From 95ca6599a589ee84c69f02d0e1d928c8d1367fb1 Mon Sep 17 00:00:00 2001 From: Naohiro Aota Date: Thu, 29 Jun 2023 17:37:31 +0900 Subject: [PATCH 1/5] btrfs: zoned: do not enable async discard The zoned mode need to reset a zone before using it. We rely on btrfs's original discard functionality (discarding unused block group range) to do the resetting. While the commit 63a7cb130718 ("btrfs: auto enable discard=async when possible") made the discard done in an async manner, a zoned reset do not need to be async, as it is fast enough. Even worth, delaying zone rests prevents using those zones again. So, let's disable async discard on the zoned mode. Fixes: 63a7cb130718 ("btrfs: auto enable discard=async when possible") CC: stable@vger.kernel.org # 6.3+ Reviewed-by: Damien Le Moal Reviewed-by: Christoph Hellwig Signed-off-by: Naohiro Aota Reviewed-by: David Sterba [ update message text ] Signed-off-by: David Sterba --- fs/btrfs/disk-io.c | 7 ++++++- fs/btrfs/zoned.c | 3 +++ 2 files changed, 9 insertions(+), 1 deletion(-) diff --git a/fs/btrfs/disk-io.c b/fs/btrfs/disk-io.c index 7513388b0567..9b9914e5f03d 100644 --- a/fs/btrfs/disk-io.c +++ b/fs/btrfs/disk-io.c @@ -3438,11 +3438,16 @@ int __cold open_ctree(struct super_block *sb, struct btrfs_fs_devices *fs_device * For devices supporting discard turn on discard=async automatically, * unless it's already set or disabled. This could be turned off by * nodiscard for the same mount. + * + * The zoned mode piggy backs on the discard functionality for + * resetting a zone. There is no reason to delay the zone reset as it is + * fast enough. So, do not enable async discard for zoned mode. */ if (!(btrfs_test_opt(fs_info, DISCARD_SYNC) || btrfs_test_opt(fs_info, DISCARD_ASYNC) || btrfs_test_opt(fs_info, NODISCARD)) && - fs_info->fs_devices->discardable) { + fs_info->fs_devices->discardable && + !btrfs_is_zoned(fs_info)) { btrfs_set_and_info(fs_info, DISCARD_ASYNC, "auto enabling async discard"); } diff --git a/fs/btrfs/zoned.c b/fs/btrfs/zoned.c index 85b8b332add9..72b90bc19a19 100644 --- a/fs/btrfs/zoned.c +++ b/fs/btrfs/zoned.c @@ -805,6 +805,9 @@ int btrfs_check_mountopts_zoned(struct btrfs_fs_info *info) return -EINVAL; } + btrfs_clear_and_info(info, DISCARD_ASYNC, + "zoned: async discard ignored and disabled for zoned mode"); + return 0; } From 8dbfc14fc736eb701089aff09645c3d4ad3decb1 Mon Sep 17 00:00:00 2001 From: Filipe Manana Date: Thu, 20 Jul 2023 12:44:33 +0100 Subject: [PATCH 2/5] btrfs: account block group tree when calculating global reserve size When using the block group tree feature, this tree is a critical tree just like the extent, csum and free space trees, and just like them it uses the delayed refs block reserve. So take into account the block group tree, and its current size, when calculating the size for the global reserve. CC: stable@vger.kernel.org # 6.1+ Signed-off-by: Filipe Manana Reviewed-by: David Sterba Signed-off-by: David Sterba --- fs/btrfs/block-rsv.c | 5 +++++ 1 file changed, 5 insertions(+) diff --git a/fs/btrfs/block-rsv.c b/fs/btrfs/block-rsv.c index 6279d200cf83..77684c5e0c8b 100644 --- a/fs/btrfs/block-rsv.c +++ b/fs/btrfs/block-rsv.c @@ -349,6 +349,11 @@ void btrfs_update_global_block_rsv(struct btrfs_fs_info *fs_info) } read_unlock(&fs_info->global_root_lock); + if (btrfs_fs_compat_ro(fs_info, BLOCK_GROUP_TREE)) { + num_bytes += btrfs_root_used(&fs_info->block_group_root->root_item); + min_items++; + } + /* * But we also want to reserve enough space so we can do the fallback * global reserve for an unlink, which is an additional From d8ccbd21918fd7fa6ce3226cffc22c444228e8ad Mon Sep 17 00:00:00 2001 From: Filipe Manana Date: Fri, 30 Jun 2023 16:03:44 +0100 Subject: [PATCH 3/5] btrfs: remove BUG_ON()'s in add_new_free_space() At add_new_free_space() we have these BUG_ON()'s that are there to deal with any failure to add free space to the in memory free space cache. Such failures are mostly -ENOMEM that should be very rare. However there's no need to have these BUG_ON()'s, we can just return any error to the caller and all callers and their upper call chain are already dealing with errors. So just make add_new_free_space() return any errors, while removing the BUG_ON()'s, and returning the total amount of added free space to an optional u64 pointer argument. Reported-by: syzbot+3ba856e07b7127889d8c@syzkaller.appspotmail.com Link: https://lore.kernel.org/linux-btrfs/000000000000e9cb8305ff4e8327@google.com/ Signed-off-by: Filipe Manana Reviewed-by: David Sterba Signed-off-by: David Sterba --- fs/btrfs/block-group.c | 51 +++++++++++++++++++++++++------------- fs/btrfs/block-group.h | 4 +-- fs/btrfs/free-space-tree.c | 24 ++++++++++++------ 3 files changed, 53 insertions(+), 26 deletions(-) diff --git a/fs/btrfs/block-group.c b/fs/btrfs/block-group.c index b7da14848686..63c3b7172ba5 100644 --- a/fs/btrfs/block-group.c +++ b/fs/btrfs/block-group.c @@ -499,12 +499,16 @@ 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, @@ -517,10 +521,12 @@ 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; @@ -529,13 +535,15 @@ 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; } /* @@ -779,8 +787,13 @@ 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; @@ -795,11 +808,10 @@ next: } 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; @@ -2294,9 +2306,11 @@ 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); @@ -2740,9 +2754,12 @@ 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 diff --git a/fs/btrfs/block-group.h b/fs/btrfs/block-group.h index 381c54a56417..aba5dff66c19 100644 --- a/fs/btrfs/block-group.h +++ b/fs/btrfs/block-group.h @@ -289,8 +289,8 @@ int btrfs_cache_block_group(struct btrfs_block_group *cache, bool wait); void btrfs_put_caching_control(struct btrfs_caching_control *ctl); struct btrfs_caching_control *btrfs_get_caching_control( struct btrfs_block_group *cache); -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_trans_handle *btrfs_start_trans_remove_block_group( struct btrfs_fs_info *fs_info, const u64 chunk_offset); diff --git a/fs/btrfs/free-space-tree.c b/fs/btrfs/free-space-tree.c index 045ddce32eca..f169378e2ca6 100644 --- a/fs/btrfs/free-space-tree.c +++ b/fs/btrfs/free-space-tree.c @@ -1515,9 +1515,13 @@ static int load_free_space_bitmaps(struct btrfs_caching_control *caching_ctl, if (prev_bit == 0 && bit == 1) { extent_start = offset; } else if (prev_bit == 1 && bit == 0) { - total_found += add_new_free_space(block_group, - extent_start, - offset); + u64 space_added; + + ret = add_new_free_space(block_group, extent_start, + offset, &space_added); + if (ret) + goto out; + total_found += space_added; if (total_found > CACHING_CTL_WAKE_UP) { total_found = 0; wake_up(&caching_ctl->wait); @@ -1529,8 +1533,9 @@ static int load_free_space_bitmaps(struct btrfs_caching_control *caching_ctl, } } if (prev_bit == 1) { - total_found += add_new_free_space(block_group, extent_start, - end); + ret = add_new_free_space(block_group, extent_start, end, NULL); + if (ret) + goto out; extent_count++; } @@ -1569,6 +1574,8 @@ static int load_free_space_extents(struct btrfs_caching_control *caching_ctl, end = block_group->start + block_group->length; while (1) { + u64 space_added; + ret = btrfs_next_item(root, path); if (ret < 0) goto out; @@ -1583,8 +1590,11 @@ static int load_free_space_extents(struct btrfs_caching_control *caching_ctl, ASSERT(key.type == BTRFS_FREE_SPACE_EXTENT_KEY); ASSERT(key.objectid < end && key.objectid + key.offset <= end); - total_found += add_new_free_space(block_group, key.objectid, - key.objectid + key.offset); + ret = add_new_free_space(block_group, key.objectid, + key.objectid + key.offset, &space_added); + if (ret) + goto out; + total_found += space_added; if (total_found > CACHING_CTL_WAKE_UP) { total_found = 0; wake_up(&caching_ctl->wait); From bf7ecbe9875061bf3fce1883e3b26b77f847d1e8 Mon Sep 17 00:00:00 2001 From: Filipe Manana Date: Fri, 21 Jul 2023 10:49:20 +0100 Subject: [PATCH 4/5] btrfs: check if the transaction was aborted at btrfs_wait_for_commit() At btrfs_wait_for_commit() we wait for a transaction to finish and then always return 0 (success) without checking if it was aborted, in which case the transaction didn't happen due to some critical error. Fix this by checking if the transaction was aborted. Fixes: 462045928bda ("Btrfs: add START_SYNC, WAIT_SYNC ioctls") CC: stable@vger.kernel.org # 4.19+ Reviewed-by: Qu Wenruo Signed-off-by: Filipe Manana Reviewed-by: David Sterba Signed-off-by: David Sterba --- fs/btrfs/transaction.c | 1 + 1 file changed, 1 insertion(+) diff --git a/fs/btrfs/transaction.c b/fs/btrfs/transaction.c index cf306351b148..f11d803b2c4e 100644 --- a/fs/btrfs/transaction.c +++ b/fs/btrfs/transaction.c @@ -931,6 +931,7 @@ int btrfs_wait_for_commit(struct btrfs_fs_info *fs_info, u64 transid) } wait_for_commit(cur_trans, TRANS_STATE_COMPLETED); + ret = cur_trans->aborted; btrfs_put_transaction(cur_trans); out: return ret; From b28ff3a7d7e97456fd86b68d24caa32e1cfa7064 Mon Sep 17 00:00:00 2001 From: Filipe Manana Date: Fri, 21 Jul 2023 10:49:21 +0100 Subject: [PATCH 5/5] btrfs: check for commit error at btrfs_attach_transaction_barrier() btrfs_attach_transaction_barrier() is used to get a handle pointing to the current running transaction if the transaction has not started its commit yet (its state is < TRANS_STATE_COMMIT_START). If the transaction commit has started, then we wait for the transaction to commit and finish before returning - however we completely ignore if the transaction was aborted due to some error during its commit, we simply return ERR_PT(-ENOENT), which makes the caller assume everything is fine and no errors happened. This could make an fsync return success (0) to user space when in fact we had a transaction abort and the target inode changes were therefore not persisted. Fix this by checking for the return value from btrfs_wait_for_commit(), and if it returned an error, return it back to the caller. Fixes: d4edf39bd5db ("Btrfs: fix uncompleted transaction") CC: stable@vger.kernel.org # 4.19+ Reviewed-by: Qu Wenruo Signed-off-by: Filipe Manana Reviewed-by: David Sterba Signed-off-by: David Sterba --- fs/btrfs/transaction.c | 9 +++++++-- 1 file changed, 7 insertions(+), 2 deletions(-) diff --git a/fs/btrfs/transaction.c b/fs/btrfs/transaction.c index f11d803b2c4e..91b6c2fdc420 100644 --- a/fs/btrfs/transaction.c +++ b/fs/btrfs/transaction.c @@ -826,8 +826,13 @@ btrfs_attach_transaction_barrier(struct btrfs_root *root) trans = start_transaction(root, 0, TRANS_ATTACH, BTRFS_RESERVE_NO_FLUSH, true); - if (trans == ERR_PTR(-ENOENT)) - btrfs_wait_for_commit(root->fs_info, 0); + if (trans == ERR_PTR(-ENOENT)) { + int ret; + + ret = btrfs_wait_for_commit(root->fs_info, 0); + if (ret) + return ERR_PTR(ret); + } return trans; }