diff options
| author | Linus Torvalds <torvalds@linux-foundation.org> | 2022-03-24 10:06:43 -0700 |
|---|---|---|
| committer | Linus Torvalds <torvalds@linux-foundation.org> | 2022-03-24 10:06:43 -0700 |
| commit | 2e2d4650b34ffe0a39f70acc9429a58d94e39236 (patch) | |
| tree | 547082f61a28eda2ddf2ddd71daeaa276fc05e5f | |
| parent | 15f2e3d6c1f713050d2d51f822fc4253f67ae7ac (diff) | |
| parent | 0f8821da48458982cf379eb4432f23958f2e3a6c (diff) | |
Merge tag 'fs.rt.v5.18' of git://git.kernel.org/pub/scm/linux/kernel/git/brauner/linux
Pull mount attributes PREEMPT_RT update from Christian Brauner:
"This contains Sebastian's fix to make changing mount
attributes/getting write access compatible with CONFIG_PREEMPT_RT.
The change only applies when users explicitly opt-in to real-time via
CONFIG_PREEMPT_RT otherwise things are exactly as before. We've waited
quite a long time with this to make sure folks could take a good look"
* tag 'fs.rt.v5.18' of git://git.kernel.org/pub/scm/linux/kernel/git/brauner/linux:
fs/namespace: Boost the mount_lock.lock owner instead of spinning on PREEMPT_RT.
| -rw-r--r-- | fs/namespace.c | 20 |
1 files changed, 18 insertions, 2 deletions
diff --git a/fs/namespace.c b/fs/namespace.c index 627db2e031e9..6e9844b8c6fb 100644 --- a/fs/namespace.c +++ b/fs/namespace.c @@ -344,8 +344,24 @@ int __mnt_want_write(struct vfsmount *m) * incremented count after it has set MNT_WRITE_HOLD. */ smp_mb(); - while (READ_ONCE(mnt->mnt.mnt_flags) & MNT_WRITE_HOLD) - cpu_relax(); + might_lock(&mount_lock.lock); + while (READ_ONCE(mnt->mnt.mnt_flags) & MNT_WRITE_HOLD) { + if (!IS_ENABLED(CONFIG_PREEMPT_RT)) { + cpu_relax(); + } else { + /* + * This prevents priority inversion, if the task + * setting MNT_WRITE_HOLD got preempted on a remote + * CPU, and it prevents life lock if the task setting + * MNT_WRITE_HOLD has a lower priority and is bound to + * the same CPU as the task that is spinning here. + */ + preempt_enable(); + lock_mount_hash(); + unlock_mount_hash(); + preempt_disable(); + } + } /* * After the slowpath clears MNT_WRITE_HOLD, mnt_is_readonly will * be set to match its requirements. So we must not load that until |