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: <2170fe67-2e65-630b-22dc-a8f6d3f7966f@gmail.com>
Date:   Fri, 7 Apr 2023 21:16:18 +0200
MIME-Version: 1.0
User-Agent: Mozilla/5.0 (X11; Linux x86_64; rv:102.0) Gecko/20100101
 Thunderbird/102.9.0
Subject: Re: [PATCH] scsi: target: cxgbit: check skb dequeue result
To:     Denis Plotnikov <den-plotnikov@yandex-team.ru>,
        linux-scsi@vger.kernel.org
Cc:     linux-kernel@vger.kernel.org, target-devel@vger.kernel.org,
        varun@chelsio.com, nab@linux-iscsi.org, martin.petersen@oracle.com
References: <20230406150405.300909-1-den-plotnikov@yandex-team.ru>
Content-Language: en-US
From:   Bodo Stroesser <bostroesser@gmail.com>
In-Reply-To: <20230406150405.300909-1-den-plotnikov@yandex-team.ru>
Content-Type: text/plain; charset=UTF-8; format=flowed
Content-Transfer-Encoding: 7bit
Precedence: bulk

On 06.04.23 17:04, Denis Plotnikov wrote:
> On a couple of abort packet paths skb dequeuing may end up with
> returning NULL, which, in turn, may end up with further null
> pointer dereference.
> 
> Fix it by checking the return value of skb dequeuing.
> 
> Found by Linux Verification Center(linuxtesting.org) with SVACE.
> 
> Fixes: 9730ffcb8957 (cxgbit: add files for cxgbit.ko)
> Signed-off-by: Denis Plotnikov <den-plotnikov@yandex-team.ru>
> ---
>   drivers/target/iscsi/cxgbit/cxgbit_cm.c | 10 +++++++---
>   1 file changed, 7 insertions(+), 3 deletions(-)
> 
> diff --git a/drivers/target/iscsi/cxgbit/cxgbit_cm.c b/drivers/target/iscsi/cxgbit/cxgbit_cm.c
> index 518ded214e74..d43fd761c20a 100644
> --- a/drivers/target/iscsi/cxgbit/cxgbit_cm.c
> +++ b/drivers/target/iscsi/cxgbit/cxgbit_cm.c
> @@ -669,6 +669,9 @@ static int cxgbit_send_abort_req(struct cxgbit_sock *csk)
>   		cxgbit_send_tx_flowc_wr(csk);
>   
>   	skb = __skb_dequeue(&csk->skbq);
> +	if (!skb)
> +		return 0;
> +
>   	cxgb_mk_abort_req(skb, len, csk->tid, csk->txq_idx,
>   			  csk->com.cdev, cxgbit_abort_arp_failure);
>   
> @@ -1769,9 +1772,10 @@ static void cxgbit_abort_req_rss(struct cxgbit_sock *csk, struct sk_buff *skb)
>   		cxgbit_send_tx_flowc_wr(csk);
>   
>   	rpl_skb = __skb_dequeue(&csk->skbq);
> -
> -	cxgb_mk_abort_rpl(rpl_skb, len, csk->tid, csk->txq_idx);
> -	cxgbit_ofld_send(csk->com.cdev, rpl_skb);
> +	if (!rpl_skb) {

Honestly I have no clue about cxgbit, but to avoid null pointer
dereference, shouldn't it be "if (rpl_skb) {"?

Bodo


> +		cxgb_mk_abort_rpl(rpl_skb, len, csk->tid, csk->txq_idx);
> +		cxgbit_ofld_send(csk->com.cdev, rpl_skb);
> +	}
>   
>   	if (wakeup_thread) {
>   		cxgbit_queue_rx_skb(csk, skb);