Re: [PATCH 1/2] IB/ipoib: Clean up send-only multicast joins

[Hal Rosenstock <hal-LDSdmyG8hGV8YrgS2mwiifqBs+8SCbDb@public.gmane.org>]

On 8/21/2015 7:34 PM, Jason Gunthorpe wrote:
> Even though we don't expect the group to be created by the SM we
> sill need to provide all the parameters to force the SM to validate
> they are correct.

Out of curiosity, has it been observed that there was inconsistency in
these additional IPoIB parameters between broadcast and non broadcast
groups on client side or is this just to be defensive to make sure this
does not occur ?

-- Hal

> 
> Signed-off-by: Jason Gunthorpe <jgunthorpe-ePGOBjL8dl3ta4EC/59zMFaTQe2KTcn/@public.gmane.org>
> ---
>  drivers/infiniband/ulp/ipoib/ipoib_multicast.c | 47 +++++++++++++++++---------
>  1 file changed, 31 insertions(+), 16 deletions(-)
> 
> diff --git a/drivers/infiniband/ulp/ipoib/ipoib_multicast.c b/drivers/infiniband/ulp/ipoib/ipoib_multicast.c
> index 0d23e0568deb..c0e702c577d5 100644
> --- a/drivers/infiniband/ulp/ipoib/ipoib_multicast.c
> +++ b/drivers/infiniband/ulp/ipoib/ipoib_multicast.c
> @@ -448,8 +448,8 @@ out_locked:
>  	return status;
>  }
>  
> -static void ipoib_mcast_join(struct net_device *dev, struct ipoib_mcast *mcast,
> -			     int create)
> +/* priv->lock must be held when calling */
> +static void ipoib_mcast_join(struct net_device *dev, struct ipoib_mcast *mcast)
>  {
>  	struct ipoib_dev_priv *priv = netdev_priv(dev);
>  	struct ib_sa_multicast *multicast;
> @@ -471,7 +471,14 @@ static void ipoib_mcast_join(struct net_device *dev, struct ipoib_mcast *mcast,
>  		IB_SA_MCMEMBER_REC_PKEY		|
>  		IB_SA_MCMEMBER_REC_JOIN_STATE;
>  
> -	if (create) {
> +	if (mcast != priv->broadcast) {
> +		/*
> +		 * RFC 4391:
> +		 *  The MGID MUST use the same P_Key, Q_Key, SL, MTU,
> +		 *  and HopLimit as those used in the broadcast-GID.  The rest
> +		 *  of attributes SHOULD follow the values used in the
> +		 *  broadcast-GID as well.
> +		 */
>  		comp_mask |=
>  			IB_SA_MCMEMBER_REC_QKEY			|
>  			IB_SA_MCMEMBER_REC_MTU_SELECTOR		|
> @@ -492,19 +499,35 @@ static void ipoib_mcast_join(struct net_device *dev, struct ipoib_mcast *mcast,
>  		rec.sl		  = priv->broadcast->mcmember.sl;
>  		rec.flow_label	  = priv->broadcast->mcmember.flow_label;
>  		rec.hop_limit	  = priv->broadcast->mcmember.hop_limit;
> +
> +		/*
> +		 * Historically Linux IPoIB has never properly supported SEND
> +		 * ONLY join. It emulated it by not providing all the required
> +		 * attributes, which is enough to prevent group creation and
> +		 * detect if there are full members or not. A major problem
> +		 * with supporting SEND ONLY is detecting when the group is
> +		 * auto-destroyed as IPoIB will cache the MLID..
> +		 */
> +#if 1
> +		if (test_bit(IPOIB_MCAST_FLAG_SENDONLY, &mcast->flags))
> +			comp_mask &= ~IB_SA_MCMEMBER_REC_TRAFFIC_CLASS;
> +#else
> +		if (test_bit(IPOIB_MCAST_FLAG_SENDONLY, &mcast->flags))
> +			rec.join_state = 4;
> +#endif
>  	}
>  
> +	spin_unlock_irq(&priv->lock);
>  	multicast = ib_sa_join_multicast(&ipoib_sa_client, priv->ca, priv->port,
>  					 &rec, comp_mask, GFP_KERNEL,
>  					 ipoib_mcast_join_complete, mcast);
> +	spin_lock_irq(&priv->lock);
>  	if (IS_ERR(multicast)) {
>  		ret = PTR_ERR(multicast);
>  		ipoib_warn(priv, "ib_sa_join_multicast failed, status %d\n", ret);
> -		spin_lock_irq(&priv->lock);
>  		/* Requeue this join task with a backoff delay */
>  		__ipoib_mcast_schedule_join_thread(priv, mcast, 1);
>  		clear_bit(IPOIB_MCAST_FLAG_BUSY, &mcast->flags);
> -		spin_unlock_irq(&priv->lock);
>  		complete(&mcast->done);
>  	}
>  }
> @@ -517,7 +540,6 @@ void ipoib_mcast_join_task(struct work_struct *work)
>  	struct ib_port_attr port_attr;
>  	unsigned long delay_until = 0;
>  	struct ipoib_mcast *mcast = NULL;
> -	int create = 1;
>  
>  	if (!test_bit(IPOIB_FLAG_OPER_UP, &priv->flags))
>  		return;
> @@ -566,7 +588,6 @@ void ipoib_mcast_join_task(struct work_struct *work)
>  		if (IS_ERR_OR_NULL(priv->broadcast->mc) &&
>  		    !test_bit(IPOIB_MCAST_FLAG_BUSY, &priv->broadcast->flags)) {
>  			mcast = priv->broadcast;
> -			create = 0;
>  			if (mcast->backoff > 1 &&
>  			    time_before(jiffies, mcast->delay_until)) {
>  				delay_until = mcast->delay_until;
> @@ -590,13 +611,7 @@ void ipoib_mcast_join_task(struct work_struct *work)
>  				/* Found the next unjoined group */
>  				init_completion(&mcast->done);
>  				set_bit(IPOIB_MCAST_FLAG_BUSY, &mcast->flags);
> -				if (test_bit(IPOIB_MCAST_FLAG_SENDONLY, &mcast->flags))
> -					create = 0;
> -				else
> -					create = 1;
> -				spin_unlock_irq(&priv->lock);
> -				ipoib_mcast_join(dev, mcast, create);
> -				spin_lock_irq(&priv->lock);
> +				ipoib_mcast_join(dev, mcast);
>  			} else if (!delay_until ||
>  				 time_before(mcast->delay_until, delay_until))
>  				delay_until = mcast->delay_until;
> @@ -616,9 +631,9 @@ out:
>  		init_completion(&mcast->done);
>  		set_bit(IPOIB_MCAST_FLAG_BUSY, &mcast->flags);
>  	}
> -	spin_unlock_irq(&priv->lock);
>  	if (mcast)
> -		ipoib_mcast_join(dev, mcast, create);
> +		ipoib_mcast_join(dev, mcast);
> +	spin_unlock_irq(&priv->lock);
>  }
>  
>  int ipoib_mcast_start_thread(struct net_device *dev)

--
To unsubscribe from this list: send the line "unsubscribe linux-rdma" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html