Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 2620:137:e000::1:20 as permitted sender) client-ip=2620:137:e000::1:20;
Message-ID: <305fbf18-6759-9b00-6fc0-93a4a2e40be2@gmail.com>
Date:   Sun, 2 Oct 2022 10:12:15 -0700
MIME-Version: 1.0
User-Agent: Mozilla/5.0 (Windows NT 10.0; Win64; x64; rv:102.0) Gecko/20100101
 Thunderbird/102.3.0
Subject: Re: [PATCH] nvme-fc: fix sleep-in-atomic-context bug caused by
 nvme_fc_rcv_ls_req
Content-Language: en-US
To:     Duoming Zhou <duoming@zju.edu.cn>, linux-nvme@lists.infradead.org
Cc:     linux-kernel@vger.kernel.org, james.smart@broadcom.com,
        kbusch@kernel.org, axboe@fb.com, hch@lst.de, sagi@grimberg.me
References: <20221002001909.20070-1-duoming@zju.edu.cn>
From:   James Smart <jsmart2021@gmail.com>
In-Reply-To: <20221002001909.20070-1-duoming@zju.edu.cn>
Content-Type: text/plain; charset=UTF-8; format=flowed
Content-Transfer-Encoding: 7bit
Precedence: bulk

On 10/1/2022 5:19 PM, Duoming Zhou wrote:
> The function lpfc_poll_timeout() is a timer handler that runs in an
> atomic context, but it calls "kzalloc(.., GFP_KERNEL)" that may sleep.
> As a result, the sleep-in-atomic-context bug will happen. The processes
> is shown below:
> 
> lpfc_poll_timeout()
>   lpfc_sli_handle_fast_ring_event()
>    lpfc_sli_process_unsol_iocb()
>     lpfc_complete_unsol_iocb()
>      lpfc_nvme_unsol_ls_handler()
>       lpfc_nvme_handle_lsreq()
>        nvme_fc_rcv_ls_req()
>         kzalloc(sizeof(.., GFP_KERNEL) //may sleep
> 
> This patch changes the gfp_t parameter of kzalloc() from GFP_KERNEL to
> GFP_ATOMIC in order to mitigate the bug.
> 
> Fixes: 14fd1e98afaf ("nvme-fc: Add Disconnect Association Rcv support")
> Signed-off-by: Duoming Zhou <duoming@zju.edu.cn>
> ---
>   drivers/nvme/host/fc.c | 2 +-
>   1 file changed, 1 insertion(+), 1 deletion(-)
> 
> diff --git a/drivers/nvme/host/fc.c b/drivers/nvme/host/fc.c
> index 127abaf9ba5..36698dfc8b3 100644
> --- a/drivers/nvme/host/fc.c
> +++ b/drivers/nvme/host/fc.c
> @@ -1754,7 +1754,7 @@ nvme_fc_rcv_ls_req(struct nvme_fc_remote_port *portptr,
>   	lsop = kzalloc(sizeof(*lsop) +
>   			sizeof(union nvmefc_ls_requests) +
>   			sizeof(union nvmefc_ls_responses),
> -			GFP_KERNEL);
> +			GFP_ATOMIC);
>   	if (!lsop) {
>   		dev_info(lport->dev,
>   			"RCV %s LS failed: No memory\n",

I would prefer this was fixed within lpfc rather than introducing atomic 
allocations (1st in either host or target transport).  It was introduced 
by lpfc change in irq handling style.

-- james


-- james