Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 2620:137:e000::3:2 as permitted sender) client-ip=2620:137:e000::3:2;
Date:   Fri, 17 Nov 2023 09:36:40 +0100
From:   Mariusz Tkaczyk <mariusz.tkaczyk@linux.intel.com>
To:     junxiao.bi@oracle.com
Cc:     Song Liu <song@kernel.org>, Tejun Heo <tj@kernel.org>,
        linux-kernel@vger.kernel.org, linux-raid@vger.kernel.org,
        gregkh@linuxfoundation.org
Subject: Re: [PATCH] kernfs: support kernfs notify in memory recliam context
Message-ID: <20231117093640.00006b70@linux.intel.com>
In-Reply-To: <c13b580b-434f-4a4b-b0b4-917f8b042de3@oracle.com>
References: <20231114185947.42829-1-junxiao.bi@oracle.com>
        <ZVPFMzHAx9JVz2ak@slm.duckdns.org>
        <c71f1cb7-14d6-45e4-9df1-dc9bc82deda8@oracle.com>
        <ZVPXd-3TshjeScek@slm.duckdns.org>
        <443775e1-ed94-452e-8f06-eadb777bede4@oracle.com>
        <20231115162915.000064f8@linux.intel.com>
        <c13b580b-434f-4a4b-b0b4-917f8b042de3@oracle.com>
MIME-Version: 1.0
Content-Type: text/plain; charset=US-ASCII
Content-Transfer-Encoding: 7bit
Precedence: bulk

On Thu, 16 Nov 2023 09:04:34 -0800
junxiao.bi@oracle.com wrote:

> On 11/15/23 7:30 AM, Mariusz Tkaczyk wrote:
> 
> > On Tue, 14 Nov 2023 15:53:47 -0800
> > junxiao.bi@oracle.com wrote:
> >  
> >> Understood, thanks. Sound like depending on Userspace on memory reclaim
> >> path is really bad idea and the only option for fixing it is to remove
> >> that dependency, but i am not sure that is possible without breaking the
> >> consistency of metadata.
> >>
> >> Thanks,
> >>
> >> Junxiao.  
> > Indeed the project of external metadata management if fragile. You cares
> > about IMSM here (same as me) so ideally we should implement metadata
> > management in kernel- I think that IMSM deserved that after 10 years on the
> > market. There is no better option, other options are just "workarounds" for
> > the lack of metadata management in kernel.  
> Agree, sound like that's the way to proceed.
> >
> > Song, any comments here?
> >
> >  From the second hand, there is native raid which should just work, so
> > maybe you can switch to the native raid?  
> 
> Unfortunately that's is not possible, it's a production setup to use 
> imsm raid.

Implementing IMSM in kernel is a goal for months/ years so I don't
see it in a timeline you would need that on to fix your production setup. It
will be a big feature, but let's wait for Song voice first.

Thanks,
Mariusz