From: David Sterba <dsterba@suse.cz>
To: Filipe Manana <fdmanana@kernel.org>
Cc: Yangtao Li <frank.li@vivo.com>,
clm@fb.com, josef@toxicpanda.com, dsterba@suse.com,
linux-btrfs@vger.kernel.org, linux-kernel@vger.kernel.org
Subject: Re: [PATCH 4/4] btrfs: Fix transaction abort during failure in del_balance_item()
Date: Wed, 9 Apr 2025 01:04:13 +0200 [thread overview]
Message-ID: <20250408230413.GE13292@twin.jikos.cz> (raw)
In-Reply-To: <CAL3q7H7BS6juCS0eRdo6sqM4jzeMMi1o=huG38wgKYumD7qmmw@mail.gmail.com>
On Tue, Apr 08, 2025 at 03:53:04PM +0100, Filipe Manana wrote:
> On Tue, Apr 8, 2025 at 1:19 PM Yangtao Li <frank.li@vivo.com> wrote:
> >
> > Handle errors by adding explicit btrfs_abort_transaction
> > and btrfs_end_transaction calls.
>
> Again, like in the previous patch, why?
> This provides no reason at all why we should abort.
> And the same comment below.
>
> >
> > Signed-off-by: Yangtao Li <frank.li@vivo.com>
> > ---
> > fs/btrfs/volumes.c | 11 +++++++----
> > 1 file changed, 7 insertions(+), 4 deletions(-)
> >
> > diff --git a/fs/btrfs/volumes.c b/fs/btrfs/volumes.c
> > index 347c475028e0..23739d18d833 100644
> > --- a/fs/btrfs/volumes.c
> > +++ b/fs/btrfs/volumes.c
> > @@ -3777,7 +3777,7 @@ static int del_balance_item(struct btrfs_fs_info *fs_info)
> > struct btrfs_trans_handle *trans;
> > BTRFS_PATH_AUTO_FREE(path);
> > struct btrfs_key key;
> > - int ret, err;
> > + int ret;
> >
> > path = btrfs_alloc_path();
> > if (!path)
> > @@ -3800,10 +3800,13 @@ static int del_balance_item(struct btrfs_fs_info *fs_info)
> > }
> >
> > ret = btrfs_del_item(trans, root, path);
> > + if (ret)
> > + goto out;
> > +
> > + return btrfs_commit_transaction(trans);
> > out:
> > - err = btrfs_commit_transaction(trans);
> > - if (err && !ret)
> > - ret = err;
> > + btrfs_abort_transaction(trans, ret);
> > + btrfs_end_transaction(trans);
>
> A transaction abort will turn the fs into RO mode, and it's meant to
> be used when we can't proceed with changes to the fs after we did
> partial changes, to avoid leaving things in an inconsistent state.
> Here we don't abort because we haven't done any changes before using
> the transaction handle, so an abort is pointless and will turn the fs
> into RO mode unnecessarily.
The del_balance_item() case seems to be unique, there's only one caller
reset_balance_state() that calls btrfs_handle_fs_error() in case of an
error. This is almost the same as a transaction abort, but
del_balance_item() may be called from various contexts.
The error handling may be improved here, e.g. some callers may care
about the actual error of del_balance_item/reset_balance_state, but
rather a hard transaction abort it could be better to handle it more
gracefully for operations that are restartable, like return an EAGAIN.
next prev parent reply other threads:[~2025-04-08 23:04 UTC|newest]
Thread overview: 18+ messages / expand[flat|nested] mbox.gz Atom feed top
2025-04-08 12:29 [PATCH 1/4] btrfs: use BTRFS_PATH_AUTO_FREE in insert_balance_item() Yangtao Li
2025-04-08 12:29 ` [PATCH 2/4] btrfs: use BTRFS_PATH_AUTO_FREE in del_balance_item() Yangtao Li
2025-04-08 14:47 ` Filipe Manana
2025-04-08 12:29 ` [PATCH 3/4] btrfs: Fix transaction abort during failure in insert_balance_item() Yangtao Li
2025-04-08 14:50 ` Filipe Manana
2025-04-08 12:29 ` [PATCH 4/4] btrfs: Fix transaction abort during failure in del_balance_item() Yangtao Li
2025-04-08 14:53 ` Filipe Manana
2025-04-08 23:04 ` David Sterba [this message]
2025-04-09 9:50 ` Filipe Manana
2025-04-08 14:44 ` [PATCH 1/4] btrfs: use BTRFS_PATH_AUTO_FREE in insert_balance_item() Filipe Manana
2025-04-08 17:36 ` Daniel Vacek
2025-04-08 17:40 ` Filipe Manana
2025-04-08 17:47 ` Daniel Vacek
2025-04-08 17:52 ` Filipe Manana
2025-04-08 18:00 ` Daniel Vacek
2025-04-09 12:19 ` 回复: " 李扬韬
2025-04-08 22:47 ` David Sterba
2025-04-09 9:45 ` Filipe Manana
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=20250408230413.GE13292@twin.jikos.cz \
--to=dsterba@suse.cz \
--cc=clm@fb.com \
--cc=dsterba@suse.com \
--cc=fdmanana@kernel.org \
--cc=frank.li@vivo.com \
--cc=josef@toxicpanda.com \
--cc=linux-btrfs@vger.kernel.org \
--cc=linux-kernel@vger.kernel.org \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox
all inboxes | Powered by JetHome®