Message ID | d794156bd3368d635913610dbe03c1fc727e297f.1621523846.git.josef@toxicpanda.com (mailing list archive) |
---|---|
State | New, archived |
Headers | show |
Series | Rework fs error handling | expand |
On Thu, May 20, 2021 at 11:21:31AM -0400, Josef Bacik wrote: > While stress testing our error handling I noticed that sometimes we > would still commit the transaction even though we had aborted the > transaction. > > Currently we track if a trans handle has dirtied any metadata, and if it > hasn't we mark the FS as having an error (so no new transactions can be > started), but we will allow the current transaction to complete as we do > not mark the transaction itself as having been aborted. > > This sounds good in theory, but we were not properly tracking IO errors > in btrfs_finish_ordered_io, and thus committing the transaction with > bogus free space data. This isn't necessarily a problem per-se with the > free space cache, as the other guards in place would have kept us from > accepting the free space cache as valid, but hi-lights a real world case > where we had a bug and could have corrupted the filesystem because of > it. > > This "skip abort on empty trans handle" is nice in theory, but assumes > we have perfect error handling everywhere, which we clearly do not. > Also we do not allow further transactions to be started, so all this > does is save the last transaction that was happening, which doesn't > necessarily gain us anything other than the potential for real > corruption. > > Remove this particular bit of code, if we decide we need to abort the > transaction then abort the current one and keep us from doing real harm > to the file system, regardless of whether this specific trans handle > dirtied anything or not. > > Signed-off-by: Josef Bacik <josef@toxicpanda.com> I've checked logs what would be the effects of leaving out the message printed by the unused transaction. Getting rid of it sounds like an improvement as it's not adding any information, the first transaction is noisy and that's where the problem happens. Additional messages about the abort are confusing, so yeah. Besides, the updates to trans->dirty lack any serialization so it's quite unreliable anyway.
diff --git a/fs/btrfs/ctree.c b/fs/btrfs/ctree.c index a484fb72a01f..4bc3ca2cbd7d 100644 --- a/fs/btrfs/ctree.c +++ b/fs/btrfs/ctree.c @@ -596,7 +596,6 @@ noinline int btrfs_cow_block(struct btrfs_trans_handle *trans, trans->transid, fs_info->generation); if (!should_cow_block(trans, root, buf)) { - trans->dirty = true; *cow_ret = buf; return 0; } @@ -1788,10 +1787,8 @@ int btrfs_search_slot(struct btrfs_trans_handle *trans, struct btrfs_root *root, * then we don't want to set the path blocking, * so we test it here */ - if (!should_cow_block(trans, root, b)) { - trans->dirty = true; + if (!should_cow_block(trans, root, b)) goto cow_done; - } /* * must have write locks on this node and the diff --git a/fs/btrfs/extent-tree.c b/fs/btrfs/extent-tree.c index 24b5e54935a9..790de24576ac 100644 --- a/fs/btrfs/extent-tree.c +++ b/fs/btrfs/extent-tree.c @@ -4784,7 +4784,6 @@ btrfs_init_new_buffer(struct btrfs_trans_handle *trans, struct btrfs_root *root, set_extent_dirty(&trans->transaction->dirty_pages, buf->start, buf->start + buf->len - 1, GFP_NOFS); } - trans->dirty = true; /* this returns a buffer locked for blocking */ return buf; } diff --git a/fs/btrfs/super.c b/fs/btrfs/super.c index 4a396c1147f1..bc613218c8c5 100644 --- a/fs/btrfs/super.c +++ b/fs/btrfs/super.c @@ -299,17 +299,6 @@ void __btrfs_abort_transaction(struct btrfs_trans_handle *trans, struct btrfs_fs_info *fs_info = trans->fs_info; WRITE_ONCE(trans->aborted, errno); - /* Nothing used. The other threads that have joined this - * transaction may be able to continue. */ - if (!trans->dirty && list_empty(&trans->new_bgs)) { - const char *errstr; - - errstr = btrfs_decode_error(errno); - btrfs_warn(fs_info, - "%s:%d: Aborting unused transaction(%s).", - function, line, errstr); - return; - } WRITE_ONCE(trans->transaction->aborted, errno); /* Wake up anybody who may be waiting on this transaction */ wake_up(&fs_info->transaction_wait); diff --git a/fs/btrfs/transaction.c b/fs/btrfs/transaction.c index f75de9f6c0ad..e0a82aa7da89 100644 --- a/fs/btrfs/transaction.c +++ b/fs/btrfs/transaction.c @@ -2074,14 +2074,6 @@ int btrfs_commit_transaction(struct btrfs_trans_handle *trans) ASSERT(refcount_read(&trans->use_count) == 1); - /* - * Some places just start a transaction to commit it. We need to make - * sure that if this commit fails that the abort code actually marks the - * transaction as failed, so set trans->dirty to make the abort code do - * the right thing. - */ - trans->dirty = true; - /* Stop the commit early if ->aborted is set */ if (TRANS_ABORTED(cur_trans)) { ret = cur_trans->aborted; diff --git a/fs/btrfs/transaction.h b/fs/btrfs/transaction.h index 364cfbb4c5c5..c49e2266b28b 100644 --- a/fs/btrfs/transaction.h +++ b/fs/btrfs/transaction.h @@ -143,7 +143,6 @@ struct btrfs_trans_handle { bool allocating_chunk; bool can_flush_pending_bgs; bool reloc_reserved; - bool dirty; bool in_fsync; struct btrfs_root *root; struct btrfs_fs_info *fs_info;
While stress testing our error handling I noticed that sometimes we would still commit the transaction even though we had aborted the transaction. Currently we track if a trans handle has dirtied any metadata, and if it hasn't we mark the FS as having an error (so no new transactions can be started), but we will allow the current transaction to complete as we do not mark the transaction itself as having been aborted. This sounds good in theory, but we were not properly tracking IO errors in btrfs_finish_ordered_io, and thus committing the transaction with bogus free space data. This isn't necessarily a problem per-se with the free space cache, as the other guards in place would have kept us from accepting the free space cache as valid, but hi-lights a real world case where we had a bug and could have corrupted the filesystem because of it. This "skip abort on empty trans handle" is nice in theory, but assumes we have perfect error handling everywhere, which we clearly do not. Also we do not allow further transactions to be started, so all this does is save the last transaction that was happening, which doesn't necessarily gain us anything other than the potential for real corruption. Remove this particular bit of code, if we decide we need to abort the transaction then abort the current one and keep us from doing real harm to the file system, regardless of whether this specific trans handle dirtied anything or not. Signed-off-by: Josef Bacik <josef@toxicpanda.com> --- fs/btrfs/ctree.c | 5 +---- fs/btrfs/extent-tree.c | 1 - fs/btrfs/super.c | 11 ----------- fs/btrfs/transaction.c | 8 -------- fs/btrfs/transaction.h | 1 - 5 files changed, 1 insertion(+), 25 deletions(-)