aboutsummaryrefslogtreecommitdiff
path: root/fs/btrfs/block-group.c
diff options
context:
space:
mode:
Diffstat (limited to 'fs/btrfs/block-group.c')
-rw-r--r--fs/btrfs/block-group.c139
1 files changed, 111 insertions, 28 deletions
diff --git a/fs/btrfs/block-group.c b/fs/btrfs/block-group.c
index 957ad1c31c4f..82324c327a50 100644
--- a/fs/btrfs/block-group.c
+++ b/fs/btrfs/block-group.c
@@ -95,14 +95,21 @@ static u64 btrfs_reduce_alloc_profile(struct btrfs_fs_info *fs_info, u64 flags)
}
allowed &= flags;
- if (allowed & BTRFS_BLOCK_GROUP_RAID6)
+ /* Select the highest-redundancy RAID level. */
+ if (allowed & BTRFS_BLOCK_GROUP_RAID1C4)
+ allowed = BTRFS_BLOCK_GROUP_RAID1C4;
+ else if (allowed & BTRFS_BLOCK_GROUP_RAID6)
allowed = BTRFS_BLOCK_GROUP_RAID6;
+ else if (allowed & BTRFS_BLOCK_GROUP_RAID1C3)
+ allowed = BTRFS_BLOCK_GROUP_RAID1C3;
else if (allowed & BTRFS_BLOCK_GROUP_RAID5)
allowed = BTRFS_BLOCK_GROUP_RAID5;
else if (allowed & BTRFS_BLOCK_GROUP_RAID10)
allowed = BTRFS_BLOCK_GROUP_RAID10;
else if (allowed & BTRFS_BLOCK_GROUP_RAID1)
allowed = BTRFS_BLOCK_GROUP_RAID1;
+ else if (allowed & BTRFS_BLOCK_GROUP_DUP)
+ allowed = BTRFS_BLOCK_GROUP_DUP;
else if (allowed & BTRFS_BLOCK_GROUP_RAID0)
allowed = BTRFS_BLOCK_GROUP_RAID0;
@@ -434,13 +441,23 @@ void btrfs_wait_block_group_cache_progress(struct btrfs_block_group *cache,
u64 num_bytes)
{
struct btrfs_caching_control *caching_ctl;
+ int progress;
caching_ctl = btrfs_get_caching_control(cache);
if (!caching_ctl)
return;
+ /*
+ * We've already failed to allocate from this block group, so even if
+ * there's enough space in the block group it isn't contiguous enough to
+ * allow for an allocation, so wait for at least the next wakeup tick,
+ * or for the thing to be done.
+ */
+ progress = atomic_read(&caching_ctl->progress);
+
wait_event(caching_ctl->wait, btrfs_block_group_done(cache) ||
- (cache->free_space_ctl->free_space >= num_bytes));
+ (progress != atomic_read(&caching_ctl->progress) &&
+ (cache->free_space_ctl->free_space >= num_bytes)));
btrfs_put_caching_control(caching_ctl);
}
@@ -492,12 +509,16 @@ static void fragment_free_space(struct btrfs_block_group *block_group)
* used yet since their free space will be released as soon as the transaction
* commits.
*/
-u64 add_new_free_space(struct btrfs_block_group *block_group, u64 start, u64 end)
+int add_new_free_space(struct btrfs_block_group *block_group, u64 start, u64 end,
+ u64 *total_added_ret)
{
struct btrfs_fs_info *info = block_group->fs_info;
- u64 extent_start, extent_end, size, total_added = 0;
+ u64 extent_start, extent_end, size;
int ret;
+ if (total_added_ret)
+ *total_added_ret = 0;
+
while (start < end) {
ret = find_first_extent_bit(&info->excluded_extents, start,
&extent_start, &extent_end,
@@ -510,10 +531,12 @@ u64 add_new_free_space(struct btrfs_block_group *block_group, u64 start, u64 end
start = extent_end + 1;
} else if (extent_start > start && extent_start < end) {
size = extent_start - start;
- total_added += size;
ret = btrfs_add_free_space_async_trimmed(block_group,
start, size);
- BUG_ON(ret); /* -ENOMEM or logic error */
+ if (ret)
+ return ret;
+ if (total_added_ret)
+ *total_added_ret += size;
start = extent_end + 1;
} else {
break;
@@ -522,13 +545,15 @@ u64 add_new_free_space(struct btrfs_block_group *block_group, u64 start, u64 end
if (start < end) {
size = end - start;
- total_added += size;
ret = btrfs_add_free_space_async_trimmed(block_group, start,
size);
- BUG_ON(ret); /* -ENOMEM or logic error */
+ if (ret)
+ return ret;
+ if (total_added_ret)
+ *total_added_ret += size;
}
- return total_added;
+ return 0;
}
/*
@@ -772,8 +797,13 @@ next:
if (key.type == BTRFS_EXTENT_ITEM_KEY ||
key.type == BTRFS_METADATA_ITEM_KEY) {
- total_found += add_new_free_space(block_group, last,
- key.objectid);
+ u64 space_added;
+
+ ret = add_new_free_space(block_group, last, key.objectid,
+ &space_added);
+ if (ret)
+ goto out;
+ total_found += space_added;
if (key.type == BTRFS_METADATA_ITEM_KEY)
last = key.objectid +
fs_info->nodesize;
@@ -782,17 +812,18 @@ next:
if (total_found > CACHING_CTL_WAKE_UP) {
total_found = 0;
- if (wakeup)
+ if (wakeup) {
+ atomic_inc(&caching_ctl->progress);
wake_up(&caching_ctl->wait);
+ }
}
}
path->slots[0]++;
}
- ret = 0;
-
- total_found += add_new_free_space(block_group, last,
- block_group->start + block_group->length);
+ ret = add_new_free_space(block_group, last,
+ block_group->start + block_group->length,
+ NULL);
out:
btrfs_free_path(path);
return ret;
@@ -891,6 +922,7 @@ int btrfs_cache_block_group(struct btrfs_block_group *cache, bool wait)
init_waitqueue_head(&caching_ctl->wait);
caching_ctl->block_group = cache;
refcount_set(&caching_ctl->count, 2);
+ atomic_set(&caching_ctl->progress, 0);
btrfs_init_work(&caching_ctl->work, caching_thread, NULL, NULL);
spin_lock(&cache->lock);
@@ -1638,6 +1670,10 @@ void btrfs_mark_bg_unused(struct btrfs_block_group *bg)
btrfs_get_block_group(bg);
trace_btrfs_add_unused_block_group(bg);
list_add_tail(&bg->bg_list, &fs_info->unused_bgs);
+ } else if (!test_bit(BLOCK_GROUP_FLAG_NEW, &bg->runtime_flags)) {
+ /* Pull out the block group from the reclaim_bgs list. */
+ trace_btrfs_add_unused_block_group(bg);
+ list_move_tail(&bg->bg_list, &fs_info->unused_bgs);
}
spin_unlock(&fs_info->unused_bgs_lock);
}
@@ -1791,8 +1827,15 @@ void btrfs_reclaim_bgs_work(struct work_struct *work)
}
spin_unlock(&bg->lock);
- /* Get out fast, in case we're unmounting the filesystem */
- if (btrfs_fs_closing(fs_info)) {
+ /*
+ * Get out fast, in case we're read-only or unmounting the
+ * filesystem. It is OK to drop block groups from the list even
+ * for the read-only case. As we did sb_start_write(),
+ * "mount -o remount,ro" won't happen and read-only filesystem
+ * means it is forced read-only due to a fatal error. So, it
+ * never gets back to read-write to let us reclaim again.
+ */
+ if (btrfs_need_cleaner_sleep(fs_info)) {
up_write(&space_info->groups_sem);
goto next;
}
@@ -1823,11 +1866,27 @@ void btrfs_reclaim_bgs_work(struct work_struct *work)
}
next:
+ if (ret)
+ btrfs_mark_bg_to_reclaim(bg);
btrfs_put_block_group(bg);
+
+ mutex_unlock(&fs_info->reclaim_bgs_lock);
+ /*
+ * Reclaiming all the block groups in the list can take really
+ * long. Prioritize cleaning up unused block groups.
+ */
+ btrfs_delete_unused_bgs(fs_info);
+ /*
+ * If we are interrupted by a balance, we can just bail out. The
+ * cleaner thread restart again if necessary.
+ */
+ if (!mutex_trylock(&fs_info->reclaim_bgs_lock))
+ goto end;
spin_lock(&fs_info->unused_bgs_lock);
}
spin_unlock(&fs_info->unused_bgs_lock);
mutex_unlock(&fs_info->reclaim_bgs_lock);
+end:
btrfs_exclop_finish(fs_info);
sb_end_write(fs_info->sb);
}
@@ -1973,7 +2032,7 @@ int btrfs_rmap_block(struct btrfs_fs_info *fs_info, u64 chunk_start,
/* For RAID5/6 adjust to a full IO stripe length */
if (map->type & BTRFS_BLOCK_GROUP_RAID56_MASK)
- io_stripe_size = nr_data_stripes(map) << BTRFS_STRIPE_LEN_SHIFT;
+ io_stripe_size = btrfs_stripe_nr_to_offset(nr_data_stripes(map));
buf = kcalloc(map->num_stripes, sizeof(u64), GFP_NOFS);
if (!buf) {
@@ -2054,6 +2113,7 @@ static int exclude_super_stripes(struct btrfs_block_group *cache)
/* Shouldn't have super stripes in sequential zones */
if (zoned && nr) {
+ kfree(logical);
btrfs_err(fs_info,
"zoned: block group %llu must not contain super block",
cache->start);
@@ -2259,9 +2319,11 @@ static int read_one_block_group(struct btrfs_fs_info *info,
btrfs_free_excluded_extents(cache);
} else if (cache->used == 0) {
cache->cached = BTRFS_CACHE_FINISHED;
- add_new_free_space(cache, cache->start,
- cache->start + cache->length);
+ ret = add_new_free_space(cache, cache->start,
+ cache->start + cache->length, NULL);
btrfs_free_excluded_extents(cache);
+ if (ret)
+ goto error;
}
ret = btrfs_add_block_group_cache(info, cache);
@@ -2635,6 +2697,7 @@ void btrfs_create_pending_block_groups(struct btrfs_trans_handle *trans)
next:
btrfs_delayed_refs_rsv_release(fs_info, 1);
list_del_init(&block_group->bg_list);
+ clear_bit(BLOCK_GROUP_FLAG_NEW, &block_group->runtime_flags);
}
btrfs_trans_release_chunk_metadata(trans);
}
@@ -2674,6 +2737,13 @@ struct btrfs_block_group *btrfs_make_block_group(struct btrfs_trans_handle *tran
if (!cache)
return ERR_PTR(-ENOMEM);
+ /*
+ * Mark it as new before adding it to the rbtree of block groups or any
+ * list, so that no other task finds it and calls btrfs_mark_bg_unused()
+ * before the new flag is set.
+ */
+ set_bit(BLOCK_GROUP_FLAG_NEW, &cache->runtime_flags);
+
cache->length = size;
set_free_space_tree_thresholds(cache);
cache->flags = type;
@@ -2697,9 +2767,12 @@ struct btrfs_block_group *btrfs_make_block_group(struct btrfs_trans_handle *tran
return ERR_PTR(ret);
}
- add_new_free_space(cache, chunk_offset, chunk_offset + size);
-
+ ret = add_new_free_space(cache, chunk_offset, chunk_offset + size, NULL);
btrfs_free_excluded_extents(cache);
+ if (ret) {
+ btrfs_put_block_group(cache);
+ return ERR_PTR(ret);
+ }
/*
* Ensure the corresponding space_info object is created and
@@ -2818,10 +2891,20 @@ int btrfs_inc_block_group_ro(struct btrfs_block_group *cache,
}
ret = inc_block_group_ro(cache, 0);
- if (!do_chunk_alloc || ret == -ETXTBSY)
- goto unlock_out;
if (!ret)
goto out;
+ if (ret == -ETXTBSY)
+ goto unlock_out;
+
+ /*
+ * Skip chunk alloction if the bg is SYSTEM, this is to avoid system
+ * chunk allocation storm to exhaust the system chunk array. Otherwise
+ * we still want to try our best to mark the block group read-only.
+ */
+ if (!do_chunk_alloc && ret == -ENOSPC &&
+ (cache->flags & BTRFS_BLOCK_GROUP_SYSTEM))
+ goto unlock_out;
+
alloc_flags = btrfs_get_alloc_profile(fs_info, cache->space_info->flags);
ret = btrfs_chunk_alloc(trans, alloc_flags, CHUNK_ALLOC_FORCE);
if (ret < 0)
@@ -3511,9 +3594,9 @@ int btrfs_update_block_group(struct btrfs_trans_handle *trans,
spin_unlock(&cache->lock);
spin_unlock(&space_info->lock);
- set_extent_dirty(&trans->transaction->pinned_extents,
- bytenr, bytenr + num_bytes - 1,
- GFP_NOFS | __GFP_NOFAIL);
+ set_extent_bit(&trans->transaction->pinned_extents,
+ bytenr, bytenr + num_bytes - 1,
+ EXTENT_DIRTY, NULL);
}
spin_lock(&trans->transaction->dirty_bgs_lock);
Powered by cgit, Gruvboxed by Blaster4385.