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;
Subject: Re: [PATCH v2] nvme-multipath: Early exit if no path is available
To:     Chao Leng <lengchao@huawei.com>, Daniel Wagner <dwagner@suse.de>
Cc:     linux-nvme@lists.infradead.org, linux-kernel@vger.kernel.org,
        Jens Axboe <axboe@fb.com>, Hannes Reinecke <hare@suse.de>,
        Keith Busch <kbusch@kernel.org>, Christoph Hellwig <hch@lst.de>
References: <20210127103033.15318-1-dwagner@suse.de>
 <db9baae0-547c-7ff4-8b2c-0b95f14be67c@huawei.com>
 <20210128075837.u5u56t23fq5gu6ou@beryllium.lan>
 <69575290-200e-b4a1-4269-c71e4c2cc37b@huawei.com>
 <20210128094004.erwnszjqcxlsi2kd@beryllium.lan>
 <ebb1d098-3ded-e592-4419-e905aabe824f@huawei.com>
From:   Sagi Grimberg <sagi@grimberg.me>
Message-ID: <675d3cf7-1ae8-adc5-b6d0-359fe10f6b23@grimberg.me>
Date:   Thu, 28 Jan 2021 17:42:34 -0800
User-Agent: Mozilla/5.0 (X11; Linux x86_64; rv:68.0) Gecko/20100101
 Thunderbird/68.10.0
MIME-Version: 1.0
In-Reply-To: <ebb1d098-3ded-e592-4419-e905aabe824f@huawei.com>
Content-Type: text/plain; charset=utf-8; format=flowed
Content-Language: en-US
Content-Transfer-Encoding: 7bit
Precedence: bulk


>> You can't see exactly where it dies but I followed the assembly to
>> nvme_round_robin_path(). Maybe it's not the initial nvme_next_ns(head,
>> old) which returns NULL but nvme_next_ns() is returning NULL eventually
>> (list_next_or_null_rcu()).
> So there is other bug cause nvme_next_ns abormal.
> I review the code about head->list and head->current_path, I find 2 bugs
> may cause the bug:
> First, I already send the patch. see:
> https://lore.kernel.org/linux-nvme/20210128033351.22116-1-lengchao@huawei.com/ 
> 
> Second, in nvme_ns_remove, list_del_rcu is before
> nvme_mpath_clear_current_path. This may cause "old" is deleted from the
> "head", but still use "old". I'm not sure there's any other
> consideration here, I will check it and try to fix it.

The reason why we first remove from head->list and only then clear
current_path is because the other way around there is no way
to guarantee that that the ns won't be assigned as current_path
again (because it is in head->list).

nvme_ns_remove fences continue of deletion of the ns by synchronizing
the srcu such that for sure the current_path clearance is visible.