Re: [PATCH] fanotify: Make sure to check event_len when copying

From: Amir Goldstein
Date: Wed Dec 05 2018 - 02:16:02 EST


On Wed, Dec 5, 2018 at 1:44 AM Kees Cook <keescook@xxxxxxxxxxxx> wrote:
>
> As a precaution, make sure we check event_len when copying to userspace.
> Based on old feedback: https://lkml.kernel.org/r/542D9FE5.3010009@xxxxxx
>

This precaution serves the new fanotify_event_info_fid format well:
https://marc.info/?l=linux-fsdevel&m=154375073318106&w=2
but also conflicts with my series.

I can weave your patch into my series and add a similar sanity check in
copy_fid_to_user() or if Jan applies your patch and has more comments
on my v4 series, I'll rebase on top of that.

Thanks,
Amir.

> Signed-off-by: Kees Cook <keescook@xxxxxxxxxxxx>
> ---
> fs/notify/fanotify/fanotify_user.c | 10 ++++++++--
> 1 file changed, 8 insertions(+), 2 deletions(-)
>
> diff --git a/fs/notify/fanotify/fanotify_user.c b/fs/notify/fanotify/fanotify_user.c
> index e03be5071362..d9484a0ac6b3 100644
> --- a/fs/notify/fanotify/fanotify_user.c
> +++ b/fs/notify/fanotify/fanotify_user.c
> @@ -206,7 +206,7 @@ static int process_access_response(struct fsnotify_group *group,
>
> static ssize_t copy_event_to_user(struct fsnotify_group *group,
> struct fsnotify_event *event,
> - char __user *buf)
> + char __user *buf, size_t count)
> {
> struct fanotify_event_metadata fanotify_event_metadata;
> struct file *f;
> @@ -220,6 +220,12 @@ static ssize_t copy_event_to_user(struct fsnotify_group *group,
>
> fd = fanotify_event_metadata.fd;
> ret = -EFAULT;
> + /*
> + * Sanity check copy size in case get_one_event() and
> + * fill_event_metadata() event_len sizes ever get out of sync.
> + */
> + if (WARN_ON_ONCE(fanotify_event_metadata.event_len > count))
> + goto out_close_fd;
> if (copy_to_user(buf, &fanotify_event_metadata,
> fanotify_event_metadata.event_len))
> goto out_close_fd;
> @@ -295,7 +301,7 @@ static ssize_t fanotify_read(struct file *file, char __user *buf,
> continue;
> }
>
> - ret = copy_event_to_user(group, kevent, buf);
> + ret = copy_event_to_user(group, kevent, buf, count);
> if (unlikely(ret == -EOPENSTALE)) {
> /*
> * We cannot report events with stale fd so drop it.
> --
> 2.17.1
>
>
> --
> Kees Cook