tracefs: Remove unneeded buggy tracefs iput callback

Commit Message

Steven Rostedt May 22, 2024, 4:45 p.m. UTC

From: "Steven Rostedt (Google)" <[e-mail]>

The iput callback was added because the remount could call into the
eventfs code and touch the ei->entry_attrs array, which could have been
freed when an eventfs directory is freed (via a synthetic event). But the
entry_attrs was freed incorrectly and since been fixed to be freed after
the last reference of the ei is done.

The iput clears the TRACEFS_EVENT_INODE flag of the tracefs_inode
preventing it from calling the eventfs_remount() function. But the iput
can be called after the last reference to the inode is done but the
eventfs_inode still exists, causing the eventfs_remount() not to be called
on an tracefs_inode when it should be.

Link: https://lore.kernel.org/all/[e-mail]/

Cc: [e-mail]
Reported-by: Masahiro Yamada <[e-mail]>
Fixes: ee4e0379475e4 ("eventfs: Free all of the eventfs_inode after RCU")
Signed-off-by: Steven Rostedt (Google) <[e-mail]>
---
 fs/tracefs/inode.c | 15 ---------------
 1 file changed, 15 deletions(-)

Comments

Steven Rostedt May 22, 2024, 6:05 p.m. UTC | #1

On Wed, 22 May 2024 12:45:04 -0400
Steven Rostedt <[e-mail]> wrote:

> From: "Steven Rostedt (Google)" <[e-mail]>
> 
> The iput callback was added because the remount could call into the
> eventfs code and touch the ei->entry_attrs array, which could have been
> freed when an eventfs directory is freed (via a synthetic event). But the
> entry_attrs was freed incorrectly and since been fixed to be freed after
> the last reference of the ei is done.
> 
> The iput clears the TRACEFS_EVENT_INODE flag of the tracefs_inode
> preventing it from calling the eventfs_remount() function. But the iput
> can be called after the last reference to the inode is done but the
> eventfs_inode still exists, causing the eventfs_remount() not to be called
> on an tracefs_inode when it should be.

Testing this more, I found that the iput is still needed, as the deletion
of the eventfs inodes can happen before the inode is released.

Will produce a v2 that handles this properly.

-- Steve

Patch

diff --git a/fs/tracefs/inode.c b/fs/tracefs/inode.c
index 9252e0d78ea2..62ca9c23b93c 100644
--- a/fs/tracefs/inode.c
+++ b/fs/tracefs/inode.c
@@ -455,22 +455,7 @@  static int tracefs_d_revalidate(struct dentry *dentry, unsigned int flags)
 	return !(ei && ei->is_freed);
 }
 
-static void tracefs_d_iput(struct dentry *dentry, struct inode *inode)
-{
-	struct tracefs_inode *ti = get_tracefs(inode);
-
-	/*
-	 * This inode is being freed and cannot be used for
-	 * eventfs. Clear the flag so that it doesn't call into
-	 * eventfs during the remount flag updates. The eventfs_inode
-	 * gets freed after an RCU cycle, so the content will still
-	 * be safe if the iteration is going on now.
-	 */
-	ti->flags &= ~TRACEFS_EVENT_INODE;
-}
-
 static const struct dentry_operations tracefs_dentry_operations = {
-	.d_iput = tracefs_d_iput,
 	.d_revalidate = tracefs_d_revalidate,
 	.d_release = tracefs_d_release,
 };