Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 23.128.96.18 as permitted sender) client-ip=23.128.96.18;
MIME-Version: 1.0
References: <20210120130233.15932-1-ericcurtin17@gmail.com>
In-Reply-To: <20210120130233.15932-1-ericcurtin17@gmail.com>
From:   Eric Curtin <ericcurtin17@gmail.com>
Date:   Wed, 20 Jan 2021 13:39:36 +0000
Message-ID: <CANpvso7mbc9C=xjxLDivb6Qm8_FDu2fLjA-StMw4Z-209PHLqw@mail.gmail.com>
Subject: Re: [PATCH] Increase limit of max_user_watches from 1/25 to 1/16
Cc:     Jonathan Corbet <corbet@lwn.net>,
        Alexander Viro <viro@zeniv.linux.org.uk>,
        "Alexander A. Klimov" <grandmaster@al2klimov.de>,
        "open list:DOCUMENTATION" <linux-doc@vger.kernel.org>,
        open list <linux-kernel@vger.kernel.org>,
        "open list:FILESYSTEMS (VFS and infrastructure)" 
        <linux-fsdevel@vger.kernel.org>
Content-Type: text/plain; charset="UTF-8"
To:     unlisted-recipients:; (no To-header on input)
Precedence: bulk

On Wed, 20 Jan 2021 at 13:02, Eric Curtin <ericcurtin17@gmail.com> wrote:
>
> The current default value for  max_user_watches  is the 1/16 (6.25%) of
> the available low memory, divided for the "watch" cost in bytes.
>
> Tools like inotify-tools and visual studio code, seem to hit these
> limits a little to easy.
>
> Also amending the documentation, it referred to an old value for this.
>
> Signed-off-by: Eric Curtin <ericcurtin17@gmail.com>
> ---
>  Documentation/admin-guide/sysctl/fs.rst | 4 ++--
>  fs/eventpoll.c                          | 4 ++--
>  2 files changed, 4 insertions(+), 4 deletions(-)
>
> diff --git a/Documentation/admin-guide/sysctl/fs.rst b/Documentation/admin-guide/sysctl/fs.rst
> index f48277a0a850..f7fe45e69c41 100644
> --- a/Documentation/admin-guide/sysctl/fs.rst
> +++ b/Documentation/admin-guide/sysctl/fs.rst
> @@ -380,5 +380,5 @@ This configuration option sets the maximum number of "watches" that are
>  allowed for each user.
>  Each "watch" costs roughly 90 bytes on a 32bit kernel, and roughly 160 bytes
>  on a 64bit one.
> -The current default value for  max_user_watches  is the 1/32 of the available
> -low memory, divided for the "watch" cost in bytes.
> +The current default value for  max_user_watches  is the 1/16 (6.25%) of the
> +available low memory, divided for the "watch" cost in bytes.
> diff --git a/fs/eventpoll.c b/fs/eventpoll.c
> index a829af074eb5..de9ef8f6d0b2 100644
> --- a/fs/eventpoll.c
> +++ b/fs/eventpoll.c
> @@ -2352,9 +2352,9 @@ static int __init eventpoll_init(void)
>
>         si_meminfo(&si);
>         /*
> -        * Allows top 4% of lomem to be allocated for epoll watches (per user).
> +        * Allows top 6.25% of lomem to be allocated for epoll watches (per user).
>          */
> -       max_user_watches = (((si.totalram - si.totalhigh) / 25) << PAGE_SHIFT) /
> +       max_user_watches = (((si.totalram - si.totalhigh) / 16) << PAGE_SHIFT) /
>                 EP_ITEM_COST;
>         BUG_ON(max_user_watches < 0);
>
> --
> 2.25.1
>

Please ignore this, this is the wrong limit (an epoll one), I sent
another patch just
to update the documentation to be correct. Weiman Long already kindly solved the
issue in 92890123749bafc317bbfacbe0a62ce08d78efb7

Separate patch is titled "[PATCH] Update
Documentation/admin-guide/sysctl/fs.rst"