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;
Subject: Re: [PATCH v7 08/14] nvmet-core: allow one host per passthru-ctrl
From:   Sagi Grimberg <sagi@grimberg.me>
To:     Max Gurtovoy <maxg@mellanox.com>,
        Logan Gunthorpe <logang@deltatee.com>,
        linux-kernel@vger.kernel.org, linux-nvme@lists.infradead.org,
        linux-block@vger.kernel.org, linux-fsdevel@vger.kernel.org
Cc:     Chaitanya Kulkarni <Chaitanya.Kulkarni@wdc.com>,
        Stephen Bates <sbates@raithlin.com>, Jens Axboe <axboe@fb.com>,
        Keith Busch <kbusch@kernel.org>, Christoph Hellwig <hch@lst.de>
References: <20190801234514.7941-1-logang@deltatee.com>
 <20190801234514.7941-9-logang@deltatee.com>
 <05a74e81-1dbd-725f-1369-5ca5c5918db1@mellanox.com>
 <a6b9db95-a7f0-d1f6-1fa2-8dc13a6aa29e@deltatee.com>
 <5717f515-e051-c420-07b7-299bcfcd1f32@mellanox.com>
 <b0921c72-93f1-f67a-c4b3-31baeb1c39cb@grimberg.me>
 <b352c7f1-2629-e72f-9c85-785e0cf7c2c1@mellanox.com>
 <24e2ddd0-4b2a-8092-cf91-df8c0fb482e5@grimberg.me>
Message-ID: <e4430207-7def-8776-0289-0d58689dc0cd@grimberg.me>
Date:   Thu, 22 Aug 2019 12:17:39 -0700
User-Agent: Mozilla/5.0 (X11; Linux x86_64; rv:60.0) Gecko/20100101
 Thunderbird/60.8.0
MIME-Version: 1.0
In-Reply-To: <24e2ddd0-4b2a-8092-cf91-df8c0fb482e5@grimberg.me>
Content-Type: text/plain; charset=utf-8; format=flowed
Content-Language: en-US
Content-Transfer-Encoding: 7bit
Sender: linux-kernel-owner@vger.kernel.org
Precedence: bulk


>>>> I don't understand why we don't limit a regular ctrl to single 
>>>> access and we do it for the PT ctrl.
>>>>
>>>> I guess the block layer helps to sync between multiple access in 
>>>> parallel but we can do it as well.
>>>>
>>>> Also, let's say you limit the access to this subsystem to 1 user, 
>>>> the bdev is still accessibly for local user and also you can create 
>>>> a different subsystem that will use this device (PT and non-PT ctrl).
>>>>
>>>> Sagi,
>>>>
>>>> can you explain the trouble you meant and how this limitation solve 
>>>> it ?
>>>
>>> Its different to emulate the controller with all its admin
>>> commands vs. passing it through to the nvme device.. (think of format 
>>> nvm)
>>>
>>>
>>>
>> we don't need to support format command for PT ctrl as we don't 
>> support other commands such create_sq/cq.
> 
> That is just an example, basically every command that we are not aware
> of we simply passthru to the drive without knowing the implications
> on a multi-host environment..

If we were to change the logic of nvmet_parse_passthru_admin_cmd to
have the default case do nvmet_parse_admin_cmd, and only have
the vendor-specific space opcodes do nvmet_passthru_execute_cmd
then I could not see at the moment how we can break a multi-host
export...