#include <linux/vmalloc.h>
#include <linux/workqueue.h>
+#define DM_MSG_PREFIX "raid1"
+
static struct workqueue_struct *_kmirrord_wq;
static struct work_struct _kmirrord_work;
+static DECLARE_WAIT_QUEUE_HEAD(_kmirrord_recovery_stopped);
static inline void wake(void)
{
struct list_head *buckets;
spinlock_t region_lock;
+ atomic_t recovery_in_flight;
struct semaphore recovery_count;
struct list_head clean_regions;
struct list_head quiesced_regions;
spin_lock_init(&rh->region_lock);
sema_init(&rh->recovery_count, 0);
+ atomic_set(&rh->recovery_in_flight, 0);
INIT_LIST_HEAD(&rh->clean_regions);
INIT_LIST_HEAD(&rh->quiesced_regions);
INIT_LIST_HEAD(&rh->recovered_regions);
struct region *reg, *nreg;
read_unlock(&rh->hash_lock);
- nreg = mempool_alloc(rh->region_pool, GFP_NOIO);
+ nreg = mempool_alloc(rh->region_pool, GFP_ATOMIC);
+ if (unlikely(!nreg))
+ nreg = kmalloc(sizeof(struct region), GFP_NOIO);
nreg->state = rh->log->type->in_sync(rh->log, region, 1) ?
RH_CLEAN : RH_NOSYNC;
nreg->rh = rh;
}
}
+static void complete_resync_work(struct region *reg, int success)
+{
+ struct region_hash *rh = reg->rh;
+
+ rh->log->type->set_region_sync(rh->log, reg->key, success);
+ dispatch_bios(rh->ms, ®->delayed_bios);
+ if (atomic_dec_and_test(&rh->recovery_in_flight))
+ wake_up_all(&_kmirrord_recovery_stopped);
+ up(&rh->recovery_count);
+}
+
static void rh_update_states(struct region_hash *rh)
{
struct region *reg, *next;
*/
list_for_each_entry_safe (reg, next, &recovered, list) {
rh->log->type->clear_region(rh->log, reg->key);
- rh->log->type->complete_resync_work(rh->log, reg->key, 1);
- dispatch_bios(rh->ms, ®->delayed_bios);
- up(&rh->recovery_count);
+ complete_resync_work(reg, 1);
mempool_free(reg, rh->region_pool);
}
static void rh_recovery_prepare(struct region_hash *rh)
{
- while (!down_trylock(&rh->recovery_count))
+ /* Extra reference to avoid race with rh_stop_recovery */
+ atomic_inc(&rh->recovery_in_flight);
+
+ while (!down_trylock(&rh->recovery_count)) {
+ atomic_inc(&rh->recovery_in_flight);
if (__rh_recovery_prepare(rh) <= 0) {
+ atomic_dec(&rh->recovery_in_flight);
up(&rh->recovery_count);
break;
}
+ }
+
+ /* Drop the extra reference */
+ if (atomic_dec_and_test(&rh->recovery_in_flight))
+ wake_up_all(&_kmirrord_recovery_stopped);
}
/*
do_writes(ms, &writes);
}
-static void do_work(void *ignored)
+static void do_work(struct work_struct *ignored)
{
struct mirror_set *ms;
ms = kmalloc(len, GFP_KERNEL);
if (!ms) {
- ti->error = "dm-mirror: Cannot allocate mirror context";
+ ti->error = "Cannot allocate mirror context";
return NULL;
}
ms->default_mirror = &ms->mirror[DEFAULT_MIRROR];
if (rh_init(&ms->rh, ms, dl, region_size, ms->nr_regions)) {
- ti->error = "dm-mirror: Error creating dirty region hash";
+ ti->error = "Error creating dirty region hash";
kfree(ms);
return NULL;
}
unsigned long long offset;
if (sscanf(argv[1], "%llu", &offset) != 1) {
- ti->error = "dm-mirror: Invalid offset";
+ ti->error = "Invalid offset";
return -EINVAL;
}
if (dm_get_device(ti, argv[0], offset, ti->len,
dm_table_get_mode(ti->table),
&ms->mirror[mirror].dev)) {
- ti->error = "dm-mirror: Device lookup failure";
+ ti->error = "Device lookup failure";
return -ENXIO;
}
struct dirty_log *dl;
if (argc < 2) {
- ti->error = "dm-mirror: Insufficient mirror log arguments";
+ ti->error = "Insufficient mirror log arguments";
return NULL;
}
if (sscanf(argv[1], "%u", ¶m_count) != 1) {
- ti->error = "dm-mirror: Invalid mirror log argument count";
+ ti->error = "Invalid mirror log argument count";
return NULL;
}
*args_used = 2 + param_count;
if (argc < *args_used) {
- ti->error = "dm-mirror: Insufficient mirror log arguments";
+ ti->error = "Insufficient mirror log arguments";
return NULL;
}
dl = dm_create_dirty_log(argv[0], ti, param_count, argv + 2);
if (!dl) {
- ti->error = "dm-mirror: Error creating mirror dirty log";
+ ti->error = "Error creating mirror dirty log";
return NULL;
}
if (!_check_region_size(ti, dl->type->get_region_size(dl))) {
- ti->error = "dm-mirror: Invalid region size";
+ ti->error = "Invalid region size";
dm_destroy_dirty_log(dl);
return NULL;
}
if (!argc || sscanf(argv[0], "%u", &nr_mirrors) != 1 ||
nr_mirrors < 2 || nr_mirrors > KCOPYD_MAX_REGIONS + 1) {
- ti->error = "dm-mirror: Invalid number of mirrors";
+ ti->error = "Invalid number of mirrors";
dm_destroy_dirty_log(dl);
return -EINVAL;
}
argv++, argc--;
if (argc != nr_mirrors * 2) {
- ti->error = "dm-mirror: Wrong number of mirror arguments";
+ ti->error = "Wrong number of mirror arguments";
dm_destroy_dirty_log(dl);
return -EINVAL;
}
if (rw == WRITE) {
queue_bio(ms, bio, rw);
- return 0;
+ return DM_MAPIO_SUBMITTED;
}
r = ms->rh.log->type->in_sync(ms->rh.log,
return r;
if (r == -EWOULDBLOCK) /* FIXME: ugly */
- r = 0;
+ r = DM_MAPIO_SUBMITTED;
/*
* We don't want to fast track a recovery just for a read
if (!r) {
/* Pass this io over to the daemon */
queue_bio(ms, bio, rw);
- return 0;
+ return DM_MAPIO_SUBMITTED;
}
m = choose_mirror(ms, bio->bi_sector);
return -EIO;
map_bio(ms, m, bio);
- return 1;
+ return DM_MAPIO_REMAPPED;
}
static int mirror_end_io(struct dm_target *ti, struct bio *bio,
struct dirty_log *log = ms->rh.log;
rh_stop_recovery(&ms->rh);
+
+ /* Wait for all I/O we generated to complete */
+ wait_event(_kmirrord_recovery_stopped,
+ !atomic_read(&ms->rh.recovery_in_flight));
+
if (log->type->suspend && log->type->suspend(log))
/* FIXME: need better error handling */
DMWARN("log suspend failed");
break;
case STATUSTYPE_TABLE:
- DMEMIT("%d ", ms->nr_mirrors);
+ DMEMIT("%d", ms->nr_mirrors);
for (m = 0; m < ms->nr_mirrors; m++)
- DMEMIT("%s %llu ", ms->mirror[m].dev->name,
+ DMEMIT(" %s %llu", ms->mirror[m].dev->name,
(unsigned long long)ms->mirror[m].offset);
}
dm_dirty_log_exit();
return r;
}
- INIT_WORK(&_kmirrord_work, do_work, NULL);
+ INIT_WORK(&_kmirrord_work, do_work);
r = dm_register_target(&mirror_target);
if (r < 0) {
git.openpandora.org / pandora-kernel.git / blobdiff