Re: [PATCH] tracing: Fix a memory leak bug

[Steven Rostedt]

On Fri, 19 Apr 2019 21:22:59 -0500
Wenwen Wang <wang6495@xxxxxxx> wrote:

> In trace_pid_write(), the buffer for trace parser is allocated through
> kmalloc() in trace_parser_get_init(). Later on, after the buffer is used,
> it is then freed through kfree() in trace_parser_put(). However, it is
> possible that trace_pid_write() is terminated due to unexpected errors,
> e.g., ENOMEM. In that case, the allocated buffer will not be freed, which
> is a memory leak bug.
> 
> To fix this issue, free the allocated buffer when an error is encountered.

Thanks for the patch. Did you find this through manual inspection,
running KASAN or via one of the static analyzers?

-- Steve

> 
> Signed-off-by: Wenwen Wang <wang6495@xxxxxxx>
> ---
>  kernel/trace/trace.c | 5 ++++-
>  1 file changed, 4 insertions(+), 1 deletion(-)
> 
> diff --git a/kernel/trace/trace.c b/kernel/trace/trace.c
> index 6c24755..fd12c9c 100644
> --- a/kernel/trace/trace.c
> +++ b/kernel/trace/trace.c
> @@ -496,8 +496,10 @@ int trace_pid_write(struct trace_pid_list
> *filtered_pids,
>  	 * not modified.
>  	 */
>  	pid_list = kmalloc(sizeof(*pid_list), GFP_KERNEL);
> -	if (!pid_list)
> +	if (!pid_list) {
> +		trace_parser_put(&parser);
>  		return -ENOMEM;
> +	}
>  
>  	pid_list->pid_max = READ_ONCE(pid_max);
>  
> @@ -507,6 +509,7 @@ int trace_pid_write(struct trace_pid_list
> *filtered_pids, 
>  	pid_list->pids = vzalloc((pid_list->pid_max + 7) >> 3);
>  	if (!pid_list->pids) {
> +		trace_parser_put(&parser);
>  		kfree(pid_list);
>  		return -ENOMEM;
>  	}