Received-SPF: pass (google.com: best guess record for domain of linux-kernel-owner@vger.kernel.org designates 209.132.180.67 as permitted sender) client-ip=209.132.180.67;
Date:   Thu, 21 Nov 2019 12:07:43 -0800 (PST)
From:   Hugh Dickins <hughd@google.com>
To:     "J. R. Okajima" <hooanon05g@gmail.com>
cc:     Hugh Dickins <hughd@google.com>,
        "zhengbin (A)" <zhengbin13@huawei.com>,
        Matthew Wilcox <willy@infradead.org>, viro@zeniv.linux.org.uk,
        linux-mm@kvack.org, linux-kernel@vger.kernel.org,
        houtao1@huawei.com, yi.zhang@huawei.com
Subject: Re: [PATCH] tmpfs: use ida to get inode number
In-Reply-To: <32188.1574336426@jrobl>
Message-ID: <alpine.LSU.2.11.1911211154090.1697@eggly.anvils>
References: <1574259798-144561-1-git-send-email-zhengbin13@huawei.com> <20191120154552.GS20752@bombadil.infradead.org> <1c64e7c2-6460-49cf-6db0-ec5f5f7e09c4@huawei.com> <alpine.LSU.2.11.1911202026040.1825@eggly.anvils> <32188.1574336426@jrobl>
User-Agent: Alpine 2.11 (LSU 23 2013-08-11)
MIME-Version: 1.0
Content-Type: TEXT/PLAIN; charset=US-ASCII
Sender: linux-kernel-owner@vger.kernel.org
Precedence: bulk

On Thu, 21 Nov 2019, J. R. Okajima wrote:
> Hugh Dickins:
> > Internally (in Google) we do rely on good tmpfs inode numbers more
> > than on those of other get_next_ino() filesystems, and carry a patch
> > to mm/shmem.c for it to use 64-bit inode numbers (and separate inode
> > number space for each superblock) - essentially,
> >
> > =09ino =3D sbinfo->next_ino++;
> > =09/* Avoid 0 in the low 32 bits: might appear deleted */
> > =09if (unlikely((unsigned int)ino =3D=3D 0))
> > =09=09ino =3D sbinfo->next_ino++;
> 
> I agree with that "per superblock inum space", but I don't see your
> point.  How can you manage it fully?  I mean how can you decide whether
> the new inum is in use or not?
> For example,
> - you create a file which is assigned inum#10.
> - you or other people create and unlink over and over on the same tmpfs.
> - then sbinfo->next_ino will become zero, skipped, ok.
> - and then it will be 10.
> I don't think you want to share the same inum by two inodes.

64 bits. I haven't done the arithmetic to work out the amusing number,
but zhengbin mentioned the script taking 10 days to duplicate an inode
number in 32 bits, so: a larger number of years than I need to care about.

> 
> Moreover, SysV SHM uses tmpfs and shmget(2) overwrite inum internally.
> It will be another seed of a similar problem.

I was totally ignorant of that peculiarity in ipc/shm.c, thanks for
alerting me to it.  But it doesn't affect what we're doing in tmpfs,
and apparently suits the users of SysV SHM: I don't see any need to
worry about it.

Hugh