#include <linux/module.h>
#include <linux/slab.h>
#include <linux/vmalloc.h>
+#include <linux/log2.h>
+#include <linux/dm-kcopyd.h>
#include "dm-snap.h"
#include "dm-bio-list.h"
-#include "kcopyd.h"
#define DM_MSG_PREFIX "snapshots"
#define SNAPSHOT_COPY_PRIORITY 2
/*
- * Each snapshot reserves this many pages for io
+ * Reserve 1MB for each snapshot initially (with minimum of 1 page).
*/
-#define SNAPSHOT_PAGES 256
+#define SNAPSHOT_PAGES (((1UL << 20) >> PAGE_SHIFT) ? : 1)
-struct workqueue_struct *ksnapd;
+static struct workqueue_struct *ksnapd;
static void flush_queued_bios(struct work_struct *work);
-struct pending_exception {
- struct exception e;
+struct dm_snap_pending_exception {
+ struct dm_snap_exception e;
/*
* Origin buffers waiting for this to complete are held
* group of pending_exceptions. It is always last to get freed.
* These fields get set up when writing to the origin.
*/
- struct pending_exception *primary_pe;
+ struct dm_snap_pending_exception *primary_pe;
/*
* Number of pending_exceptions processing this chunk.
* Hash table mapping origin volumes to lists of snapshots and
* a lock to protect it
*/
-static kmem_cache_t *exception_cache;
-static kmem_cache_t *pending_cache;
+static struct kmem_cache *exception_cache;
+static struct kmem_cache *pending_cache;
static mempool_t *pending_pool;
/*
kfree(_origins);
}
-static inline unsigned int origin_hash(struct block_device *bdev)
+static unsigned origin_hash(struct block_device *bdev)
{
return bdev->bd_dev & ORIGIN_MASK;
}
/*
* Implementation of the exception hash tables.
+ * The lowest hash_shift bits of the chunk number are ignored, allowing
+ * some consecutive chunks to be grouped together.
*/
-static int init_exception_table(struct exception_table *et, uint32_t size)
+static int init_exception_table(struct exception_table *et, uint32_t size,
+ unsigned hash_shift)
{
unsigned int i;
+ et->hash_shift = hash_shift;
et->hash_mask = size - 1;
et->table = dm_vcalloc(size, sizeof(struct list_head));
if (!et->table)
return 0;
}
-static void exit_exception_table(struct exception_table *et, kmem_cache_t *mem)
+static void exit_exception_table(struct exception_table *et, struct kmem_cache *mem)
{
struct list_head *slot;
- struct exception *ex, *next;
+ struct dm_snap_exception *ex, *next;
int i, size;
size = et->hash_mask + 1;
vfree(et->table);
}
-static inline uint32_t exception_hash(struct exception_table *et, chunk_t chunk)
+static uint32_t exception_hash(struct exception_table *et, chunk_t chunk)
{
- return chunk & et->hash_mask;
+ return (chunk >> et->hash_shift) & et->hash_mask;
}
-static void insert_exception(struct exception_table *eh, struct exception *e)
+static void insert_exception(struct exception_table *eh,
+ struct dm_snap_exception *e)
{
struct list_head *l = &eh->table[exception_hash(eh, e->old_chunk)];
list_add(&e->hash_list, l);
}
-static inline void remove_exception(struct exception *e)
+static void remove_exception(struct dm_snap_exception *e)
{
list_del(&e->hash_list);
}
* Return the exception data for a sector, or NULL if not
* remapped.
*/
-static struct exception *lookup_exception(struct exception_table *et,
- chunk_t chunk)
+static struct dm_snap_exception *lookup_exception(struct exception_table *et,
+ chunk_t chunk)
{
struct list_head *slot;
- struct exception *e;
+ struct dm_snap_exception *e;
slot = &et->table[exception_hash(et, chunk)];
list_for_each_entry (e, slot, hash_list)
- if (e->old_chunk == chunk)
+ if (chunk >= e->old_chunk &&
+ chunk <= e->old_chunk + dm_consecutive_chunk_count(e))
return e;
return NULL;
}
-static inline struct exception *alloc_exception(void)
+static struct dm_snap_exception *alloc_exception(void)
{
- struct exception *e;
+ struct dm_snap_exception *e;
e = kmem_cache_alloc(exception_cache, GFP_NOIO);
if (!e)
return e;
}
-static inline void free_exception(struct exception *e)
+static void free_exception(struct dm_snap_exception *e)
{
kmem_cache_free(exception_cache, e);
}
-static inline struct pending_exception *alloc_pending_exception(void)
+static struct dm_snap_pending_exception *alloc_pending_exception(void)
{
return mempool_alloc(pending_pool, GFP_NOIO);
}
-static inline void free_pending_exception(struct pending_exception *pe)
+static void free_pending_exception(struct dm_snap_pending_exception *pe)
{
mempool_free(pe, pending_pool);
}
+static void insert_completed_exception(struct dm_snapshot *s,
+ struct dm_snap_exception *new_e)
+{
+ struct exception_table *eh = &s->complete;
+ struct list_head *l;
+ struct dm_snap_exception *e = NULL;
+
+ l = &eh->table[exception_hash(eh, new_e->old_chunk)];
+
+ /* Add immediately if this table doesn't support consecutive chunks */
+ if (!eh->hash_shift)
+ goto out;
+
+ /* List is ordered by old_chunk */
+ list_for_each_entry_reverse(e, l, hash_list) {
+ /* Insert after an existing chunk? */
+ if (new_e->old_chunk == (e->old_chunk +
+ dm_consecutive_chunk_count(e) + 1) &&
+ new_e->new_chunk == (dm_chunk_number(e->new_chunk) +
+ dm_consecutive_chunk_count(e) + 1)) {
+ dm_consecutive_chunk_count_inc(e);
+ free_exception(new_e);
+ return;
+ }
+
+ /* Insert before an existing chunk? */
+ if (new_e->old_chunk == (e->old_chunk - 1) &&
+ new_e->new_chunk == (dm_chunk_number(e->new_chunk) - 1)) {
+ dm_consecutive_chunk_count_inc(e);
+ e->old_chunk--;
+ e->new_chunk--;
+ free_exception(new_e);
+ return;
+ }
+
+ if (new_e->old_chunk > e->old_chunk)
+ break;
+ }
+
+out:
+ list_add(&new_e->hash_list, e ? &e->hash_list : l);
+}
+
int dm_add_exception(struct dm_snapshot *s, chunk_t old, chunk_t new)
{
- struct exception *e;
+ struct dm_snap_exception *e;
e = alloc_exception();
if (!e)
return -ENOMEM;
e->old_chunk = old;
+
+ /* Consecutive_count is implicitly initialised to zero */
e->new_chunk = new;
- insert_exception(&s->complete, e);
+
+ insert_completed_exception(s, e);
+
return 0;
}
return mem;
}
-/*
- * Rounds a number down to a power of 2.
- */
-static inline uint32_t round_down(uint32_t n)
-{
- while (n & (n - 1))
- n &= (n - 1);
- return n;
-}
-
/*
* Allocate room for a suitable hash table.
*/
hash_size = min(origin_dev_size, cow_dev_size) >> s->chunk_shift;
hash_size = min(hash_size, max_buckets);
- /* Round it down to a power of 2 */
- hash_size = round_down(hash_size);
- if (init_exception_table(&s->complete, hash_size))
+ hash_size = rounddown_pow_of_two(hash_size);
+ if (init_exception_table(&s->complete, hash_size,
+ DM_CHUNK_CONSECUTIVE_BITS))
return -ENOMEM;
/*
if (hash_size < 64)
hash_size = 64;
- if (init_exception_table(&s->pending, hash_size)) {
+ if (init_exception_table(&s->pending, hash_size, 0)) {
exit_exception_table(&s->complete, exception_cache);
return -ENOMEM;
}
* Round a number up to the nearest 'size' boundary. size must
* be a power of 2.
*/
-static inline ulong round_up(ulong n, ulong size)
+static ulong round_up(ulong n, ulong size)
{
size--;
return (n + size) & ~size;
chunk_size = round_up(chunk_size, PAGE_SIZE >> 9);
/* Check chunk_size is a power of 2 */
- if (chunk_size & (chunk_size - 1)) {
+ if (!is_power_of_2(chunk_size)) {
*error = "Chunk size is not a power of 2";
return -EINVAL;
}
s->last_percent = 0;
init_rwsem(&s->lock);
spin_lock_init(&s->pe_lock);
- s->table = ti->table;
+ s->ti = ti;
/* Allocate hash table for COW data */
if (init_hash_tables(s)) {
goto bad4;
}
- r = kcopyd_client_create(SNAPSHOT_PAGES, &s->kcopyd_client);
+ r = dm_kcopyd_client_create(SNAPSHOT_PAGES, &s->kcopyd_client);
if (r) {
ti->error = "Could not create kcopyd client";
goto bad5;
/* Metadata must only be loaded into one table at once */
r = s->store.read_metadata(&s->store);
- if (r) {
+ if (r < 0) {
ti->error = "Failed to read snapshot metadata";
goto bad6;
+ } else if (r > 0) {
+ s->valid = 0;
+ DMWARN("Snapshot is marked invalid.");
}
bio_list_init(&s->queued_bios);
return 0;
bad6:
- kcopyd_client_destroy(s->kcopyd_client);
+ dm_kcopyd_client_destroy(s->kcopyd_client);
bad5:
s->store.destroy(&s->store);
return r;
}
+static void __free_exceptions(struct dm_snapshot *s)
+{
+ dm_kcopyd_client_destroy(s->kcopyd_client);
+ s->kcopyd_client = NULL;
+
+ exit_exception_table(&s->pending, pending_cache);
+ exit_exception_table(&s->complete, exception_cache);
+
+ s->store.destroy(&s->store);
+}
+
static void snapshot_dtr(struct dm_target *ti)
{
- struct dm_snapshot *s = (struct dm_snapshot *) ti->private;
+ struct dm_snapshot *s = ti->private;
flush_workqueue(ksnapd);
/* After this returns there can be no new kcopyd jobs. */
unregister_snapshot(s);
- kcopyd_client_destroy(s->kcopyd_client);
-
- exit_exception_table(&s->pending, pending_cache);
- exit_exception_table(&s->complete, exception_cache);
-
- /* Deallocate memory used */
- s->store.destroy(&s->store);
+ __free_exceptions(s);
dm_put_device(ti, s->origin);
dm_put_device(ti, s->cow);
while (bio) {
n = bio->bi_next;
bio->bi_next = NULL;
- bio_io_error(bio, bio->bi_size);
+ bio_io_error(bio);
bio = n;
}
}
s->valid = 0;
- dm_table_event(s->table);
+ dm_table_event(s->ti->table);
}
-static void get_pending_exception(struct pending_exception *pe)
+static void get_pending_exception(struct dm_snap_pending_exception *pe)
{
atomic_inc(&pe->ref_count);
}
-static struct bio *put_pending_exception(struct pending_exception *pe)
+static struct bio *put_pending_exception(struct dm_snap_pending_exception *pe)
{
- struct pending_exception *primary_pe;
+ struct dm_snap_pending_exception *primary_pe;
struct bio *origin_bios = NULL;
primary_pe = pe->primary_pe;
return origin_bios;
}
-static void pending_complete(struct pending_exception *pe, int success)
+static void pending_complete(struct dm_snap_pending_exception *pe, int success)
{
- struct exception *e;
+ struct dm_snap_exception *e;
struct dm_snapshot *s = pe->snap;
struct bio *origin_bios = NULL;
struct bio *snapshot_bios = NULL;
* Add a proper exception, and remove the
* in-flight exception from the list.
*/
- insert_exception(&s->complete, e);
+ insert_completed_exception(s, e);
out:
remove_exception(&pe->e);
static void commit_callback(void *context, int success)
{
- struct pending_exception *pe = (struct pending_exception *) context;
+ struct dm_snap_pending_exception *pe = context;
+
pending_complete(pe, success);
}
* Called when the copy I/O has finished. kcopyd actually runs
* this code so don't block.
*/
-static void copy_callback(int read_err, unsigned int write_err, void *context)
+static void copy_callback(int read_err, unsigned long write_err, void *context)
{
- struct pending_exception *pe = (struct pending_exception *) context;
+ struct dm_snap_pending_exception *pe = context;
struct dm_snapshot *s = pe->snap;
if (read_err || write_err)
/*
* Dispatches the copy operation to kcopyd.
*/
-static void start_copy(struct pending_exception *pe)
+static void start_copy(struct dm_snap_pending_exception *pe)
{
struct dm_snapshot *s = pe->snap;
- struct io_region src, dest;
+ struct dm_io_region src, dest;
struct block_device *bdev = s->origin->bdev;
sector_t dev_size;
dest.count = src.count;
/* Hand over to kcopyd */
- kcopyd_copy(s->kcopyd_client,
+ dm_kcopyd_copy(s->kcopyd_client,
&src, 1, &dest, 0, copy_callback, pe);
}
* NOTE: a write lock must be held on snap->lock before calling
* this.
*/
-static struct pending_exception *
+static struct dm_snap_pending_exception *
__find_pending_exception(struct dm_snapshot *s, struct bio *bio)
{
- struct exception *e;
- struct pending_exception *pe;
+ struct dm_snap_exception *e;
+ struct dm_snap_pending_exception *pe;
chunk_t chunk = sector_to_chunk(s, bio->bi_sector);
/*
e = lookup_exception(&s->pending, chunk);
if (e) {
/* cast the exception to a pending exception */
- pe = container_of(e, struct pending_exception, e);
+ pe = container_of(e, struct dm_snap_pending_exception, e);
goto out;
}
e = lookup_exception(&s->pending, chunk);
if (e) {
free_pending_exception(pe);
- pe = container_of(e, struct pending_exception, e);
+ pe = container_of(e, struct dm_snap_pending_exception, e);
goto out;
}
return pe;
}
-static inline void remap_exception(struct dm_snapshot *s, struct exception *e,
- struct bio *bio)
+static void remap_exception(struct dm_snapshot *s, struct dm_snap_exception *e,
+ struct bio *bio, chunk_t chunk)
{
bio->bi_bdev = s->cow->bdev;
- bio->bi_sector = chunk_to_sector(s, e->new_chunk) +
- (bio->bi_sector & s->chunk_mask);
+ bio->bi_sector = chunk_to_sector(s, dm_chunk_number(e->new_chunk) +
+ (chunk - e->old_chunk)) +
+ (bio->bi_sector & s->chunk_mask);
}
static int snapshot_map(struct dm_target *ti, struct bio *bio,
union map_info *map_context)
{
- struct exception *e;
- struct dm_snapshot *s = (struct dm_snapshot *) ti->private;
- int r = 1;
+ struct dm_snap_exception *e;
+ struct dm_snapshot *s = ti->private;
+ int r = DM_MAPIO_REMAPPED;
chunk_t chunk;
- struct pending_exception *pe = NULL;
+ struct dm_snap_pending_exception *pe = NULL;
chunk = sector_to_chunk(s, bio->bi_sector);
if (!s->valid)
return -EIO;
- if (unlikely(bio_barrier(bio)))
- return -EOPNOTSUPP;
-
/* FIXME: should only take write lock if we need
* to copy an exception */
down_write(&s->lock);
/* If the block is already remapped - use that, else remap it */
e = lookup_exception(&s->complete, chunk);
if (e) {
- remap_exception(s, e, bio);
+ remap_exception(s, e, bio, chunk);
goto out_unlock;
}
goto out_unlock;
}
- remap_exception(s, &pe->e, bio);
+ remap_exception(s, &pe->e, bio, chunk);
bio_list_add(&pe->snapshot_bios, bio);
- r = 0;
+ r = DM_MAPIO_SUBMITTED;
if (!pe->started) {
/* this is protected by snap->lock */
static void snapshot_resume(struct dm_target *ti)
{
- struct dm_snapshot *s = (struct dm_snapshot *) ti->private;
+ struct dm_snapshot *s = ti->private;
down_write(&s->lock);
s->active = 1;
static int snapshot_status(struct dm_target *ti, status_type_t type,
char *result, unsigned int maxlen)
{
- struct dm_snapshot *snap = (struct dm_snapshot *) ti->private;
+ struct dm_snapshot *snap = ti->private;
switch (type) {
case STATUSTYPE_INFO:
*---------------------------------------------------------------*/
static int __origin_write(struct list_head *snapshots, struct bio *bio)
{
- int r = 1, first = 0;
+ int r = DM_MAPIO_REMAPPED, first = 0;
struct dm_snapshot *snap;
- struct exception *e;
- struct pending_exception *pe, *next_pe, *primary_pe = NULL;
+ struct dm_snap_exception *e;
+ struct dm_snap_pending_exception *pe, *next_pe, *primary_pe = NULL;
chunk_t chunk;
LIST_HEAD(pe_queue);
goto next_snapshot;
/* Nothing to do if writing beyond end of snapshot */
- if (bio->bi_sector >= dm_table_get_size(snap->table))
+ if (bio->bi_sector >= dm_table_get_size(snap->ti->table))
goto next_snapshot;
/*
bio_list_add(&primary_pe->origin_bios, bio);
- r = 0;
+ r = DM_MAPIO_SUBMITTED;
}
if (!pe->primary_pe) {
static int do_origin(struct dm_dev *origin, struct bio *bio)
{
struct origin *o;
- int r = 1;
+ int r = DM_MAPIO_REMAPPED;
down_read(&_origins_lock);
o = __lookup_origin(origin->bdev);
static void origin_dtr(struct dm_target *ti)
{
- struct dm_dev *dev = (struct dm_dev *) ti->private;
+ struct dm_dev *dev = ti->private;
dm_put_device(ti, dev);
}
static int origin_map(struct dm_target *ti, struct bio *bio,
union map_info *map_context)
{
- struct dm_dev *dev = (struct dm_dev *) ti->private;
+ struct dm_dev *dev = ti->private;
bio->bi_bdev = dev->bdev;
- if (unlikely(bio_barrier(bio)))
- return -EOPNOTSUPP;
-
/* Only tell snapshots if this is a write */
- return (bio_rw(bio) == WRITE) ? do_origin(dev, bio) : 1;
+ return (bio_rw(bio) == WRITE) ? do_origin(dev, bio) : DM_MAPIO_REMAPPED;
}
#define min_not_zero(l, r) (l == 0) ? r : ((r == 0) ? l : min(l, r))
*/
static void origin_resume(struct dm_target *ti)
{
- struct dm_dev *dev = (struct dm_dev *) ti->private;
+ struct dm_dev *dev = ti->private;
struct dm_snapshot *snap;
struct origin *o;
chunk_t chunk_size = 0;
static int origin_status(struct dm_target *ti, status_type_t type, char *result,
unsigned int maxlen)
{
- struct dm_dev *dev = (struct dm_dev *) ti->private;
+ struct dm_dev *dev = ti->private;
switch (type) {
case STATUSTYPE_INFO:
static struct target_type origin_target = {
.name = "snapshot-origin",
- .version = {1, 5, 0},
+ .version = {1, 6, 0},
.module = THIS_MODULE,
.ctr = origin_ctr,
.dtr = origin_dtr,
static struct target_type snapshot_target = {
.name = "snapshot",
- .version = {1, 5, 0},
+ .version = {1, 6, 0},
.module = THIS_MODULE,
.ctr = snapshot_ctr,
.dtr = snapshot_dtr,
goto bad2;
}
- exception_cache = kmem_cache_create("dm-snapshot-ex",
- sizeof(struct exception),
- __alignof__(struct exception),
- 0, NULL, NULL);
+ exception_cache = KMEM_CACHE(dm_snap_exception, 0);
if (!exception_cache) {
DMERR("Couldn't create exception cache.");
r = -ENOMEM;
goto bad3;
}
- pending_cache =
- kmem_cache_create("dm-snapshot-in",
- sizeof(struct pending_exception),
- __alignof__(struct pending_exception),
- 0, NULL, NULL);
+ pending_cache = KMEM_CACHE(dm_snap_pending_exception, 0);
if (!pending_cache) {
DMERR("Couldn't create pending cache.");
r = -ENOMEM;
projects / linux-2.6-omap-h63xx.git / blobdiff