Re: [PATCH rdma-rc] RDMA/cm: Fix missing RDMA_CM_EVENT_REJECTED event after receiving REJ message

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

 



On Mon, Apr 06, 2020 at 08:32:42PM +0300, Leon Romanovsky wrote:
> From: Leon Romanovsky <leonro@xxxxxxxxxxxx>
> 
> The cm_reset_to_idle() call before formatting event changed the CM_ID
> state from IB_CM_REQ_RCVD to be IB_CM_IDLE. It caused to wrong value
> of CM_REJ_MESSAGE_REJECTED field.
> 
> The result of that was that rdma_reject() calls in the passive side
> didn't generate RDMA_CM_EVENT_REJECTED event in the active side.
> 
> Fixes: 81ddb41f876d ("RDMA/cm: Allow ib_send_cm_rej() to be done under lock")
> Signed-off-by: Leon Romanovsky <leonro@xxxxxxxxxxxx>
>  drivers/infiniband/core/cm.c | 24 +++++++++++++-----------
>  1 file changed, 13 insertions(+), 11 deletions(-)
> 
> diff --git a/drivers/infiniband/core/cm.c b/drivers/infiniband/core/cm.c
> index bbbfa77dbce7..06f8eeba423a 100644
> +++ b/drivers/infiniband/core/cm.c
> @@ -1843,11 +1843,9 @@ static void cm_format_mra(struct cm_mra_msg *mra_msg,
> 
>  static void cm_format_rej(struct cm_rej_msg *rej_msg,
>  			  struct cm_id_private *cm_id_priv,
> -			  enum ib_cm_rej_reason reason,
> -			  void *ari,
> -			  u8 ari_length,
> -			  const void *private_data,
> -			  u8 private_data_len)
> +			  enum ib_cm_rej_reason reason, void *ari,
> +			  u8 ari_length, const void *private_data,
> +			  u8 private_data_len, enum ib_cm_state state)
>  {
>  	lockdep_assert_held(&cm_id_priv->lock);
> 
> @@ -1855,7 +1853,7 @@ static void cm_format_rej(struct cm_rej_msg *rej_msg,
>  	IBA_SET(CM_REJ_REMOTE_COMM_ID, rej_msg,
>  		be32_to_cpu(cm_id_priv->id.remote_id));
> 
> -	switch(cm_id_priv->id.state) {
> +	switch (state) {
>  	case IB_CM_REQ_RCVD:
>  		IBA_SET(CM_REJ_LOCAL_COMM_ID, rej_msg, be32_to_cpu(0));
>  		IBA_SET(CM_REJ_MESSAGE_REJECTED, rej_msg, CM_MSG_RESPONSE_REQ);
> @@ -1920,8 +1918,9 @@ static void cm_dup_req_handler(struct cm_work *work,
>  			      cm_id_priv->private_data_len);
>  		break;
>  	case IB_CM_TIMEWAIT:
> -		cm_format_rej((struct cm_rej_msg *) msg->mad, cm_id_priv,
> -			      IB_CM_REJ_STALE_CONN, NULL, 0, NULL, 0);
> +		cm_format_rej((struct cm_rej_msg *)msg->mad, cm_id_priv,
> +			      IB_CM_REJ_STALE_CONN, NULL, 0, NULL, 0,
> +			      cm_id_priv->id.state);

This can just be IB_CM_TIMEWAIT instead of cm_id_priv->id.state

Jason



[Index of Archives]     [Linux USB Devel]     [Video for Linux]     [Linux Audio Users]     [Photo]     [Yosemite News]     [Yosemite Photos]     [Linux Kernel]     [Linux SCSI]     [XFree86]

  Powered by Linux