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;
MIME-Version: 1.0
In-Reply-To: <FEA49C2E-639E-4B1E-A4A1-0E0879F1214A@fb.com>
References: <20180927135211.31641-1-dschatzberg@fb.com> <CAJfpegsRNv3Uby2w-B0+d4fE8ecSRkT0ef2AfQ_QnnEVT-nkmA@mail.gmail.com>
 <FEA49C2E-639E-4B1E-A4A1-0E0879F1214A@fb.com>
From:   Miklos Szeredi <miklos@szeredi.hu>
Date:   Wed, 10 Oct 2018 10:29:39 +0200
Message-ID: <CAJfpegugRrA7-U4+6zRWZBEuTiYa=GgLqCeJWvhBz15-5szubQ@mail.gmail.com>
Subject: Re: [PATCH] fuse: enable caching of symlinks
To:     Dan Schatzberg <dschatzberg@fb.com>
Cc:     Tejun Heo <tj@kernel.org>, Kernel Team <Kernel-team@fb.com>,
        "linux-fsdevel@vger.kernel.org" <linux-fsdevel@vger.kernel.org>,
        "linux-kernel@vger.kernel.org" <linux-kernel@vger.kernel.org>
Content-Type: text/plain; charset="UTF-8"
Sender: linux-kernel-owner@vger.kernel.org
Precedence: bulk

On Tue, Oct 2, 2018 at 11:33 PM, Dan Schatzberg <dschatzberg@fb.com> wrote:
> On 9/27/18, 9:59 AM, "Miklos Szeredi" <miklos@szeredi.hu> wrote:
>
>>  On Thu, Sep 27, 2018 at 3:52 PM, Dan Schatzberg <dschatzberg@fb.com> wrote:
>>> FUSE file reads are cached in the page cache, but symlink reads are
>>> not. This patch enables FUSE READLINK operations to be cached which
>>> can improve performance of some FUSE workloads.
>>>
>>> In particular, I'm working on a FUSE filesystem for access to source
>>> code and discovered that about a 10% improvement to build times is
>>> achieved with this patch (there are a lot of symlinks in the source
>>> tree).
>>>
>>> Please provide feedback, I'm looking to flesh this out more.
>>
>> Need to think about how/when to invalidate the cached symlink
>> contents.  I think treating it like metadata (i.e. look at
>> attr_timeout for validity) would be the simplest.
>>
>> Thanks,
>> Miklos
>>
>
> Any further thoughts on this? Just so I can confirm my understanding,
> the attached patch will unconditionally cache READLINK responses
> However, based on the entry timeout provided by the LOOKUP response,
> Userspace can decide to invalidate this by providing a different inode on
> a subsequent LOOKUP (e.g. during fuse_dentry_revalidate())

Indeed.

> Would you like there to be a way to invalidate the cached symlink without
>  changing the inode via LOOKUP?

No, that would make no sense, since symlink contents are not mutable.
This was just a braino on my part.

I still worry about backward compatibility, though.  So we should add
a flag to negotiate symlink caching in with the INIT request.

Thanks,
Miklos