IB/umad: Add P_Key index support

Add support for setting the P_Key index of sent MADs and getting the
P_Key index of received MADs. This requires a change to the layout of
the ABI structure struct ib_user_mad_hdr, so to avoid breaking
compatibility, we default to the old (unchanged) ABI and add a new
ioctl IB_USER_MAD_ENABLE_PKEY that allows applications that are aware
of the new ABI to opt into using it.

We plan on switching to the new ABI by default in a year or so, and
this patch adds a warning that is printed when an application uses the
old ABI, to push people towards converting to the new ABI.

Signed-off-by: Roland Dreier <[e-mail]>
Reviewed-by: Sean Hefty <[e-mail]>
Reviewed-by: Hal Rosenstock <[e-mail]>

Documentation/infiniband/user_mad.txt

···
   request/response pairs.  The upper 32 bits are reserved for use by
   the kernel and will be overwritten before a MAD is sent.
 
+P_Key Index Handling
+
+  The old ib_umad interface did not allow setting the P_Key index for
+  MADs that are sent and did not provide a way for obtaining the P_Key
+  index of received MADs.  A new layout for struct ib_user_mad_hdr
+  with a pkey_index member has been defined; however, to preserve
+  binary compatibility with older applications, this new layout will
+  not be used unless the IB_USER_MAD_ENABLE_PKEY ioctl is called
+  before a file descriptor is used for anything else.
+
+  In September 2008, the IB_USER_MAD_ABI_VERSION will be incremented
+  to 6, the new layout of struct ib_user_mad_hdr will be used by
+  default, and the IB_USER_MAD_ENABLE_PKEY ioctl will be removed.
+
 Setting IsSM Capability Bit
 
   To set the IsSM capability bit for a port, simply open the

drivers/infiniband/core/user_mad.c

···
 	wait_queue_head_t	recv_wait;
 	struct ib_mad_agent    *agent[IB_UMAD_MAX_AGENTS];
 	int			agents_dead;
+	u8			use_pkey_index;
+	u8			already_used;
 };
 
 struct ib_umad_packet {
···
 		container_of(ref, struct ib_umad_device, ref);
 
 	kfree(dev);
+}
+
+static int hdr_size(struct ib_umad_file *file)
+{
+	return file->use_pkey_index ? sizeof (struct ib_user_mad_hdr) :
+		sizeof (struct ib_user_mad_hdr_old);
 }
 
 /* caller must hold port->mutex at least for reading */
···
 	packet->length = mad_recv_wc->mad_len;
 	packet->recv_wc = mad_recv_wc;
 
-	packet->mad.hdr.status    = 0;
-	packet->mad.hdr.length    = sizeof (struct ib_user_mad) +
-				    mad_recv_wc->mad_len;
-	packet->mad.hdr.qpn 	  = cpu_to_be32(mad_recv_wc->wc->src_qp);
-	packet->mad.hdr.lid 	  = cpu_to_be16(mad_recv_wc->wc->slid);
-	packet->mad.hdr.sl  	  = mad_recv_wc->wc->sl;
-	packet->mad.hdr.path_bits = mad_recv_wc->wc->dlid_path_bits;
+	packet->mad.hdr.status	   = 0;
+	packet->mad.hdr.length	   = hdr_size(file) + mad_recv_wc->mad_len;
+	packet->mad.hdr.qpn	   = cpu_to_be32(mad_recv_wc->wc->src_qp);
+	packet->mad.hdr.lid	   = cpu_to_be16(mad_recv_wc->wc->slid);
+	packet->mad.hdr.sl	   = mad_recv_wc->wc->sl;
+	packet->mad.hdr.path_bits  = mad_recv_wc->wc->dlid_path_bits;
+	packet->mad.hdr.pkey_index = mad_recv_wc->wc->pkey_index;
 	packet->mad.hdr.grh_present = !!(mad_recv_wc->wc->wc_flags & IB_WC_GRH);
 	if (packet->mad.hdr.grh_present) {
 		struct ib_ah_attr ah_attr;
···
 	ib_free_recv_mad(mad_recv_wc);
 }
 
-static ssize_t copy_recv_mad(char __user *buf, struct ib_umad_packet *packet,
-			     size_t count)
+static ssize_t copy_recv_mad(struct ib_umad_file *file, char __user *buf,
+			     struct ib_umad_packet *packet, size_t count)
 {
 	struct ib_mad_recv_buf *recv_buf;
 	int left, seg_payload, offset, max_seg_payload;
···
 	/* We need enough room to copy the first (or only) MAD segment. */
 	recv_buf = &packet->recv_wc->recv_buf;
 	if ((packet->length <= sizeof (*recv_buf->mad) &&
-	     count < sizeof (packet->mad) + packet->length) ||
+	     count < hdr_size(file) + packet->length) ||
 	    (packet->length > sizeof (*recv_buf->mad) &&
-	     count < sizeof (packet->mad) + sizeof (*recv_buf->mad)))
+	     count < hdr_size(file) + sizeof (*recv_buf->mad)))
 		return -EINVAL;
 
-	if (copy_to_user(buf, &packet->mad, sizeof (packet->mad)))
+	if (copy_to_user(buf, &packet->mad, hdr_size(file)))
 		return -EFAULT;
 
-	buf += sizeof (packet->mad);
+	buf += hdr_size(file);
 	seg_payload = min_t(int, packet->length, sizeof (*recv_buf->mad));
 	if (copy_to_user(buf, recv_buf->mad, seg_payload))
 		return -EFAULT;
···
 		 * Multipacket RMPP MAD message. Copy remainder of message.
 		 * Note that last segment may have a shorter payload.
 		 */
-		if (count < sizeof (packet->mad) + packet->length) {
+		if (count < hdr_size(file) + packet->length) {
 			/*
 			 * The buffer is too small, return the first RMPP segment,
 			 * which includes the RMPP message length.
···
 				return -EFAULT;
 		}
 	}
-	return sizeof (packet->mad) + packet->length;
+	return hdr_size(file) + packet->length;
 }
 
-static ssize_t copy_send_mad(char __user *buf, struct ib_umad_packet *packet,
-			     size_t count)
+static ssize_t copy_send_mad(struct ib_umad_file *file, char __user *buf,
+			     struct ib_umad_packet *packet, size_t count)
 {
-	ssize_t size = sizeof (packet->mad) + packet->length;
+	ssize_t size = hdr_size(file) + packet->length;
 
 	if (count < size)
 		return -EINVAL;
 
-	if (copy_to_user(buf, &packet->mad, size))
+	if (copy_to_user(buf, &packet->mad, hdr_size(file)))
+		return -EFAULT;
+
+	buf += hdr_size(file);
+
+	if (copy_to_user(buf, packet->mad.data, packet->length))
 		return -EFAULT;
 
 	return size;
···
 	struct ib_umad_packet *packet;
 	ssize_t ret;
 
-	if (count < sizeof (struct ib_user_mad))
+	if (count < hdr_size(file))
 		return -EINVAL;
 
 	spin_lock_irq(&file->recv_lock);
···
 	spin_unlock_irq(&file->recv_lock);
 
 	if (packet->recv_wc)
-		ret = copy_recv_mad(buf, packet, count);
+		ret = copy_recv_mad(file, buf, packet, count);
 	else
-		ret = copy_send_mad(buf, packet, count);
+		ret = copy_send_mad(file, buf, packet, count);
 
 	if (ret < 0) {
 		/* Requeue packet */
···
 	__be64 *tid;
 	int ret, data_len, hdr_len, copy_offset, rmpp_active;
 
-	if (count < sizeof (struct ib_user_mad) + IB_MGMT_RMPP_HDR)
+	if (count < hdr_size(file) + IB_MGMT_RMPP_HDR)
 		return -EINVAL;
 
 	packet = kzalloc(sizeof *packet + IB_MGMT_RMPP_HDR, GFP_KERNEL);
 	if (!packet)
 		return -ENOMEM;
 
-	if (copy_from_user(&packet->mad, buf,
-			    sizeof (struct ib_user_mad) + IB_MGMT_RMPP_HDR)) {
+	if (copy_from_user(&packet->mad, buf, hdr_size(file))) {
 		ret = -EFAULT;
 		goto err;
 	}
···
 	if (packet->mad.hdr.id < 0 ||
 	    packet->mad.hdr.id >= IB_UMAD_MAX_AGENTS) {
 		ret = -EINVAL;
+		goto err;
+	}
+
+	buf += hdr_size(file);
+
+	if (copy_from_user(packet->mad.data, buf, IB_MGMT_RMPP_HDR)) {
+		ret = -EFAULT;
 		goto err;
 	}
 
···
 			      IB_MGMT_RMPP_FLAG_ACTIVE;
 	}
 
-	data_len = count - sizeof (struct ib_user_mad) - hdr_len;
+	data_len = count - hdr_size(file) - hdr_len;
 	packet->msg = ib_create_send_mad(agent,
 					 be32_to_cpu(packet->mad.hdr.qpn),
-					 0, rmpp_active, hdr_len,
-					 data_len, GFP_KERNEL);
+					 packet->mad.hdr.pkey_index, rmpp_active,
+					 hdr_len, data_len, GFP_KERNEL);
 	if (IS_ERR(packet->msg)) {
 		ret = PTR_ERR(packet->msg);
 		goto err_ah;
···
 
 	/* Copy MAD header.  Any RMPP header is already in place. */
 	memcpy(packet->msg->mad, packet->mad.data, IB_MGMT_MAD_HDR);
-	buf += sizeof (struct ib_user_mad);
 
 	if (!rmpp_active) {
 		if (copy_from_user(packet->msg->mad + copy_offset,
···
 		goto out;
 	}
 
+	if (!file->already_used) {
+		file->already_used = 1;
+		if (!file->use_pkey_index) {
+			printk(KERN_WARNING "user_mad: process %s did not enable "
+			       "P_Key index support.\n", current->comm);
+			printk(KERN_WARNING "user_mad:   Documentation/infiniband/user_mad.txt "
+			       "has info on the new ABI.\n");
+		}
+	}
+
 	file->agent[agent_id] = agent;
 	ret = 0;
 
···
 	return ret;
 }
 
+static long ib_umad_enable_pkey(struct ib_umad_file *file)
+{
+	int ret = 0;
+
+	down_write(&file->port->mutex);
+	if (file->already_used)
+		ret = -EINVAL;
+	else
+		file->use_pkey_index = 1;
+	up_write(&file->port->mutex);
+
+	return ret;
+}
+
 static long ib_umad_ioctl(struct file *filp, unsigned int cmd,
 			  unsigned long arg)
 {
···
 		return ib_umad_reg_agent(filp->private_data, arg);
 	case IB_USER_MAD_UNREGISTER_AGENT:
 		return ib_umad_unreg_agent(filp->private_data, arg);
+	case IB_USER_MAD_ENABLE_PKEY:
+		return ib_umad_enable_pkey(filp->private_data);
 	default:
 		return -ENOIOCTLCMD;
 	}

include/rdma/ib_user_mad.h

···
  */
 
 /**
- * ib_user_mad_hdr - MAD packet header
+ * ib_user_mad_hdr_old - Old version of MAD packet header without pkey_index
  * @id - ID of agent MAD received with/to be sent with
  * @status - 0 on successful receive, ETIMEDOUT if no response
  *   received (transaction ID in data[] will be set to TID of original
···
  * @gid - Remote GID in GRH
  * @flow_label - Flow label in GRH
  */
+struct ib_user_mad_hdr_old {
+	__u32	id;
+	__u32	status;
+	__u32	timeout_ms;
+	__u32	retries;
+	__u32	length;
+	__be32	qpn;
+	__be32  qkey;
+	__be16	lid;
+	__u8	sl;
+	__u8	path_bits;
+	__u8	grh_present;
+	__u8	gid_index;
+	__u8	hop_limit;
+	__u8	traffic_class;
+	__u8	gid[16];
+	__be32	flow_label;
+};
+
+/**
+ * ib_user_mad_hdr - MAD packet header
+ *   This layout allows specifying/receiving the P_Key index.  To use
+ *   this capability, an application must call the
+ *   IB_USER_MAD_ENABLE_PKEY ioctl on the user MAD file handle before
+ *   any other actions with the file handle.
+ * @id - ID of agent MAD received with/to be sent with
+ * @status - 0 on successful receive, ETIMEDOUT if no response
+ *   received (transaction ID in data[] will be set to TID of original
+ *   request) (ignored on send)
+ * @timeout_ms - Milliseconds to wait for response (unset on receive)
+ * @retries - Number of automatic retries to attempt
+ * @qpn - Remote QP number received from/to be sent to
+ * @qkey - Remote Q_Key to be sent with (unset on receive)
+ * @lid - Remote lid received from/to be sent to
+ * @sl - Service level received with/to be sent with
+ * @path_bits - Local path bits received with/to be sent with
+ * @grh_present - If set, GRH was received/should be sent
+ * @gid_index - Local GID index to send with (unset on receive)
+ * @hop_limit - Hop limit in GRH
+ * @traffic_class - Traffic class in GRH
+ * @gid - Remote GID in GRH
+ * @flow_label - Flow label in GRH
+ * @pkey_index - P_Key index
+ */
 struct ib_user_mad_hdr {
 	__u32	id;
 	__u32	status;
···
 	__u8	traffic_class;
 	__u8	gid[16];
 	__be32	flow_label;
+	__u16	pkey_index;
+	__u8	reserved[6];
 };
 
 /**
···
 					      struct ib_user_mad_reg_req)
 
 #define IB_USER_MAD_UNREGISTER_AGENT	_IOW(IB_IOCTL_MAGIC, 2, __u32)
+
+#define IB_USER_MAD_ENABLE_PKEY		_IO(IB_IOCTL_MAGIC, 3)
 
 #endif /* IB_USER_MAD_H */