Re: [PATCH 4.16 231/272] fanotify: Avoid lost events due to ENOMEM for unlimited queues

From: Amir Goldstein
Date: Mon May 28 2018 - 08:39:15 EST


On Mon, May 28, 2018 at 1:04 PM, Greg Kroah-Hartman
<gregkh@xxxxxxxxxxxxxxxxxxx> wrote:
> 4.16-stable review patch. If anyone has any objections, please let me know.
>

I do not have objections for applying this patch to stable, but AFAIK
it is a correctness patch that doesn't fix any bug and it was mainly added
as a prerequisite to memcg accounting of event allocations, which is not
yet merged and not destined for stable.

Jan? do you agree with my statements above?

Thanks,
Amir.

> ------------------
>
> From: Jan Kara <jack@xxxxxxx>
>
> [ Upstream commit 1f5eaa90010ed7cf0ae90a526c48657d02c6086f ]
>
> Fanotify queues of unlimited length do not expect events can be lost.
> Since these queues are used for system auditing and other security
> related tasks, loosing events can even have security implications.
> Currently, since the allocation is small (32-bytes), it cannot fail
> however when we start accounting events in memcgs, allocation can start
> failing. So avoid loosing events due to failure to allocate memory by
> making event allocation use __GFP_NOFAIL.
>
> Reviewed-by: Amir Goldstein <amir73il@xxxxxxxxx>
> Signed-off-by: Jan Kara <jack@xxxxxxx>
> Signed-off-by: Sasha Levin <alexander.levin@xxxxxxxxxxxxx>
> Signed-off-by: Greg Kroah-Hartman <gregkh@xxxxxxxxxxxxxxxxxxx>
> ---
> fs/notify/fanotify/fanotify.c | 19 ++++++++++++++-----
> fs/notify/fanotify/fanotify.h | 3 ++-
> fs/notify/fanotify/fanotify_user.c | 2 +-
> 3 files changed, 17 insertions(+), 7 deletions(-)
>
> --- a/fs/notify/fanotify/fanotify.c
> +++ b/fs/notify/fanotify/fanotify.c
> @@ -135,23 +135,32 @@ static bool fanotify_should_send_event(s
> return false;
> }
>
> -struct fanotify_event_info *fanotify_alloc_event(struct inode *inode, u32 mask,
> +struct fanotify_event_info *fanotify_alloc_event(struct fsnotify_group *group,
> + struct inode *inode, u32 mask,
> const struct path *path)
> {
> struct fanotify_event_info *event;
> + gfp_t gfp = GFP_KERNEL;
> +
> + /*
> + * For queues with unlimited length lost events are not expected and
> + * can possibly have security implications. Avoid losing events when
> + * memory is short.
> + */
> + if (group->max_events == UINT_MAX)
> + gfp |= __GFP_NOFAIL;
>
> if (fanotify_is_perm_event(mask)) {
> struct fanotify_perm_event_info *pevent;
>
> - pevent = kmem_cache_alloc(fanotify_perm_event_cachep,
> - GFP_KERNEL);
> + pevent = kmem_cache_alloc(fanotify_perm_event_cachep, gfp);
> if (!pevent)
> return NULL;
> event = &pevent->fae;
> pevent->response = 0;
> goto init;
> }
> - event = kmem_cache_alloc(fanotify_event_cachep, GFP_KERNEL);
> + event = kmem_cache_alloc(fanotify_event_cachep, gfp);
> if (!event)
> return NULL;
> init: __maybe_unused
> @@ -206,7 +215,7 @@ static int fanotify_handle_event(struct
> return 0;
> }
>
> - event = fanotify_alloc_event(inode, mask, data);
> + event = fanotify_alloc_event(group, inode, mask, data);
> ret = -ENOMEM;
> if (unlikely(!event))
> goto finish;
> --- a/fs/notify/fanotify/fanotify.h
> +++ b/fs/notify/fanotify/fanotify.h
> @@ -52,5 +52,6 @@ static inline struct fanotify_event_info
> return container_of(fse, struct fanotify_event_info, fse);
> }
>
> -struct fanotify_event_info *fanotify_alloc_event(struct inode *inode, u32 mask,
> +struct fanotify_event_info *fanotify_alloc_event(struct fsnotify_group *group,
> + struct inode *inode, u32 mask,
> const struct path *path);
> --- a/fs/notify/fanotify/fanotify_user.c
> +++ b/fs/notify/fanotify/fanotify_user.c
> @@ -757,7 +757,7 @@ SYSCALL_DEFINE2(fanotify_init, unsigned
> group->fanotify_data.user = user;
> atomic_inc(&user->fanotify_listeners);
>
> - oevent = fanotify_alloc_event(NULL, FS_Q_OVERFLOW, NULL);
> + oevent = fanotify_alloc_event(group, NULL, FS_Q_OVERFLOW, NULL);
> if (unlikely(!oevent)) {
> fd = -ENOMEM;
> goto out_destroy_group;
>
>