linux-ext4 - [PATCHv2 REBASED 0/2] jbd2: Kill t_handle

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [thread-next>] [day] [month] [year] [list]

Date:   Wed, 16 Feb 2022 12:30:34 +0530
From:   Ritesh Harjani <riteshh@...ux.ibm.com>
To:     linux-ext4@...r.kernel.org
Cc:     Jan Kara <jack@...e.cz>, "Theodore Ts'o" <tytso@....edu>,
        Andreas Dilger <adilger.kernel@...ger.ca>,
        linux-fsdevel@...r.kernel.org,
        Ritesh Harjani <riteshh@...ux.ibm.com>
Subject: [PATCHv2 REBASED 0/2] jbd2: Kill t_handle_lock transaction spinlock

Hello,

This is mainly just rebasing the patch series on top of recent use-after-free
fix submitted [4], due to conflict in function jbd2_journal_wait_updates().
No other changes apart from that.

Testing
========
I have again tested xfstests with -g log,metadata,auto group with 4k bs
config (COFIG_KASAN disabled). I haven't found any regression due to these
patches in my testing.

Original cover letter
======================
This small patch series kills the age-old t_handle_lock transaction spinlock,
which on careful inspection, came out to be not very useful anymore.
At some of the places it isn't required at all now and for the rest
(e.g. update_t_max_wait()), we could make use of atomic cmpxchg to make the
code path lockless.

This was tested with fstests with -g quick and -g log on my qemu setup.
I had also done some extensive fsmark testing to see that we don't see any
bottleneck resulting from removal of CONFIG_JBD2_DEBUG to update t_max_wait
in patch-2. None of my test showed any bottleneck.

Note that there had been several patches in the past over time which had led to
t_handle_lock becoming obselete now e.g. [1-2]
In this work, couple of the code paths to remove this spinlock were observed
while doing code review and to get completely rid of it was something which was
suggested by Jan [3].
Thanks to Jan for thorough review and suggestions :)

References
===========
[v1]: https://lore.kernel.org/all/cover.1642953021.git.riteshh@linux.ibm.com/
[1]: https://lore.kernel.org/linux-ext4/1280939957-3277-4-git-send-email-tytso@mit.edu/
[2]: https://lore.kernel.org/linux-ext4/20120103153245.GE31457@quack.suse.cz/
[3]: https://lore.kernel.org/linux-ext4/20220113112749.d5tfszcksvxvshnn@quack3.lan/
[4]: https://lore.kernel.org/all/948c2fed518ae739db6a8f7f83f1d58b504f87d0.1644497105.git.ritesh.list@gmail.com/

Ritesh Harjani (2):
  jbd2: Kill t_handle_lock transaction spinlock
  jbd2: Remove CONFIG_JBD2_DEBUG to update t_max_wait

 fs/jbd2/transaction.c | 35 ++++++++++++-----------------------
 include/linux/jbd2.h  |  3 ---
 2 files changed, 12 insertions(+), 26 deletions(-)

--
2.31.1