Re: [PATCH] scsi: RDMA/srpt: Fix incorrect pointer dereference

[Bart Van Assche]

On 12/17/19 11:26 AM, Aditya Pakki wrote:
> In srpt_queue_response(), the rdma channel ch is first
> dereferenced and then checked for NULL. This renders the
> assertion ineffective. This patch removes the assertion and
> avoids potential NULL pointer dereference.
>
> Signed-off-by: Aditya Pakki <pakki001@xxxxxxx>
> ---
>   drivers/infiniband/ulp/srpt/ib_srpt.c | 8 +++++---
>   1 file changed, 5 insertions(+), 3 deletions(-)
>
> diff --git a/drivers/infiniband/ulp/srpt/ib_srpt.c b/drivers/infiniband/ulp/srpt/ib_srpt.c
> index 23c782e3d49a..bbc6729c81c0 100644
> --- a/drivers/infiniband/ulp/srpt/ib_srpt.c
> +++ b/drivers/infiniband/ulp/srpt/ib_srpt.c
> @@ -2803,15 +2803,17 @@ static void srpt_queue_response(struct se_cmd *cmd)
>   	struct srpt_send_ioctx *ioctx =
>   		container_of(cmd, struct srpt_send_ioctx, cmd);
>   	struct srpt_rdma_ch *ch = ioctx->ch;
> -	struct srpt_device *sdev = ch->sport->sdev;
>   	struct ib_send_wr send_wr, *first_wr = &send_wr;
> -	struct ib_sge sge;
>   	enum srpt_command_state state;
> +	struct srpt_device *sdev;
>   	int resp_len, ret, i;
> +	struct ib_sge sge;
>   	u8 srp_tm_status;
>   
> -	BUG_ON(!ch);
> +	if (WARN_ON(!ch))
> +		return;
>   
> +	sdev = ch->sport->sdev;
>   	state = ioctx->state;
>   	switch (state) {
>   	case SRPT_STATE_NEW:

Instead of making all these changes, please remove the BUG_ON(!ch) 
statement. If the condition ioctx->ch == NULL would ever be encountered 
then the call trace reported on the console will be sufficient to figure 
out what happened.

Thanks,

Bart.