Commit e49e5829 authored by Joe Thornber's avatar Joe Thornber Committed by Alasdair G Kergon
Browse files

dm thin: add read only and fail io modes



Add read-only and fail-io modes to thin provisioning.

If a transaction commit fails the pool's metadata device will transition
to "read-only" mode.  If a commit fails once already in read-only mode
the transition to "fail-io" mode occurs.

Once in fail-io mode the pool and all associated thin devices will
report a status of "Fail".
Signed-off-by: default avatarJoe Thornber <ejt@redhat.com>
Signed-off-by: default avatarMike Snitzer <snitzer@redhat.com>
Signed-off-by: default avatarAlasdair G Kergon <agk@redhat.com>
parent da105ed5
......@@ -231,6 +231,9 @@ i) Constructor
no_discard_passdown: Don't pass discards down to the underlying
data device, but just remove the mapping.
read_only: Don't allow any changes to be made to the pool
metadata.
Data block size must be between 64KB (128 sectors) and 1GB
(2097152 sectors) inclusive.
......@@ -239,7 +242,7 @@ ii) Status
<transaction id> <used metadata blocks>/<total metadata blocks>
<used data blocks>/<total data blocks> <held metadata root>
[no_]discard_passdown ro|rw
transaction id:
A 64-bit number used by userspace to help synchronise with metadata
......@@ -257,6 +260,21 @@ ii) Status
held root. This feature is not yet implemented so '-' is
always returned.
discard_passdown|no_discard_passdown
Whether or not discards are actually being passed down to the
underlying device. When this is enabled when loading the table,
it can get disabled if the underlying device doesn't support it.
ro|rw
If the pool encounters certain types of device failures it will
drop into a read-only metadata mode in which no changes to
the pool metadata (like allocating new blocks) are permitted.
In serious cases where even a read-only mode is deemed unsafe
no further I/O will be permitted and the status will just
contain the string 'Fail'. The userspace recovery tools
should then be used.
iii) Messages
create_thin <dev id>
......@@ -329,3 +347,7 @@ regain some space then send the 'trim' message to the pool.
ii) Status
<nr mapped sectors> <highest mapped sector>
If the pool has encountered device errors and failed, the status
will just contain the string 'Fail'. The userspace recovery
tools should then be used.
/*
* Copyright (C) 2011 Red Hat UK.
* Copyright (C) 2011-2012 Red Hat UK.
*
* This file is released under the GPL.
*/
......@@ -496,12 +496,27 @@ static void build_virtual_key(struct dm_thin_device *td, dm_block_t b,
*/
struct dm_thin_new_mapping;
/*
* The pool runs in 3 modes. Ordered in degraded order for comparisons.
*/
enum pool_mode {
PM_WRITE, /* metadata may be changed */
PM_READ_ONLY, /* metadata may not be changed */
PM_FAIL, /* all I/O fails */
};
struct pool_features {
enum pool_mode mode;
unsigned zero_new_blocks:1;
unsigned discard_enabled:1;
unsigned discard_passdown:1;
};
struct thin_c;
typedef void (*process_bio_fn)(struct thin_c *tc, struct bio *bio);
typedef void (*process_mapping_fn)(struct dm_thin_new_mapping *m);
struct pool {
struct list_head list;
struct dm_target *ti; /* Only set if a pool target is bound */
......@@ -542,8 +557,17 @@ struct pool {
struct dm_thin_new_mapping *next_mapping;
mempool_t *mapping_pool;
mempool_t *endio_hook_pool;
process_bio_fn process_bio;
process_bio_fn process_discard;
process_mapping_fn process_prepared_mapping;
process_mapping_fn process_prepared_discard;
};
static enum pool_mode get_pool_mode(struct pool *pool);
static void set_pool_mode(struct pool *pool, enum pool_mode mode);
/*
* Target context for a pool.
*/
......@@ -718,16 +742,28 @@ static void issue(struct thin_c *tc, struct bio *bio)
struct pool *pool = tc->pool;
unsigned long flags;
if (!bio_triggers_commit(tc, bio)) {
generic_make_request(bio);
return;
}
/*
* Batch together any FUA/FLUSH bios we find and then issue
* a single commit for them in process_deferred_bios().
* Complete bio with an error if earlier I/O caused changes to
* the metadata that can't be committed e.g, due to I/O errors
* on the metadata device.
*/
if (bio_triggers_commit(tc, bio)) {
spin_lock_irqsave(&pool->lock, flags);
bio_list_add(&pool->deferred_flush_bios, bio);
spin_unlock_irqrestore(&pool->lock, flags);
} else
generic_make_request(bio);
if (dm_thin_aborted_changes(tc->td)) {
bio_io_error(bio);
return;
}
/*
* Batch together any bios that trigger commits and then issue a
* single commit for them in process_deferred_bios().
*/
spin_lock_irqsave(&pool->lock, flags);
bio_list_add(&pool->deferred_flush_bios, bio);
spin_unlock_irqrestore(&pool->lock, flags);
}
static void remap_to_origin_and_issue(struct thin_c *tc, struct bio *bio)
......@@ -864,6 +900,14 @@ static void cell_defer_except(struct thin_c *tc, struct dm_bio_prison_cell *cell
wake_worker(pool);
}
static void process_prepared_mapping_fail(struct dm_thin_new_mapping *m)
{
if (m->bio)
m->bio->bi_end_io = m->saved_bi_end_io;
cell_error(m->cell);
list_del(&m->list);
mempool_free(m, m->tc->pool->mapping_pool);
}
static void process_prepared_mapping(struct dm_thin_new_mapping *m)
{
struct thin_c *tc = m->tc;
......@@ -908,18 +952,20 @@ static void process_prepared_mapping(struct dm_thin_new_mapping *m)
mempool_free(m, tc->pool->mapping_pool);
}
static void process_prepared_discard(struct dm_thin_new_mapping *m)
static void process_prepared_discard_fail(struct dm_thin_new_mapping *m)
{
int r;
struct thin_c *tc = m->tc;
r = dm_thin_remove_block(tc->td, m->virt_block);
if (r)
DMERR("dm_thin_remove_block() failed");
bio_io_error(m->bio);
cell_defer_except(tc, m->cell);
cell_defer_except(tc, m->cell2);
mempool_free(m, tc->pool->mapping_pool);
}
static void process_prepared_discard_passdown(struct dm_thin_new_mapping *m)
{
struct thin_c *tc = m->tc;
/*
* Pass the discard down to the underlying device?
*/
if (m->pass_discard)
remap_and_issue(tc, m->bio, m->data_block);
else
......@@ -930,8 +976,20 @@ static void process_prepared_discard(struct dm_thin_new_mapping *m)
mempool_free(m, tc->pool->mapping_pool);
}
static void process_prepared_discard(struct dm_thin_new_mapping *m)
{
int r;
struct thin_c *tc = m->tc;
r = dm_thin_remove_block(tc->td, m->virt_block);
if (r)
DMERR("dm_thin_remove_block() failed");
process_prepared_discard_passdown(m);
}
static void process_prepared(struct pool *pool, struct list_head *head,
void (*fn)(struct dm_thin_new_mapping *))
process_mapping_fn *fn)
{
unsigned long flags;
struct list_head maps;
......@@ -943,7 +1001,7 @@ static void process_prepared(struct pool *pool, struct list_head *head,
spin_unlock_irqrestore(&pool->lock, flags);
list_for_each_entry_safe(m, tmp, &maps, list)
fn(m);
(*fn)(m);
}
/*
......@@ -1109,6 +1167,35 @@ static void schedule_zero(struct thin_c *tc, dm_block_t virt_block,
}
}
static int commit(struct pool *pool)
{
int r;
r = dm_pool_commit_metadata(pool->pmd);
if (r)
DMERR("commit failed, error = %d", r);
return r;
}
/*
* A non-zero return indicates read_only or fail_io mode.
* Many callers don't care about the return value.
*/
static int commit_or_fallback(struct pool *pool)
{
int r;
if (get_pool_mode(pool) != PM_WRITE)
return -EINVAL;
r = commit(pool);
if (r)
set_pool_mode(pool, PM_READ_ONLY);
return r;
}
static int alloc_data_block(struct thin_c *tc, dm_block_t *result)
{
int r;
......@@ -1137,12 +1224,7 @@ static int alloc_data_block(struct thin_c *tc, dm_block_t *result)
* Try to commit to see if that will free up some
* more space.
*/
r = dm_pool_commit_metadata(pool->pmd);
if (r) {
DMERR("%s: dm_pool_commit_metadata() failed, error = %d",
__func__, r);
return r;
}
(void) commit_or_fallback(pool);
r = dm_pool_get_free_block_count(pool->pmd, &free_blocks);
if (r)
......@@ -1373,6 +1455,7 @@ static void provision_block(struct thin_c *tc, struct bio *bio, dm_block_t block
default:
DMERR("%s: alloc_data_block() failed, error = %d", __func__, r);
set_pool_mode(tc->pool, PM_READ_ONLY);
cell_error(cell);
break;
}
......@@ -1430,6 +1513,49 @@ static void process_bio(struct thin_c *tc, struct bio *bio)
}
}
static void process_bio_read_only(struct thin_c *tc, struct bio *bio)
{
int r;
int rw = bio_data_dir(bio);
dm_block_t block = get_bio_block(tc, bio);
struct dm_thin_lookup_result lookup_result;
r = dm_thin_find_block(tc->td, block, 1, &lookup_result);
switch (r) {
case 0:
if (lookup_result.shared && (rw == WRITE) && bio->bi_size)
bio_io_error(bio);
else
remap_and_issue(tc, bio, lookup_result.block);
break;
case -ENODATA:
if (rw != READ) {
bio_io_error(bio);
break;
}
if (tc->origin_dev) {
remap_to_origin_and_issue(tc, bio);
break;
}
zero_fill_bio(bio);
bio_endio(bio, 0);
break;
default:
DMERR("dm_thin_find_block() failed, error = %d", r);
bio_io_error(bio);
break;
}
}
static void process_bio_fail(struct thin_c *tc, struct bio *bio)
{
bio_io_error(bio);
}
static int need_commit_due_to_time(struct pool *pool)
{
return jiffies < pool->last_commit_jiffies ||
......@@ -1441,7 +1567,6 @@ static void process_deferred_bios(struct pool *pool)
unsigned long flags;
struct bio *bio;
struct bio_list bios;
int r;
bio_list_init(&bios);
......@@ -1468,9 +1593,9 @@ static void process_deferred_bios(struct pool *pool)
}
if (bio->bi_rw & REQ_DISCARD)
process_discard(tc, bio);
pool->process_discard(tc, bio);
else
process_bio(tc, bio);
pool->process_bio(tc, bio);
}
/*
......@@ -1486,10 +1611,7 @@ static void process_deferred_bios(struct pool *pool)
if (bio_list_empty(&bios) && !need_commit_due_to_time(pool))
return;
r = dm_pool_commit_metadata(pool->pmd);
if (r) {
DMERR("%s: dm_pool_commit_metadata() failed, error = %d",
__func__, r);
if (commit_or_fallback(pool)) {
while ((bio = bio_list_pop(&bios)))
bio_io_error(bio);
return;
......@@ -1504,8 +1626,8 @@ static void do_worker(struct work_struct *ws)
{
struct pool *pool = container_of(ws, struct pool, worker);
process_prepared(pool, &pool->prepared_mappings, process_prepared_mapping);
process_prepared(pool, &pool->prepared_discards, process_prepared_discard);
process_prepared(pool, &pool->prepared_mappings, &pool->process_prepared_mapping);
process_prepared(pool, &pool->prepared_discards, &pool->process_prepared_discard);
process_deferred_bios(pool);
}
......@@ -1522,6 +1644,52 @@ static void do_waker(struct work_struct *ws)
/*----------------------------------------------------------------*/
static enum pool_mode get_pool_mode(struct pool *pool)
{
return pool->pf.mode;
}
static void set_pool_mode(struct pool *pool, enum pool_mode mode)
{
int r;
pool->pf.mode = mode;
switch (mode) {
case PM_FAIL:
DMERR("switching pool to failure mode");
pool->process_bio = process_bio_fail;
pool->process_discard = process_bio_fail;
pool->process_prepared_mapping = process_prepared_mapping_fail;
pool->process_prepared_discard = process_prepared_discard_fail;
break;
case PM_READ_ONLY:
DMERR("switching pool to read-only mode");
r = dm_pool_abort_metadata(pool->pmd);
if (r) {
DMERR("aborting transaction failed");
set_pool_mode(pool, PM_FAIL);
} else {
dm_pool_metadata_read_only(pool->pmd);
pool->process_bio = process_bio_read_only;
pool->process_discard = process_discard;
pool->process_prepared_mapping = process_prepared_mapping_fail;
pool->process_prepared_discard = process_prepared_discard_passdown;
}
break;
case PM_WRITE:
pool->process_bio = process_bio;
pool->process_discard = process_discard;
pool->process_prepared_mapping = process_prepared_mapping;
pool->process_prepared_discard = process_prepared_discard;
break;
}
}
/*----------------------------------------------------------------*/
/*
* Mapping functions.
*/
......@@ -1567,6 +1735,12 @@ static int thin_bio_map(struct dm_target *ti, struct bio *bio,
struct dm_thin_lookup_result result;
map_context->ptr = thin_hook_bio(tc, bio);
if (get_pool_mode(tc->pool) == PM_FAIL) {
bio_io_error(bio);
return DM_MAPIO_SUBMITTED;
}
if (bio->bi_rw & (REQ_DISCARD | REQ_FLUSH | REQ_FUA)) {
thin_defer_bio(tc, bio);
return DM_MAPIO_SUBMITTED;
......@@ -1603,14 +1777,35 @@ static int thin_bio_map(struct dm_target *ti, struct bio *bio,
break;
case -ENODATA:
if (get_pool_mode(tc->pool) == PM_READ_ONLY) {
/*
* This block isn't provisioned, and we have no way
* of doing so. Just error it.
*/
bio_io_error(bio);
r = DM_MAPIO_SUBMITTED;
break;
}
/* fall through */
case -EWOULDBLOCK:
/*
* In future, the failed dm_thin_find_block above could
* provide the hint to load the metadata into cache.
*/
case -EWOULDBLOCK:
thin_defer_bio(tc, bio);
r = DM_MAPIO_SUBMITTED;
break;
default:
/*
* Must always call bio_io_error on failure.
* dm_thin_find_block can fail with -EINVAL if the
* pool is switched to fail-io mode.
*/
bio_io_error(bio);
r = DM_MAPIO_SUBMITTED;
break;
}
return r;
......@@ -1647,15 +1842,26 @@ static int bind_control_target(struct pool *pool, struct dm_target *ti)
{
struct pool_c *pt = ti->private;
/*
* We want to make sure that degraded pools are never upgraded.
*/
enum pool_mode old_mode = pool->pf.mode;
enum pool_mode new_mode = pt->pf.mode;
if (old_mode > new_mode)
new_mode = old_mode;
pool->ti = ti;
pool->low_water_blocks = pt->low_water_blocks;
pool->pf = pt->pf;
set_pool_mode(pool, new_mode);
/*
* If discard_passdown was enabled verify that the data device
* supports discards. Disable discard_passdown if not; otherwise
* -EOPNOTSUPP will be returned.
*/
/* FIXME: pull this out into a sep fn. */
if (pt->pf.discard_passdown) {
struct request_queue *q = bdev_get_queue(pt->data_dev->bdev);
if (!q || !blk_queue_discard(q)) {
......@@ -1681,6 +1887,7 @@ static void unbind_control_target(struct pool *pool, struct dm_target *ti)
/* Initialize pool features. */
static void pool_features_init(struct pool_features *pf)
{
pf->mode = PM_WRITE;
pf->zero_new_blocks = 1;
pf->discard_enabled = 1;
pf->discard_passdown = 1;
......@@ -1711,14 +1918,16 @@ static struct kmem_cache *_endio_hook_cache;
static struct pool *pool_create(struct mapped_device *pool_md,
struct block_device *metadata_dev,
unsigned long block_size, char **error)
unsigned long block_size,
int read_only, char **error)
{
int r;
void *err_p;
struct pool *pool;
struct dm_pool_metadata *pmd;
bool format_device = read_only ? false : true;
pmd = dm_pool_metadata_open(metadata_dev, block_size, true);
pmd = dm_pool_metadata_open(metadata_dev, block_size, format_device);
if (IS_ERR(pmd)) {
*error = "Error creating metadata object";
return (struct pool *)pmd;
......@@ -1835,8 +2044,8 @@ static void __pool_dec(struct pool *pool)
static struct pool *__pool_find(struct mapped_device *pool_md,
struct block_device *metadata_dev,
unsigned long block_size, char **error,
int *created)
unsigned long block_size, int read_only,
char **error, int *created)
{
struct pool *pool = __pool_table_lookup_metadata_dev(metadata_dev);
......@@ -1857,7 +2066,7 @@ static struct pool *__pool_find(struct mapped_device *pool_md,
__pool_inc(pool);
} else {
pool = pool_create(pool_md, metadata_dev, block_size, error);
pool = pool_create(pool_md, metadata_dev, block_size, read_only, error);
*created = 1;
}
}
......@@ -1908,19 +2117,23 @@ static int parse_pool_features(struct dm_arg_set *as, struct pool_features *pf,
arg_name = dm_shift_arg(as);
argc--;
if (!strcasecmp(arg_name, "skip_block_zeroing")) {
if (!strcasecmp(arg_name, "skip_block_zeroing"))
pf->zero_new_blocks = 0;
continue;
} else if (!strcasecmp(arg_name, "ignore_discard")) {
else if (!strcasecmp(arg_name, "ignore_discard"))
pf->discard_enabled = 0;
continue;
} else if (!strcasecmp(arg_name, "no_discard_passdown")) {
else if (!strcasecmp(arg_name, "no_discard_passdown"))
pf->discard_passdown = 0;
continue;
}
ti->error = "Unrecognised pool feature requested";
r = -EINVAL;
else if (!strcasecmp(arg_name, "read_only"))
pf->mode = PM_READ_ONLY;
else {
ti->error = "Unrecognised pool feature requested";
r = -EINVAL;
break;
}
}
return r;
......@@ -2013,7 +2226,7 @@ static int pool_ctr(struct dm_target *ti, unsigned argc, char **argv)
}
pool = __pool_find(dm_table_get_md(ti->table), metadata_dev->bdev,
block_size, &ti->error, &pool_created);
block_size, pf.mode == PM_READ_ONLY, &ti->error, &pool_created);
if (IS_ERR(pool)) {
r = PTR_ERR(pool);
goto out_free_pt;
......@@ -2146,15 +2359,12 @@ static int pool_preresume(struct dm_target *ti)
r = dm_pool_resize_data_dev(pool->pmd, data_size);
if (r) {
DMERR("failed to resize data device");
/* FIXME Stricter than necessary: Rollback transaction instead here */
set_pool_mode(pool, PM_READ_ONLY);
return r;
}
r = dm_pool_commit_metadata(pool->pmd);
if (r) {
DMERR("%s: dm_pool_commit_metadata() failed, error = %d",
__func__, r);
return r;
}
(void) commit_or_fallback(pool);
}
return 0;
......@@ -2177,19 +2387,12 @@ static void pool_resume(struct dm_target *ti)
static void pool_postsuspend(struct dm_target *ti)
{
int r;
struct pool_c *pt = ti->private;
struct pool *pool = pt->pool;
cancel_delayed_work(&pool->waker);
flush_workqueue(pool->wq);
r = dm_pool_commit_metadata(pool->pmd);
if (r < 0) {
DMERR("%s: dm_pool_commit_metadata() failed, error = %d",
__func__, r);
/* FIXME: invalidate device? error the next FUA or FLUSH bio ?*/
}
(void) commit_or_fallback(pool);
}
static int check_arg_count(unsigned argc, unsigned args_required)
......@@ -2323,12 +2526,7 @@ static int process_reserve_metadata_snap_mesg(unsigned argc, char **argv, struct
if (r)
return r;
r = dm_pool_commit_metadata(pool->pmd);
if (r) {
DMERR("%s: dm_pool_commit_metadata() failed, error = %d",
__func__, r);
return r;
}
(void) commit_or_fallback(pool);
r = dm_pool_reserve_metadata_snap(pool->pmd);
if (r)
......@@ -2389,16 +2587,32 @@ static int pool_message(struct dm_target *ti, unsigned argc, char **argv)
else
DMWARN("Unrecognised thin pool target message received: %s", argv[0]);
if (!r) {
r = dm_pool_commit_metadata(pool->pmd);
if (r)
DMERR("%s message: dm_pool_commit_metadata() failed, error = %d",
argv[0], r);
}
if (!r)
(void) commit_or_fallback(pool);
return r;
}
static void em