds->lock, flags);	entry = ds->entries + ds->current_entry;	entry->count++;	spin_unlock_irqrestore(&ds->lock, flags); 	return entry;} static unsigned ds_next(unsigned index){	return (index + 1) % DEFERRED_SET_SIZE;} static void __sweep(struct deferred_set *ds, struct list_head *head){	while ((ds->sweeper != ds->current_entry) &&	       !ds->entries[ds->sweeper].count) {		list_splice_init(&ds->entries[ds->sweeper].work_items, head);		ds->sweeper = ds_next(ds->sweeper);	} 	if ((ds->sweeper == ds->current_entry) && !ds->entries[ds->sweeper].count)		list_splice_init(&ds->entries[ds->sweeper].work_items, head);} static void ds_dec(struct deferred_entry *entry, struct list_head *head){	unsigned long flags; 	spin_lock_irqsave(&entry->ds->lock, flags);	BUG_ON(!entry->count);	--entry->count;	__sweep(entry->ds, head);	spin_unlock_irqrestore(&entry->ds->lock, flags);} /* * Returns 1 if deferred or 0 if no pending items to delay job. */static int ds_add_work(struct deferred_set *ds, struct list_head *work){	int r = 1;	unsigned long flags;	unsigned next_entry; 	spin_lock_irqsave(&ds->lock, flags);	if ((ds->sweeper == ds->current_entry) &&	    !ds->entries[ds->current_entry].count)		r = 0;	else {		list_add(work, &ds->entries[ds->current_entry].work_items);		next_entry = ds_next(ds->current_entry);		if (!ds->entries[next_entry].count)			ds->current_entry = next_entry;	}	spin_unlock_irqrestore(&ds->lock, flags); 	return r;} /*----------------------------------------------------------------*/ /* * Key building. */static void build_data_key(struct dm_thin_device *td,			   dm_block_t b, struct cell_key *key){	key->virtual = 0;	key->dev = dm_thin_dev_id(td);	key->block = b;} static void build_virtual_key(struct dm_thin_device *td, dm_block_t b,			      struct cell_key *key){	key->virtual = 1;	key->dev = dm_thin_dev_id(td);	key->block = b;} /*----------------------------------------------------------------*/ /* * A pool device ties together a metadata device and a data device.  It * also provides the interface for creating and destroying internal * devices. */struct dm_thin_new_mapping; struct pool_features {	unsigned zero_new_blocks:1;

Lines 1–500 of 2849.

Keyboard shortcuts

General

Show keyboard shortcuts
?
Focus the current search
/

Issues and pull requests

Open filters
f

Issues, pull requests and Inbox

Select next item
j↓
Select previous item
k↑
Open selected item
Enter
Dismiss filters or clear selection
Esc

Repository code

Go to file
t