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;
Organization: Red Hat UK Ltd. Registered Address: Red Hat UK Ltd, Amberley
        Place, 107-111 Peascod Street, Windsor, Berkshire, SI4 1TE, United
        Kingdom.
        Registered in England and Wales under Company Registration No. 3798903
From:   David Howells <dhowells@redhat.com>
In-Reply-To: <CAOQ4uxhRS3MGEnCUDcsB1RL0d1Oy0g0Rzm75hVFAJw2dJ7uKSA@mail.gmail.com>
References: <CAOQ4uxhRS3MGEnCUDcsB1RL0d1Oy0g0Rzm75hVFAJw2dJ7uKSA@mail.gmail.com> <20220128074731.1623738-1-hch@lst.de> <918225.1643364739@warthog.procyon.org.uk>
To:     Amir Goldstein <amir73il@gmail.com>
Cc:     dhowells@redhat.com, Christoph Hellwig <hch@lst.de>,
        Linus Torvalds <torvalds@linux-foundation.org>,
        Al Viro <viro@zeniv.linux.org.uk>,
        linux-fsdevel <linux-fsdevel@vger.kernel.org>,
        linux-kernel <linux-kernel@vger.kernel.org>,
        Chaitanya Kulkarni <chaitanyak@nvidia.com>,
        Miklos Szeredi <miklos@szeredi.hu>
Subject: Re: [PATCH v2] fs: rename S_KERNEL_FILE
MIME-Version: 1.0
Content-Type: text/plain; charset="us-ascii"
Content-ID: <922908.1643369759.1@warthog.procyon.org.uk>
Date:   Fri, 28 Jan 2022 11:35:59 +0000
Message-ID: <922909.1643369759@warthog.procyon.org.uk>
Precedence: bulk

Amir Goldstein <amir73il@gmail.com> wrote:

> Good idea, but then the helpers to set the flag should not be internal
> to cachefiles and the locking semantics should be clear.

I could move them out, at least partially.  They do log some information
that's private to cachefiles through the tracepoint, but it's just one number
and could be passed in as a parameter.  I could move the tracepoint to
somewhere more generic.

> FYI, overlayfs already takes an "exclusive lock" on upper/work dir
> among all ovl instances.
> 
> How do you feel about hoisting ovl_inuse_* helpers to fs.h
> and rename s/I_OVL_INUSE/I_EXCL_INUSE?

Fine by me.  Sharing a cache with or through an overlay would make for very
fun coherency management.

> Whether deny rmdir should have its own flag or not I don't know,
> but from ovl POV I *think* it should not be a problem to deny rmdir
> for the ovl upper/work dirs as long as ovl is mounted(?).

What's the consequence of someone rearranging the directories directly in the
contributing dirs whilst there's an overlay over them?

> Another problem with generic deny of rmdir is that users getting
> EBUSY have no way to figure out the reason.
> At least for a specific subsystem (i.e. cachefiles) users should be able
> to check if the denied dir is in the subsystem's inventory(?)

I could add a tracepoint for that.  It could form a set with the cachefiles
tracepoints if I move those out too.

David