summaryrefslogtreecommitdiff
path: root/drivers/md/dm-thin.c
diff options
context:
space:
mode:
authorJoe Thornber <ejt@redhat.com>2018-09-10 16:50:09 +0100
committerMike Snitzer <snitzer@redhat.com>2018-09-10 17:03:18 -0400
commit3ab91828166895600efd9cdc3a0eb32001f7204a (patch)
treeb0a357366b35d36d697c8f6f1d2f6fede0074b63 /drivers/md/dm-thin.c
parent5380c05b682991a6818c3755d450a3e87eeac0e5 (diff)
dm thin metadata: try to avoid ever aborting transactions
Committing a transaction can consume some metadata of it's own, we now reserve a small amount of metadata to cover this. Free metadata reported by the kernel will not include this reserve. If any of the reserve has been used after a commit we enter a new internal state PM_OUT_OF_METADATA_SPACE. This is reported as PM_READ_ONLY, so no userland changes are needed. If the metadata device is resized the pool will move back to PM_WRITE. These changes mean we never need to abort and rollback a transaction due to running out of metadata space. This is particularly important because there have been a handful of reports of data corruption against DM thin-provisioning that can all be attributed to the thin-pool having ran out of metadata space. Signed-off-by: Joe Thornber <ejt@redhat.com> Signed-off-by: Mike Snitzer <snitzer@redhat.com>
Diffstat (limited to 'drivers/md/dm-thin.c')
-rw-r--r--drivers/md/dm-thin.c73
1 files changed, 65 insertions, 8 deletions
diff --git a/drivers/md/dm-thin.c b/drivers/md/dm-thin.c
index 7bd60a150f8f..aaf1ad481ee8 100644
--- a/drivers/md/dm-thin.c
+++ b/drivers/md/dm-thin.c
@@ -200,7 +200,13 @@ struct dm_thin_new_mapping;
enum pool_mode {
PM_WRITE, /* metadata may be changed */
PM_OUT_OF_DATA_SPACE, /* metadata may be changed, though data may not be allocated */
+
+ /*
+ * Like READ_ONLY, except may switch back to WRITE on metadata resize. Reported as READ_ONLY.
+ */
+ PM_OUT_OF_METADATA_SPACE,
PM_READ_ONLY, /* metadata may not be changed */
+
PM_FAIL, /* all I/O fails */
};
@@ -1371,7 +1377,35 @@ static void set_pool_mode(struct pool *pool, enum pool_mode new_mode);
static void requeue_bios(struct pool *pool);
-static void check_for_space(struct pool *pool)
+static bool is_read_only_pool_mode(enum pool_mode mode)
+{
+ return (mode == PM_OUT_OF_METADATA_SPACE || mode == PM_READ_ONLY);
+}
+
+static bool is_read_only(struct pool *pool)
+{
+ return is_read_only_pool_mode(get_pool_mode(pool));
+}
+
+static void check_for_metadata_space(struct pool *pool)
+{
+ int r;
+ const char *ooms_reason = NULL;
+ dm_block_t nr_free;
+
+ r = dm_pool_get_free_metadata_block_count(pool->pmd, &nr_free);
+ if (r)
+ ooms_reason = "Could not get free metadata blocks";
+ else if (!nr_free)
+ ooms_reason = "No free metadata blocks";
+
+ if (ooms_reason && !is_read_only(pool)) {
+ DMERR("%s", ooms_reason);
+ set_pool_mode(pool, PM_OUT_OF_METADATA_SPACE);
+ }
+}
+
+static void check_for_data_space(struct pool *pool)
{
int r;
dm_block_t nr_free;
@@ -1397,14 +1431,16 @@ static int commit(struct pool *pool)
{
int r;
- if (get_pool_mode(pool) >= PM_READ_ONLY)
+ if (get_pool_mode(pool) >= PM_OUT_OF_METADATA_SPACE)
return -EINVAL;
r = dm_pool_commit_metadata(pool->pmd);
if (r)
metadata_operation_failed(pool, "dm_pool_commit_metadata", r);
- else
- check_for_space(pool);
+ else {
+ check_for_metadata_space(pool);
+ check_for_data_space(pool);
+ }
return r;
}
@@ -1470,6 +1506,19 @@ static int alloc_data_block(struct thin_c *tc, dm_block_t *result)
return r;
}
+ r = dm_pool_get_free_metadata_block_count(pool->pmd, &free_blocks);
+ if (r) {
+ metadata_operation_failed(pool, "dm_pool_get_free_metadata_block_count", r);
+ return r;
+ }
+
+ if (!free_blocks) {
+ /* Let's commit before we use up the metadata reserve. */
+ r = commit(pool);
+ if (r)
+ return r;
+ }
+
return 0;
}
@@ -1501,6 +1550,7 @@ static blk_status_t should_error_unserviceable_bio(struct pool *pool)
case PM_OUT_OF_DATA_SPACE:
return pool->pf.error_if_no_space ? BLK_STS_NOSPC : 0;
+ case PM_OUT_OF_METADATA_SPACE:
case PM_READ_ONLY:
case PM_FAIL:
return BLK_STS_IOERR;
@@ -2464,8 +2514,9 @@ static void set_pool_mode(struct pool *pool, enum pool_mode new_mode)
error_retry_list(pool);
break;
+ case PM_OUT_OF_METADATA_SPACE:
case PM_READ_ONLY:
- if (old_mode != new_mode)
+ if (!is_read_only_pool_mode(old_mode))
notify_of_pool_mode_change(pool, "read-only");
dm_pool_metadata_read_only(pool->pmd);
pool->process_bio = process_bio_read_only;
@@ -3403,6 +3454,10 @@ static int maybe_resize_metadata_dev(struct dm_target *ti, bool *need_commit)
DMINFO("%s: growing the metadata device from %llu to %llu blocks",
dm_device_name(pool->pool_md),
sb_metadata_dev_size, metadata_dev_size);
+
+ if (get_pool_mode(pool) == PM_OUT_OF_METADATA_SPACE)
+ set_pool_mode(pool, PM_WRITE);
+
r = dm_pool_resize_metadata_dev(pool->pmd, metadata_dev_size);
if (r) {
metadata_operation_failed(pool, "dm_pool_resize_metadata_dev", r);
@@ -3707,7 +3762,7 @@ static int pool_message(struct dm_target *ti, unsigned argc, char **argv,
struct pool_c *pt = ti->private;
struct pool *pool = pt->pool;
- if (get_pool_mode(pool) >= PM_READ_ONLY) {
+ if (get_pool_mode(pool) >= PM_OUT_OF_METADATA_SPACE) {
DMERR("%s: unable to service pool target messages in READ_ONLY or FAIL mode",
dm_device_name(pool->pool_md));
return -EOPNOTSUPP;
@@ -3781,6 +3836,7 @@ static void pool_status(struct dm_target *ti, status_type_t type,
dm_block_t nr_blocks_data;
dm_block_t nr_blocks_metadata;
dm_block_t held_root;
+ enum pool_mode mode;
char buf[BDEVNAME_SIZE];
char buf2[BDEVNAME_SIZE];
struct pool_c *pt = ti->private;
@@ -3851,9 +3907,10 @@ static void pool_status(struct dm_target *ti, status_type_t type,
else
DMEMIT("- ");
- if (pool->pf.mode == PM_OUT_OF_DATA_SPACE)
+ mode = get_pool_mode(pool);
+ if (mode == PM_OUT_OF_DATA_SPACE)
DMEMIT("out_of_data_space ");
- else if (pool->pf.mode == PM_READ_ONLY)
+ else if (is_read_only_pool_mode(mode))
DMEMIT("ro ");
else
DMEMIT("rw ");