Received-SPF: pass (google.com: domain of linux-ext4-owner@vger.kernel.org designates 23.128.96.18 as permitted sender) client-ip=23.128.96.18;
MIME-Version: 1.0
References: <20211019000015.1666608-1-krisman@collabora.com> <20211019000015.1666608-21-krisman@collabora.com>
In-Reply-To: <20211019000015.1666608-21-krisman@collabora.com>
From:   Amir Goldstein <amir73il@gmail.com>
Date:   Tue, 19 Oct 2021 08:50:23 +0300
Message-ID: <CAOQ4uxi3C7MQxGPc1fD8ZyRTkyJZQac3_M-0aGYzPKbJ6AK8Jg@mail.gmail.com>
Subject: Re: [PATCH v8 20/32] fanotify: Dynamically resize the FAN_FS_ERROR pool
To:     Gabriel Krisman Bertazi <krisman@collabora.com>
Cc:     Jan Kara <jack@suse.com>, "Darrick J. Wong" <djwong@kernel.org>,
        Theodore Tso <tytso@mit.edu>,
        Dave Chinner <david@fromorbit.com>,
        David Howells <dhowells@redhat.com>,
        Khazhismel Kumykov <khazhy@google.com>,
        linux-fsdevel <linux-fsdevel@vger.kernel.org>,
        Ext4 <linux-ext4@vger.kernel.org>,
        Linux API <linux-api@vger.kernel.org>, kernel@collabora.com
Content-Type: text/plain; charset="UTF-8"
Precedence: bulk

On Tue, Oct 19, 2021 at 3:03 AM Gabriel Krisman Bertazi
<krisman@collabora.com> wrote:
>
> Allow the FAN_FS_ERROR group mempool to grow up to an upper limit
> dynamically, instead of starting already at the limit.  This doesn't
> bother resizing on mark removal, but next time a mark is added, the slot
> will be either reused or resized.  Also, if several marks are being
> removed at once, most likely the group is going away anyway.
>
> Signed-off-by: Gabriel Krisman Bertazi <krisman@collabora.com>
> ---
>  fs/notify/fanotify/fanotify_user.c | 26 +++++++++++++++++++++-----
>  include/linux/fsnotify_backend.h   |  1 +
>  2 files changed, 22 insertions(+), 5 deletions(-)
>
> diff --git a/fs/notify/fanotify/fanotify_user.c b/fs/notify/fanotify/fanotify_user.c
> index f77581c5b97f..a860c286e885 100644
> --- a/fs/notify/fanotify/fanotify_user.c
> +++ b/fs/notify/fanotify/fanotify_user.c
> @@ -959,6 +959,10 @@ static int fanotify_remove_mark(struct fsnotify_group *group,
>
>         removed = fanotify_mark_remove_from_mask(fsn_mark, mask, flags,
>                                                  umask, &destroy_mark);
> +
> +       if (removed & FAN_FS_ERROR)
> +               group->fanotify_data.error_event_marks--;
> +
>         if (removed & fsnotify_conn_mask(fsn_mark->connector))
>                 fsnotify_recalc_mask(fsn_mark->connector);
>         if (destroy_mark)
> @@ -1057,12 +1061,24 @@ static struct fsnotify_mark *fanotify_add_new_mark(struct fsnotify_group *group,
>
>  static int fanotify_group_init_error_pool(struct fsnotify_group *group)
>  {
> -       if (mempool_initialized(&group->fanotify_data.error_events_pool))
> -               return 0;
> +       int ret;
> +
> +       if (group->fanotify_data.error_event_marks >=
> +           FANOTIFY_DEFAULT_MAX_FEE_POOL)
> +               return -ENOMEM;
>
> -       return mempool_init_kmalloc_pool(&group->fanotify_data.error_events_pool,
> -                                        FANOTIFY_DEFAULT_MAX_FEE_POOL,
> -                                        sizeof(struct fanotify_error_event));
> +       if (!mempool_initialized(&group->fanotify_data.error_events_pool))
> +               ret = mempool_init_kmalloc_pool(
> +                               &group->fanotify_data.error_events_pool,
> +                                1, sizeof(struct fanotify_error_event));
> +       else
> +               ret = mempool_resize(&group->fanotify_data.error_events_pool,
> +                                    group->fanotify_data.error_event_marks + 1);
> +
> +       if (!ret)
> +               group->fanotify_data.error_event_marks++;
> +
> +       return ret;
>  }

This is not what I had in mind.
I was thinking start with ~32 and double each time limit is reached.
And also, this code grows the pool to infinity with add/remove mark loop.

Anyway, since I clearly did not understand how mempool works and
Jan had some different ideas I would leave it to Jan to explain
how he wants the mempool init limit and resize to be implemented.

Thanks,
Amir.