From: "Steven Rostedt (Google)" rostedt@goodmis.org

When the inode is being dropped from the dentry, the TRACEFS_EVENT_INODE flag needs to be cleared to prevent a remount from calling eventfs_remount() on the tracefs_inode private data. There's a race between the inode is dropped (and the dentry freed) to where the inode is actually freed. If a remount happens between the two, the eventfs_inode could be accessed after it is freed (only the dentry keeps a ref count on it).

Currently the TRACEFS_EVENT_INODE flag is cleared from the dentry iput() function. But this is incorrect, as it is possible that the inode has another reference to it. The flag should only be cleared when the inode is really being dropped and has no more references. That happens in the drop_inode callback of the inode, as that gets called when the last reference of the inode is released.

Remove the tracefs_d_iput() function and move its logic to the more appropriate tracefs_drop_inode() callback function.

Link: https://lore.kernel.org/linux-trace-kernel/20240523051539.908205106@goodmis....

Cc: stable@vger.kernel.org Cc: Masami Hiramatsu mhiramat@kernel.org Cc: Mark Rutland mark.rutland@arm.com Cc: Mathieu Desnoyers mathieu.desnoyers@efficios.com Cc: Andrew Morton akpm@linux-foundation.org Cc: Masahiro Yamada masahiroy@kernel.org Fixes: baa23a8d4360d ("tracefs: Reset permissions on remount if permissions are options") Signed-off-by: Steven Rostedt (Google) rostedt@goodmis.org --- fs/tracefs/inode.c | 33 +++++++++++++++++---------------- 1 file changed, 17 insertions(+), 16 deletions(-)

diff --git a/fs/tracefs/inode.c b/fs/tracefs/inode.c index 9252e0d78ea2..7c29f4afc23d 100644 --- a/fs/tracefs/inode.c +++ b/fs/tracefs/inode.c @@ -426,10 +426,26 @@ static int tracefs_show_options(struct seq_file *m, struct dentry *root) return 0; }

+static int tracefs_drop_inode(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; + + return 1; +} + static const struct super_operations tracefs_super_operations = { .alloc_inode = tracefs_alloc_inode, .free_inode = tracefs_free_inode, - .drop_inode = generic_delete_inode, + .drop_inode = tracefs_drop_inode, .statfs = simple_statfs, .show_options = tracefs_show_options, }; @@ -455,22 +471,7 @@ 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, };