Received-SPF: pass (google.com: domain of linux-nfs-owner@vger.kernel.org designates 2620:137:e000::1:20 as permitted sender) client-ip=2620:137:e000::1:20;
From:   Benjamin Coddington <bcodding@redhat.com>
To:     Trond Myklebust <trondmy@hammerspace.com>
Cc:     Alexander Viro <viro@zeniv.linux.org.uk>,
        Christian Brauner <brauner@kernel.org>,
        Jeffrey Layton <jlayton@kernel.org>,
        Neil Brown <neilb@suse.de>,
        Dave Wysochanski <dwysocha@redhat.com>,
        linux-fsdevel <linux-fsdevel@vger.kernel.org>,
        linux-nfs <linux-nfs@vger.kernel.org>,
        David Howells <dhowells@redhat.com>,
        Christoph Hellwig <hch@lst.de>
Subject: Re: allowing for a completely cached umount(2) pathwalk
Date:   Fri, 14 Apr 2023 15:01:01 -0400
Message-ID: <6F3DB6E1-F104-492B-9AF1-5AEC8C27D267@redhat.com>
In-Reply-To: <E746F6B4-779A-4184-A2A7-5879E3D3DAEE@hammerspace.com>
References: <95ee689c76bf034fa2fe9fade0bccdb311f3a04f.camel@kernel.org>
 <168142566371.24821.15867603327393356000@noble.neil.brown.name>
 <20230414024312.GF3390869@ZenIV>
 <2631cb9c05087ddd917679b7cebc58cb42cd2de6.camel@kernel.org>
 <20230414-sowas-unmittelbar-67fdae9ca5cd@brauner>
 <9192A185-03BF-4062-B12F-E7EF52578014@hammerspace.com>
 <20230414-leiht-lektion-18f5a7a38306@brauner>
 <91D4AC04-A016-48A9-8E3A-BBB6C38E8C4B@hammerspace.com>
 <4F4F5C98-AA06-40FB-AE51-79E860CD1D76@hammerspace.com>
 <20230414162253.GL3390869@ZenIV>
 <E746F6B4-779A-4184-A2A7-5879E3D3DAEE@hammerspace.com>
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: quoted-printable
Precedence: bulk

On 14 Apr 2023, at 12:41, Trond Myklebust wrote:
>
> I mean both cases. Doing a lazy umount with a hard mounted filesystem i=
s a risk sport: if the server does become permanently borked, you can fil=
l up your page cache with stuff that can=E2=80=99t be evicted. Most users=
 don=E2=80=99t realise this, so they get confused when it happens (partic=
ularly since the filesystem is out-of-sight and hence out-of-mind).

I've been pecking away at a sysfs knob for this case.  Seemed a clearer p=
ath to destruction.

>>
>> Note, BTW, that hard vs. soft is a property of fs instance; if you hav=
e
>> it present elsewhere in the mount tree, flipping it would affect all
>> such places.  I don't see any good way to make it a per-mount thing, T=
BH=E2=80=A6
>
>
> The main use case is for when the server is permanently down, so normal=
ly it shouldn=E2=80=99t be a problem with flipping the mode on all instan=
ces.

Is there another case?  Because, if so..

> That said, it might be nice to make it per-mountpoint at some time. We =
do have the ability to declare individual RPC calls to time out, so it=E2=
=80=99s doable at the RPC level. All we would really need is the ability =
to store a per-vfsmount flag.

=2E. maybe vfsmount's mnt_root d_fsdata?

Ben