Received-SPF: pass (google.com: domain of linux-nfs+bounces-3517-linux.lists.archive=gmail.com@vger.kernel.org designates 147.75.80.249 as permitted sender) client-ip=147.75.80.249;
From: Zhu Yanjun <zyjzyj2000@gmail.com>
Message-ID: <2bc28d01-0345-4d30-870a-3c5506936401@linux.dev>
Date: Sat, 1 Jun 2024 12:48:40 +0200
Precedence: bulk
MIME-Version: 1.0
User-Agent: Mozilla Thunderbird
Subject: Re: [PATCH 2/2] svcrdma: Handle ADDR_CHANGE CM event properly
To: cel@kernel.org, linux-nfs@vger.kernel.org, linux-rdma@vger.kernel.org
Cc: Chuck Lever <chuck.lever@oracle.com>, Sagi Grimberg <sagi@grimberg.me>
References: <20240531131550.64044-4-cel@kernel.org>
 <20240531131550.64044-6-cel@kernel.org>
Content-Language: en-US
In-Reply-To: <20240531131550.64044-6-cel@kernel.org>
Content-Type: text/plain; charset=UTF-8; format=flowed
Content-Transfer-Encoding: 7bit

On 31.05.24 15:15, cel@kernel.org wrote:
> From: Chuck Lever <chuck.lever@oracle.com>
> 
> Sagi tells me that when a bonded device reports an address change,
> the consumer must destroy its listener IDs and create new ones.
> 
> See commit a032e4f6d60d ("nvmet-rdma: fix bonding failover possible
> NULL deref").
> 
> Suggested-by: Sagi Grimberg <sagi@grimberg.me>
> Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
> ---
>   net/sunrpc/xprtrdma/svc_rdma_transport.c | 16 +++++++++++++++-
>   1 file changed, 15 insertions(+), 1 deletion(-)
> 
> diff --git a/net/sunrpc/xprtrdma/svc_rdma_transport.c b/net/sunrpc/xprtrdma/svc_rdma_transport.c
> index fa50b7494a0a..a003b62fb7d5 100644
> --- a/net/sunrpc/xprtrdma/svc_rdma_transport.c
> +++ b/net/sunrpc/xprtrdma/svc_rdma_transport.c
> @@ -284,17 +284,31 @@ static void handle_connect_req(struct rdma_cm_id *new_cma_id,
>    *
>    * Return values:
>    *     %0: Do not destroy @cma_id
> - *     %1: Destroy @cma_id (never returned here)
> + *     %1: Destroy @cma_id
>    *
>    * NB: There is never a DEVICE_REMOVAL event for INADDR_ANY listeners.
>    */
>   static int svc_rdma_listen_handler(struct rdma_cm_id *cma_id,
>   				   struct rdma_cm_event *event)
>   {
> +	struct svcxprt_rdma *cma_xprt = cma_id->context;
> +	struct svc_xprt *cma_rdma = &cma_xprt->sc_xprt;
> +	struct sockaddr *sap = (struct sockaddr *)&cma_id->route.addr.src_addr;
> +	struct rdma_cm_id *listen_id;

I am not sure whether I need to suggest "Reverse Christmas Tree" or not.
This is a trivial problem. ^_^
You can ignore it. And it is not mandatory.

But if we follow this rule, the source code will become more readable.

Zhu Yanjun

> +
>   	switch (event->event) {
>   	case RDMA_CM_EVENT_CONNECT_REQUEST:
>   		handle_connect_req(cma_id, &event->param.conn);
>   		break;
> +	case RDMA_CM_EVENT_ADDR_CHANGE:
> +		listen_id = svc_rdma_create_listen_id(cma_rdma->xpt_net,
> +						      sap, cma_xprt);
> +		if (IS_ERR(listen_id)) {
> +			pr_err("Listener dead, address change failed for device %s\n",
> +				cma_id->device->name);
> +		} else
> +			cma_xprt->sc_cm_id = listen_id;
> +		return 1;
>   	default:
>   		break;
>   	}