From d04c6b88320debb403bff8d8b634a1efa48b8d3d Mon Sep 17 00:00:00 2001 From: Jeff Mahoney Date: Mon, 15 Jun 2015 09:41:14 -0400 Subject: [PATCH 1/6] btrfs: make btrfs_issue_discard return bytes discarded Initially this will just be the length argument passed to it, but the following patches will adjust that to reflect re-alignment and skipped blocks. Signed-off-by: Jeff Mahoney Reviewed-by: Filipe Manana Tested-by: Filipe Manana Signed-off-by: Chris Mason --- fs/btrfs/extent-tree.c | 19 ++++++++++++++----- 1 file changed, 14 insertions(+), 5 deletions(-) diff --git a/fs/btrfs/extent-tree.c b/fs/btrfs/extent-tree.c index 07204bf601ed..16655bb5f293 100644 --- a/fs/btrfs/extent-tree.c +++ b/fs/btrfs/extent-tree.c @@ -1883,10 +1883,17 @@ static int remove_extent_backref(struct btrfs_trans_handle *trans, return ret; } -static int btrfs_issue_discard(struct block_device *bdev, - u64 start, u64 len) +static int btrfs_issue_discard(struct block_device *bdev, u64 start, u64 len, + u64 *discarded_bytes) { - return blkdev_issue_discard(bdev, start >> 9, len >> 9, GFP_NOFS, 0); + int ret = 0; + + *discarded_bytes = 0; + ret = blkdev_issue_discard(bdev, start >> 9, len >> 9, GFP_NOFS, 0); + if (!ret) + *discarded_bytes = len; + + return ret; } int btrfs_discard_extent(struct btrfs_root *root, u64 bytenr, @@ -1907,14 +1914,16 @@ int btrfs_discard_extent(struct btrfs_root *root, u64 bytenr, for (i = 0; i < bbio->num_stripes; i++, stripe++) { + u64 bytes; if (!stripe->dev->can_discard) continue; ret = btrfs_issue_discard(stripe->dev->bdev, stripe->physical, - stripe->length); + stripe->length, + &bytes); if (!ret) - discarded_bytes += stripe->length; + discarded_bytes += bytes; else if (ret != -EOPNOTSUPP) break; /* Logic errors or -ENOMEM, or -EIO but I don't know how that could happen JDM */ From 4d89d377bbb0e34cd1571b57a984c2326cab69b5 Mon Sep 17 00:00:00 2001 From: Jeff Mahoney Date: Mon, 15 Jun 2015 09:41:15 -0400 Subject: [PATCH 2/6] btrfs: btrfs_issue_discard ensure offset/length are aligned to sector boundaries It's possible, though unexpected, to pass unaligned offsets and lengths to btrfs_issue_discard. We then shift the offset/length values to sector units. If an unaligned offset has been passed, it will result in the entire sector being discarded, possibly losing data. An unaligned length is safe but we'll end up returning an inaccurate number of discarded bytes. This patch aligns the offset to the 512B boundary, adjusts the length, and warns, since we shouldn't be discarding on an offset that isn't aligned with our sector size. Signed-off-by: Jeff Mahoney Reviewed-by: Filipe Manana Tested-by: Filipe Manana Signed-off-by: Chris Mason --- fs/btrfs/extent-tree.c | 17 +++++++++++++---- 1 file changed, 13 insertions(+), 4 deletions(-) diff --git a/fs/btrfs/extent-tree.c b/fs/btrfs/extent-tree.c index 16655bb5f293..7aa6ad1f014d 100644 --- a/fs/btrfs/extent-tree.c +++ b/fs/btrfs/extent-tree.c @@ -1887,12 +1887,21 @@ static int btrfs_issue_discard(struct block_device *bdev, u64 start, u64 len, u64 *discarded_bytes) { int ret = 0; + u64 aligned_start = ALIGN(start, 1 << 9); + + if (WARN_ON(start != aligned_start)) { + len -= aligned_start - start; + len = round_down(len, 1 << 9); + start = aligned_start; + } *discarded_bytes = 0; - ret = blkdev_issue_discard(bdev, start >> 9, len >> 9, GFP_NOFS, 0); - if (!ret) - *discarded_bytes = len; - + if (len) { + ret = blkdev_issue_discard(bdev, start >> 9, len >> 9, + GFP_NOFS, 0); + if (!ret) + *discarded_bytes = len; + } return ret; } From 86557861dfe4f8defde0df40620b97cc60285aa4 Mon Sep 17 00:00:00 2001 From: Jeff Mahoney Date: Mon, 15 Jun 2015 09:41:16 -0400 Subject: [PATCH 3/6] btrfs: skip superblocks during discard Btrfs doesn't track superblocks with extent records so there is nothing persistent on-disk to indicate that those blocks are in use. We track the superblocks in memory to ensure they don't get used by removing them from the free space cache when we load a block group from disk. Prior to 47ab2a6c6a (Btrfs: remove empty block groups automatically), that was fine since the block group would never be reclaimed so the superblock was always safe. Once we started removing the empty block groups, we were protected by the fact that discards weren't being properly issued for unused space either via FITRIM or -odiscard. The block groups were still being released, but the blocks remained on disk. In order to properly discard unused block groups, we need to filter out the superblocks from the discard range. Superblocks are located at fixed locations on each device, so it makes sense to filter them out in btrfs_issue_discard, which is used by both -odiscard and FITRIM. Signed-off-by: Jeff Mahoney Reviewed-by: Filipe Manana Tested-by: Filipe Manana Signed-off-by: Chris Mason --- fs/btrfs/extent-tree.c | 59 +++++++++++++++++++++++++++++++++++++++--- 1 file changed, 55 insertions(+), 4 deletions(-) diff --git a/fs/btrfs/extent-tree.c b/fs/btrfs/extent-tree.c index 7aa6ad1f014d..d763457b3cce 100644 --- a/fs/btrfs/extent-tree.c +++ b/fs/btrfs/extent-tree.c @@ -1883,10 +1883,12 @@ static int remove_extent_backref(struct btrfs_trans_handle *trans, return ret; } +#define in_range(b, first, len) ((b) >= (first) && (b) < (first) + (len)) static int btrfs_issue_discard(struct block_device *bdev, u64 start, u64 len, u64 *discarded_bytes) { - int ret = 0; + int j, ret = 0; + u64 bytes_left, end; u64 aligned_start = ALIGN(start, 1 << 9); if (WARN_ON(start != aligned_start)) { @@ -1896,11 +1898,60 @@ static int btrfs_issue_discard(struct block_device *bdev, u64 start, u64 len, } *discarded_bytes = 0; - if (len) { - ret = blkdev_issue_discard(bdev, start >> 9, len >> 9, + + if (!len) + return 0; + + end = start + len; + bytes_left = len; + + /* Skip any superblocks on this device. */ + for (j = 0; j < BTRFS_SUPER_MIRROR_MAX; j++) { + u64 sb_start = btrfs_sb_offset(j); + u64 sb_end = sb_start + BTRFS_SUPER_INFO_SIZE; + u64 size = sb_start - start; + + if (!in_range(sb_start, start, bytes_left) && + !in_range(sb_end, start, bytes_left) && + !in_range(start, sb_start, BTRFS_SUPER_INFO_SIZE)) + continue; + + /* + * Superblock spans beginning of range. Adjust start and + * try again. + */ + if (sb_start <= start) { + start += sb_end - start; + if (start > end) { + bytes_left = 0; + break; + } + bytes_left = end - start; + continue; + } + + if (size) { + ret = blkdev_issue_discard(bdev, start >> 9, size >> 9, + GFP_NOFS, 0); + if (!ret) + *discarded_bytes += size; + else if (ret != -EOPNOTSUPP) + return ret; + } + + start = sb_end; + if (start > end) { + bytes_left = 0; + break; + } + bytes_left = end - start; + } + + if (bytes_left) { + ret = blkdev_issue_discard(bdev, start >> 9, bytes_left >> 9, GFP_NOFS, 0); if (!ret) - *discarded_bytes = len; + *discarded_bytes += bytes_left; } return ret; } From 499f377f49f085ee4aa214c738e948e88626f39b Mon Sep 17 00:00:00 2001 From: Jeff Mahoney Date: Mon, 15 Jun 2015 09:41:17 -0400 Subject: [PATCH 4/6] btrfs: iterate over unused chunk space in FITRIM Since we now clean up block groups automatically as they become empty, iterating over block groups is no longer sufficient to discard unused space. This patch iterates over the unused chunk space and discards any regions that are unallocated, regardless of whether they were ever used. This is a change for btrfs but is consistent with other file systems. We do this in a transactionless manner since the discard process can take a substantial amount of time and a transaction would need to be started before the acquisition of the device list lock. That would mean a transaction would be held open across /all/ of the discards collectively. In order to prevent other threads from allocating or freeing chunks, we hold the chunks lock across the search and discard calls. We release it between searches to allow the file system to perform more-or-less normally. Since the running transaction can commit and disappear while we're using the transaction pointer, we take a reference to it and release it after the search. This is safe since it would happen normally at the end of the transaction commit after any locks are released anyway. We also take the commit_root_sem to protect against a transaction starting and committing while we're running. Signed-off-by: Jeff Mahoney Reviewed-by: Filipe Manana Tested-by: Filipe Manana Signed-off-by: Chris Mason --- fs/btrfs/extent-tree.c | 101 +++++++++++++++++++++++++++++++++++++++++ fs/btrfs/volumes.c | 63 +++++++++++++++---------- fs/btrfs/volumes.h | 3 ++ 3 files changed, 143 insertions(+), 24 deletions(-) diff --git a/fs/btrfs/extent-tree.c b/fs/btrfs/extent-tree.c index d763457b3cce..15411aefbfa0 100644 --- a/fs/btrfs/extent-tree.c +++ b/fs/btrfs/extent-tree.c @@ -10135,10 +10135,99 @@ int btrfs_error_unpin_extent_range(struct btrfs_root *root, u64 start, u64 end) return unpin_extent_range(root, start, end, false); } +/* + * It used to be that old block groups would be left around forever. + * Iterating over them would be enough to trim unused space. Since we + * now automatically remove them, we also need to iterate over unallocated + * space. + * + * We don't want a transaction for this since the discard may take a + * substantial amount of time. We don't require that a transaction be + * running, but we do need to take a running transaction into account + * to ensure that we're not discarding chunks that were released in + * the current transaction. + * + * Holding the chunks lock will prevent other threads from allocating + * or releasing chunks, but it won't prevent a running transaction + * from committing and releasing the memory that the pending chunks + * list head uses. For that, we need to take a reference to the + * transaction. + */ +static int btrfs_trim_free_extents(struct btrfs_device *device, + u64 minlen, u64 *trimmed) +{ + u64 start = 0, len = 0; + int ret; + + *trimmed = 0; + + /* Not writeable = nothing to do. */ + if (!device->writeable) + return 0; + + /* No free space = nothing to do. */ + if (device->total_bytes <= device->bytes_used) + return 0; + + ret = 0; + + while (1) { + struct btrfs_fs_info *fs_info = device->dev_root->fs_info; + struct btrfs_transaction *trans; + u64 bytes; + + ret = mutex_lock_interruptible(&fs_info->chunk_mutex); + if (ret) + return ret; + + down_read(&fs_info->commit_root_sem); + + spin_lock(&fs_info->trans_lock); + trans = fs_info->running_transaction; + if (trans) + atomic_inc(&trans->use_count); + spin_unlock(&fs_info->trans_lock); + + ret = find_free_dev_extent_start(trans, device, minlen, start, + &start, &len); + if (trans) + btrfs_put_transaction(trans); + + if (ret) { + up_read(&fs_info->commit_root_sem); + mutex_unlock(&fs_info->chunk_mutex); + if (ret == -ENOSPC) + ret = 0; + break; + } + + ret = btrfs_issue_discard(device->bdev, start, len, &bytes); + up_read(&fs_info->commit_root_sem); + mutex_unlock(&fs_info->chunk_mutex); + + if (ret) + break; + + start += len; + *trimmed += bytes; + + if (fatal_signal_pending(current)) { + ret = -ERESTARTSYS; + break; + } + + cond_resched(); + } + + return ret; +} + int btrfs_trim_fs(struct btrfs_root *root, struct fstrim_range *range) { struct btrfs_fs_info *fs_info = root->fs_info; struct btrfs_block_group_cache *cache = NULL; + struct btrfs_device *device; + struct list_head *devices; u64 group_trimmed; u64 start; u64 end; @@ -10193,6 +10282,18 @@ int btrfs_trim_fs(struct btrfs_root *root, struct fstrim_range *range) cache = next_block_group(fs_info->tree_root, cache); } + mutex_lock(&root->fs_info->fs_devices->device_list_mutex); + devices = &root->fs_info->fs_devices->alloc_list; + list_for_each_entry(device, devices, dev_alloc_list) { + ret = btrfs_trim_free_extents(device, range->minlen, + &group_trimmed); + if (ret) + break; + + trimmed += group_trimmed; + } + mutex_unlock(&root->fs_info->fs_devices->device_list_mutex); + range->len = trimmed; return ret; } diff --git a/fs/btrfs/volumes.c b/fs/btrfs/volumes.c index 9b95503ddd00..141c6051cf58 100644 --- a/fs/btrfs/volumes.c +++ b/fs/btrfs/volumes.c @@ -1116,15 +1116,18 @@ out: return ret; } -static int contains_pending_extent(struct btrfs_trans_handle *trans, +static int contains_pending_extent(struct btrfs_transaction *transaction, struct btrfs_device *device, u64 *start, u64 len) { + struct btrfs_fs_info *fs_info = device->dev_root->fs_info; struct extent_map *em; - struct list_head *search_list = &trans->transaction->pending_chunks; + struct list_head *search_list = &fs_info->pinned_chunks; int ret = 0; u64 physical_start = *start; + if (transaction) + search_list = &transaction->pending_chunks; again: list_for_each_entry(em, search_list, list) { struct map_lookup *map; @@ -1159,8 +1162,8 @@ again: } } } - if (search_list == &trans->transaction->pending_chunks) { - search_list = &trans->root->fs_info->pinned_chunks; + if (search_list != &fs_info->pinned_chunks) { + search_list = &fs_info->pinned_chunks; goto again; } @@ -1169,12 +1172,13 @@ again: /* - * find_free_dev_extent - find free space in the specified device - * @device: the device which we search the free space in - * @num_bytes: the size of the free space that we need - * @start: store the start of the free space. - * @len: the size of the free space. that we find, or the size of the max - * free space if we don't find suitable free space + * find_free_dev_extent_start - find free space in the specified device + * @device: the device which we search the free space in + * @num_bytes: the size of the free space that we need + * @search_start: the position from which to begin the search + * @start: store the start of the free space. + * @len: the size of the free space. that we find, or the size + * of the max free space if we don't find suitable free space * * this uses a pretty simple search, the expectation is that it is * called very infrequently and that a given device has a small number @@ -1188,9 +1192,9 @@ again: * But if we don't find suitable free space, it is used to store the size of * the max free space. */ -int find_free_dev_extent(struct btrfs_trans_handle *trans, - struct btrfs_device *device, u64 num_bytes, - u64 *start, u64 *len) +int find_free_dev_extent_start(struct btrfs_transaction *transaction, + struct btrfs_device *device, u64 num_bytes, + u64 search_start, u64 *start, u64 *len) { struct btrfs_key key; struct btrfs_root *root = device->dev_root; @@ -1200,19 +1204,11 @@ int find_free_dev_extent(struct btrfs_trans_handle *trans, u64 max_hole_start; u64 max_hole_size; u64 extent_end; - u64 search_start; u64 search_end = device->total_bytes; int ret; int slot; struct extent_buffer *l; - /* FIXME use last free of some kind */ - - /* we don't want to overwrite the superblock on the drive, - * so we make sure to start at an offset of at least 1MB - */ - search_start = max(root->fs_info->alloc_start, 1024ull * 1024); - path = btrfs_alloc_path(); if (!path) return -ENOMEM; @@ -1273,7 +1269,7 @@ again: * Have to check before we set max_hole_start, otherwise * we could end up sending back this offset anyway. */ - if (contains_pending_extent(trans, device, + if (contains_pending_extent(transaction, device, &search_start, hole_size)) { if (key.offset >= search_start) { @@ -1322,7 +1318,7 @@ next: if (search_end > search_start) { hole_size = search_end - search_start; - if (contains_pending_extent(trans, device, &search_start, + if (contains_pending_extent(transaction, device, &search_start, hole_size)) { btrfs_release_path(path); goto again; @@ -1348,6 +1344,24 @@ out: return ret; } +int find_free_dev_extent(struct btrfs_trans_handle *trans, + struct btrfs_device *device, u64 num_bytes, + u64 *start, u64 *len) +{ + struct btrfs_root *root = device->dev_root; + u64 search_start; + + /* FIXME use last free of some kind */ + + /* + * we don't want to overwrite the superblock on the drive, + * so we make sure to start at an offset of at least 1MB + */ + search_start = max(root->fs_info->alloc_start, 1024ull * 1024); + return find_free_dev_extent_start(trans->transaction, device, + num_bytes, search_start, start, len); +} + static int btrfs_free_dev_extent(struct btrfs_trans_handle *trans, struct btrfs_device *device, u64 start, u64 *dev_extent_len) @@ -4200,7 +4214,8 @@ again: u64 start = new_size; u64 len = old_size - new_size; - if (contains_pending_extent(trans, device, &start, len)) { + if (contains_pending_extent(trans->transaction, device, + &start, len)) { unlock_chunks(root); checked_pending_chunks = true; failed = 0; diff --git a/fs/btrfs/volumes.h b/fs/btrfs/volumes.h index 210a64390f40..57b0217b5300 100644 --- a/fs/btrfs/volumes.h +++ b/fs/btrfs/volumes.h @@ -455,6 +455,9 @@ int btrfs_cancel_balance(struct btrfs_fs_info *fs_info); int btrfs_create_uuid_tree(struct btrfs_fs_info *fs_info); int btrfs_check_uuid_tree(struct btrfs_fs_info *fs_info); int btrfs_chunk_readonly(struct btrfs_root *root, u64 chunk_offset); +int find_free_dev_extent_start(struct btrfs_transaction *transaction, + struct btrfs_device *device, u64 num_bytes, + u64 search_start, u64 *start, u64 *max_avail); int find_free_dev_extent(struct btrfs_trans_handle *trans, struct btrfs_device *device, u64 num_bytes, u64 *start, u64 *max_avail); From e44163e177960ee60e32a73bffdd53c3a5827406 Mon Sep 17 00:00:00 2001 From: Jeff Mahoney Date: Mon, 15 Jun 2015 09:41:18 -0400 Subject: [PATCH 5/6] btrfs: explictly delete unused block groups in close_ctree and ro-remount The cleaner thread may already be sleeping by the time we enter close_ctree. If that's the case, we'll skip removing any unused block groups queued for removal, even during a normal umount. They'll be cleaned up automatically at next mount, but users expect a umount to be a clean synchronization point, especially when used on thin-provisioned storage with -odiscard. We also explicitly remove unused block groups in the ro-remount path for the same reason. Signed-off-by: Jeff Mahoney Reviewed-by: Filipe Manana Tested-by: Filipe Manana Signed-off-by: Chris Mason --- fs/btrfs/disk-io.c | 9 +++++++++ fs/btrfs/super.c | 11 +++++++++++ 2 files changed, 20 insertions(+) diff --git a/fs/btrfs/disk-io.c b/fs/btrfs/disk-io.c index 84cbbb2d562e..053109ba26b7 100644 --- a/fs/btrfs/disk-io.c +++ b/fs/btrfs/disk-io.c @@ -3767,6 +3767,15 @@ void close_ctree(struct btrfs_root *root) cancel_work_sync(&fs_info->async_reclaim_work); if (!(fs_info->sb->s_flags & MS_RDONLY)) { + /* + * If the cleaner thread is stopped and there are + * block groups queued for removal, the deletion will be + * skipped when we quit the cleaner thread. + */ + mutex_lock(&root->fs_info->cleaner_mutex); + btrfs_delete_unused_bgs(root->fs_info); + mutex_unlock(&root->fs_info->cleaner_mutex); + ret = btrfs_commit_super(root); if (ret) btrfs_err(fs_info, "commit super ret %d", ret); diff --git a/fs/btrfs/super.c b/fs/btrfs/super.c index cd7ef34d2dce..a1077e0ffaa8 100644 --- a/fs/btrfs/super.c +++ b/fs/btrfs/super.c @@ -1650,6 +1650,17 @@ static int btrfs_remount(struct super_block *sb, int *flags, char *data) sb->s_flags |= MS_RDONLY; + /* + * Setting MS_RDONLY will put the cleaner thread to + * sleep at the next loop if it's already active. + * If it's already asleep, we'll leave unused block + * groups on disk until we're mounted read-write again + * unless we clean them up here. + */ + mutex_lock(&root->fs_info->cleaner_mutex); + btrfs_delete_unused_bgs(fs_info); + mutex_unlock(&root->fs_info->cleaner_mutex); + btrfs_dev_replace_suspend_for_unmount(fs_info); btrfs_scrub_cancel(fs_info); btrfs_pause_balance(fs_info); From e33e17ee1098d8d751552ac11c111e1c1a3db014 Mon Sep 17 00:00:00 2001 From: Jeff Mahoney Date: Mon, 15 Jun 2015 09:41:19 -0400 Subject: [PATCH 6/6] btrfs: add missing discards when unpinning extents with -o discard When we clear the dirty bits in btrfs_delete_unused_bgs for extents in the empty block group, it results in btrfs_finish_extent_commit being unable to discard the freed extents. The block group removal patch added an alternate path to forget extents other than btrfs_finish_extent_commit. As a result, any extents that would be freed when the block group is removed aren't discarded. In my test run, with a large copy of mixed sized files followed by removal, it left nearly 2/3 of extents undiscarded. To clean up the block groups, we add the removed block group onto a list that will be discarded after transaction commit. Signed-off-by: Jeff Mahoney Reviewed-by: Filipe Manana Tested-by: Filipe Manana Signed-off-by: Chris Mason --- fs/btrfs/ctree.h | 3 ++ fs/btrfs/extent-tree.c | 68 ++++++++++++++++++++++++++++++++++--- fs/btrfs/free-space-cache.c | 59 ++++++++++++++++++-------------- fs/btrfs/super.c | 2 +- fs/btrfs/transaction.c | 2 ++ fs/btrfs/transaction.h | 2 ++ 6 files changed, 106 insertions(+), 30 deletions(-) diff --git a/fs/btrfs/ctree.h b/fs/btrfs/ctree.h index aac314e14188..19ef3f306559 100644 --- a/fs/btrfs/ctree.h +++ b/fs/btrfs/ctree.h @@ -3437,6 +3437,8 @@ int btrfs_remove_block_group(struct btrfs_trans_handle *trans, struct btrfs_root *root, u64 group_start, struct extent_map *em); void btrfs_delete_unused_bgs(struct btrfs_fs_info *fs_info); +void btrfs_get_block_group_trimming(struct btrfs_block_group_cache *cache); +void btrfs_put_block_group_trimming(struct btrfs_block_group_cache *cache); void btrfs_create_pending_block_groups(struct btrfs_trans_handle *trans, struct btrfs_root *root); u64 btrfs_get_alloc_profile(struct btrfs_root *root, int data); @@ -4073,6 +4075,7 @@ __cold void __btrfs_std_error(struct btrfs_fs_info *fs_info, const char *function, unsigned int line, int errno, const char *fmt, ...); +const char *btrfs_decode_error(int errno); __cold void __btrfs_abort_transaction(struct btrfs_trans_handle *trans, diff --git a/fs/btrfs/extent-tree.c b/fs/btrfs/extent-tree.c index 15411aefbfa0..6b791f394698 100644 --- a/fs/btrfs/extent-tree.c +++ b/fs/btrfs/extent-tree.c @@ -6131,20 +6131,19 @@ int btrfs_finish_extent_commit(struct btrfs_trans_handle *trans, struct btrfs_root *root) { struct btrfs_fs_info *fs_info = root->fs_info; + struct btrfs_block_group_cache *block_group, *tmp; + struct list_head *deleted_bgs; struct extent_io_tree *unpin; u64 start; u64 end; int ret; - if (trans->aborted) - return 0; - if (fs_info->pinned_extents == &fs_info->freed_extents[0]) unpin = &fs_info->freed_extents[1]; else unpin = &fs_info->freed_extents[0]; - while (1) { + while (!trans->aborted) { mutex_lock(&fs_info->unused_bg_unpin_mutex); ret = find_first_extent_bit(unpin, 0, &start, &end, EXTENT_DIRTY, NULL); @@ -6163,6 +6162,34 @@ int btrfs_finish_extent_commit(struct btrfs_trans_handle *trans, cond_resched(); } + /* + * Transaction is finished. We don't need the lock anymore. We + * do need to clean up the block groups in case of a transaction + * abort. + */ + deleted_bgs = &trans->transaction->deleted_bgs; + list_for_each_entry_safe(block_group, tmp, deleted_bgs, bg_list) { + u64 trimmed = 0; + + ret = -EROFS; + if (!trans->aborted) + ret = btrfs_discard_extent(root, + block_group->key.objectid, + block_group->key.offset, + &trimmed); + + list_del_init(&block_group->bg_list); + btrfs_put_block_group_trimming(block_group); + btrfs_put_block_group(block_group); + + if (ret) { + const char *errstr = btrfs_decode_error(ret); + btrfs_warn(fs_info, + "Discard failed while removing blockgroup: errno=%d %s\n", + ret, errstr); + } + } + return 0; } @@ -9903,6 +9930,11 @@ int btrfs_remove_block_group(struct btrfs_trans_handle *trans, * currently running transaction might finish and a new one start, * allowing for new block groups to be created that can reuse the same * physical device locations unless we take this special care. + * + * There may also be an implicit trim operation if the file system + * is mounted with -odiscard. The same protections must remain + * in place until the extents have been discarded completely when + * the transaction commit has completed. */ remove_em = (atomic_read(&block_group->trimming) == 0); /* @@ -9977,6 +10009,7 @@ void btrfs_delete_unused_bgs(struct btrfs_fs_info *fs_info) spin_lock(&fs_info->unused_bgs_lock); while (!list_empty(&fs_info->unused_bgs)) { u64 start, end; + int trimming; block_group = list_first_entry(&fs_info->unused_bgs, struct btrfs_block_group_cache, @@ -10076,12 +10109,39 @@ void btrfs_delete_unused_bgs(struct btrfs_fs_info *fs_info) spin_unlock(&block_group->lock); spin_unlock(&space_info->lock); + /* DISCARD can flip during remount */ + trimming = btrfs_test_opt(root, DISCARD); + + /* Implicit trim during transaction commit. */ + if (trimming) + btrfs_get_block_group_trimming(block_group); + /* * Btrfs_remove_chunk will abort the transaction if things go * horribly wrong. */ ret = btrfs_remove_chunk(trans, root, block_group->key.objectid); + + if (ret) { + if (trimming) + btrfs_put_block_group_trimming(block_group); + goto end_trans; + } + + /* + * If we're not mounted with -odiscard, we can just forget + * about this block group. Otherwise we'll need to wait + * until transaction commit to do the actual discard. + */ + if (trimming) { + WARN_ON(!list_empty(&block_group->bg_list)); + spin_lock(&trans->transaction->deleted_bgs_lock); + list_move(&block_group->bg_list, + &trans->transaction->deleted_bgs); + spin_unlock(&trans->transaction->deleted_bgs_lock); + btrfs_get_block_group(block_group); + } end_trans: btrfs_end_transaction(trans, root); next: diff --git a/fs/btrfs/free-space-cache.c b/fs/btrfs/free-space-cache.c index fb5a6b1c62a6..abe3a66bd3ba 100644 --- a/fs/btrfs/free-space-cache.c +++ b/fs/btrfs/free-space-cache.c @@ -3272,35 +3272,23 @@ next: return ret; } -int btrfs_trim_block_group(struct btrfs_block_group_cache *block_group, - u64 *trimmed, u64 start, u64 end, u64 minlen) +void btrfs_get_block_group_trimming(struct btrfs_block_group_cache *cache) { - int ret; + atomic_inc(&cache->trimming); +} - *trimmed = 0; +void btrfs_put_block_group_trimming(struct btrfs_block_group_cache *block_group) +{ + struct extent_map_tree *em_tree; + struct extent_map *em; + bool cleanup; spin_lock(&block_group->lock); - if (block_group->removed) { - spin_unlock(&block_group->lock); - return 0; - } - atomic_inc(&block_group->trimming); + cleanup = (atomic_dec_and_test(&block_group->trimming) && + block_group->removed); spin_unlock(&block_group->lock); - ret = trim_no_bitmap(block_group, trimmed, start, end, minlen); - if (ret) - goto out; - - ret = trim_bitmaps(block_group, trimmed, start, end, minlen); -out: - spin_lock(&block_group->lock); - if (atomic_dec_and_test(&block_group->trimming) && - block_group->removed) { - struct extent_map_tree *em_tree; - struct extent_map *em; - - spin_unlock(&block_group->lock); - + if (cleanup) { lock_chunks(block_group->fs_info->chunk_root); em_tree = &block_group->fs_info->mapping_tree.map_tree; write_lock(&em_tree->lock); @@ -3324,10 +3312,31 @@ out: * this block group have left 1 entry each one. Free them. */ __btrfs_remove_free_space_cache(block_group->free_space_ctl); - } else { - spin_unlock(&block_group->lock); } +} +int btrfs_trim_block_group(struct btrfs_block_group_cache *block_group, + u64 *trimmed, u64 start, u64 end, u64 minlen) +{ + int ret; + + *trimmed = 0; + + spin_lock(&block_group->lock); + if (block_group->removed) { + spin_unlock(&block_group->lock); + return 0; + } + btrfs_get_block_group_trimming(block_group); + spin_unlock(&block_group->lock); + + ret = trim_no_bitmap(block_group, trimmed, start, end, minlen); + if (ret) + goto out; + + ret = trim_bitmaps(block_group, trimmed, start, end, minlen); +out: + btrfs_put_block_group_trimming(block_group); return ret; } diff --git a/fs/btrfs/super.c b/fs/btrfs/super.c index a1077e0ffaa8..8da24e242896 100644 --- a/fs/btrfs/super.c +++ b/fs/btrfs/super.c @@ -69,7 +69,7 @@ static struct file_system_type btrfs_fs_type; static int btrfs_remount(struct super_block *sb, int *flags, char *data); -static const char *btrfs_decode_error(int errno) +const char *btrfs_decode_error(int errno) { char *errstr = "unknown"; diff --git a/fs/btrfs/transaction.c b/fs/btrfs/transaction.c index f5021fcb154e..44da9299a25b 100644 --- a/fs/btrfs/transaction.c +++ b/fs/btrfs/transaction.c @@ -258,6 +258,8 @@ loop: mutex_init(&cur_trans->cache_write_mutex); cur_trans->num_dirty_bgs = 0; spin_lock_init(&cur_trans->dirty_bgs_lock); + INIT_LIST_HEAD(&cur_trans->deleted_bgs); + spin_lock_init(&cur_trans->deleted_bgs_lock); list_add_tail(&cur_trans->list, &fs_info->trans_list); extent_io_tree_init(&cur_trans->dirty_pages, fs_info->btree_inode->i_mapping); diff --git a/fs/btrfs/transaction.h b/fs/btrfs/transaction.h index eb09c2067fa8..edc2fbc262d7 100644 --- a/fs/btrfs/transaction.h +++ b/fs/btrfs/transaction.h @@ -74,6 +74,8 @@ struct btrfs_transaction { */ struct mutex cache_write_mutex; spinlock_t dirty_bgs_lock; + struct list_head deleted_bgs; + spinlock_t deleted_bgs_lock; struct btrfs_delayed_ref_root delayed_refs; int aborted; int dirty_bg_run;