On Tue, 2007-08-07 at 17:17 -0500, Serge E. Hallyn wrote:

quoted text
> diff --git a/fs/splice.c b/fs/splice.c
> index e36c003..2df95f3 100644
> --- a/fs/splice.c
> +++ b/fs/splice.c
> @@ -827,6 +827,12 @@ generic_file_splice_write(struct pipe_inode_info *pipe, struct file *out,
>  	ssize_t ret;
>  	int err;
>  
> +	mutex_lock(&inode->i_mutex);
> +	err = security_inode_killpriv(out->f_path.dentry);
> +	mutex_unlock(&inode->i_mutex);
> +	if (err)
> +		return err;

You are unconditionally taking the i_mutex whether or not you actually
have any capabilities to remove. Normally, removing capabilities due to
a write is something which occurs once every blue moon. Can't you
introduce a heuristic along the lines of should_remove_suid() in order
to optimise away the common case?

In addition, if you need to remove both the capabilities and the suid
bits, then it should be unnecessary to take the i_mutex twice.

quoted text
> +
>  	err = should_remove_suid(out->f_path.dentry);
>  	if (unlikely(err)) {
>  		mutex_lock(&inode->i_mutex);

Trond

-
unsubscribe notice
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/