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: <161581005972.2850696.12854461380574304411.stgit@warthog.procyon.org.uk>
 <CAJfpegsb9XrUct5zawN+kS_DSfowBf2BnrZzG+cQXUvsGZZuow@mail.gmail.com> <2857440.1615815708@warthog.procyon.org.uk>
In-Reply-To: <2857440.1615815708@warthog.procyon.org.uk>
From:   Miklos Szeredi <miklos@szeredi.hu>
Date:   Mon, 15 Mar 2021 15:22:47 +0100
Message-ID: <CAJfpegujN1KA=bS55RobUGphfUbY7BAnPke+dPdUnZbcWudwxQ@mail.gmail.com>
Subject: Re: [RFC][PATCH 0/3] vfs: Use an xarray instead of inserted bookmarks
 to scan mount list
To:     David Howells <dhowells@redhat.com>
Cc:     Alexander Viro <viro@zeniv.linux.org.uk>,
        Matthew Wilcox <willy@infradead.org>,
        Ian Kent <raven@themaw.net>,
        Linus Torvalds <torvalds@linux-foundation.org>,
        linux-fsdevel@vger.kernel.org, linux-kernel@vger.kernel.org
Content-Type: text/plain; charset="UTF-8"
Precedence: bulk

On Mon, Mar 15, 2021 at 2:41 PM David Howells <dhowells@redhat.com> wrote:
>
> Miklos Szeredi <miklos@szeredi.hu> wrote:
>
> > >  (2) We can use the file position to represent the mnt_id and can jump to
> > >      it directly - ie. using seek() to jump to a mount object by its ID.
> >
> > What happens if the mount at the current position is removed?
>
> umount_tree() requires the namespace_sem to be writelocked, so that should be
> fine as the patches currently read-lock that whilst doing /proc/*/mount*
>
> I'm assuming that kern_unmount() won't be a problem as it is there to deal
> with mounts made by kern_mount() which don't get added to the mount list
> (mnt_ns is MNT_NS_INTERNAL).  kern_unmount_array() seems to be the same
> because overlayfs gives it mounts generated by clone_private_mount().  It
> might be worth putting a WARN_ON() in kern_unmount() to require this.
>
> When reading through proc, m_start() calls xas_find() which returns the entry
> at the starting index or, if not present, the next higher entry.

This will break the property of new mounts always being added to the
end of the list.  That's likely a regression for nerural based parsers
(i.e. people), probably less so for machine parsers.

Thanks,
Miklos

>
> David
>