Re: [PATCH] eventfs: fix a null pointer access in eventfs_iterate

[Steven Rostedt <rostedt@goodmis.org>]

On Thu, 29 Aug 2024 11:24:36 +0800
Chi Zhiling <chizhiling@163.com> wrote:

> From: Chi Zhiling <chizhiling@kylinos.cn>
> 
> We found a null pointer accessing in tracefs[1], the reason is that
> the variable 'ei_child' is set to LIST_POISON1, that means the list
> was removed in eventfs_remove_rec. so when access the ei_child->is_freed,
> the panic triggered.
> 
> the linked list is protected by eventfs_mutex in eventfs_remove_rec,

Only writes of the link list is protected by the mutex. Reads are not.

> so when we access the list of ei_child in eventfs_iterate, we also need
> a mutex_lock in there to avoid eventfs_remove_rec modify the list.

Yes you hit a bug, but no this is *not* the solution!


> Signed-off-by: Chi Zhiling <chizhiling@kylinos.cn>
> ---
>  fs/tracefs/event_inode.c | 6 +++++-
>  1 file changed, 5 insertions(+), 1 deletion(-)
> 
> diff --git a/fs/tracefs/event_inode.c b/fs/tracefs/event_inode.c
> index 01e99e98457d..4895ed07376b 100644
> --- a/fs/tracefs/event_inode.c
> +++ b/fs/tracefs/event_inode.c
> @@ -642,6 +642,7 @@ static int eventfs_iterate(struct file *file, struct dir_context *ctx)
>  	/* Subtract the skipped entries above */
>  	c -= min((unsigned int)c, (unsigned int)ei->nr_entries);
>  
> +	mutex_unlock(&eventfs_mutex);

This list is protected by SRCU (hence the name of the iterator), if you
need to add a mutex around it, something else is broken.

>  	list_for_each_entry_srcu(ei_child, &ei->children, list,
>  				 srcu_read_lock_held(&eventfs_srcu)) {
>  
> @@ -659,9 +660,12 @@ static int eventfs_iterate(struct file *file, struct dir_context *ctx)
>  
>  		ino = eventfs_dir_ino(ei_child);
>  
> -		if (!dir_emit(ctx, name, strlen(name), ino, DT_DIR))
> +		if (!dir_emit(ctx, name, strlen(name), ino, DT_DIR)) {
> +			mutex_unlock(&eventfs_mutex);
>  			goto out_dec;
> +		}
>  	}
> +	mutex_unlock(&eventfs_mutex);
>  	ret = 1;
>   out:
>  	srcu_read_unlock(&eventfs_srcu, idx);

The real fix is:

diff --git a/fs/tracefs/event_inode.c b/fs/tracefs/event_inode.c
index 01e99e98457d..8705c77a9e75 100644
--- a/fs/tracefs/event_inode.c
+++ b/fs/tracefs/event_inode.c
@@ -862,7 +862,7 @@ static void eventfs_remove_rec(struct eventfs_inode *ei, int level)
 	list_for_each_entry(ei_child, &ei->children, list)
 		eventfs_remove_rec(ei_child, level + 1);
 
-	list_del(&ei->list);
+	list_del_rcu(&ei->list);
 	free_ei(ei);
 }
 
Can you test that and let me know if it fixes your issue. I'll just go
ahead and apply it as it is an obvious bug.

-- Steve