[<prev] [next>] [thread-next>] [day] [month] [year] [list]
Message-Id: <20250812095739.3194337-1-amarkuze@redhat.com>
Date: Tue, 12 Aug 2025 09:57:37 +0000
From: Alex Markuze <amarkuze@...hat.com>
To: ceph-devel@...r.kernel.org
Cc: linux-kernel@...r.kernel.org,
Slava.Dubeyko@....com,
idryomov@...il.com,
Alex Markuze <amarkuze@...hat.com>
Subject: [PATCH v4 0/2] ceph: fix client race conditions with stale r_parent
This patch series addresses client-side race conditions in the Ceph filesystem
where the cached parent directory inode (r_parent) can become stale during
concurrent operations like rename, leading to incorrect state application.
The first patch adds validation during reply processing to ensure the cached
parent directory inode matches the directory info in MDS replies. It refactors
the path building API to use a structured approach and prevents applying state
changes to incorrect directory inodes.
The second patch addresses cases where r_parent becomes stale between request
creation and message sending when the parent directory's i_rwsem is not locked.
It validates that r_parent matches the encoded parent inode and updates to the
correct inode if a mismatch is detected, with appropriate warnings for this
Alex Markuze (2):
ceph: fix client race condition validating r_parent before applying
state
ceph: fix client race condition where r_parent becomes stale before
sending message
fs/ceph/debugfs.c | 14 ++--
fs/ceph/dir.c | 17 ++---
fs/ceph/file.c | 24 +++----
fs/ceph/inode.c | 59 ++++++++++++++--
fs/ceph/mds_client.c | 165 ++++++++++++++++++++++++++-----------------
fs/ceph/mds_client.h | 18 +++--
6 files changed, 189 insertions(+), 108 deletions(-)
--
2.34.1
Powered by blists - more mailing lists