Merge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/roland/infiniband

* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/roland/infiniband: (76 commits)
  IB: Update MAINTAINERS with Hal's new email address
  IB/mlx4: Implement query SRQ
  IB/mlx4: Implement query QP
  IB/cm: Send no match if a SIDR REQ does not match a listen
  IB/cm: Fix handling of duplicate SIDR REQs
  IB/cm: cm_msgs.h should include ib_cm.h
  IB/cm: Include HCA ACK delay in local ACK timeout
  IB/cm: Use spin_lock_irq() instead of spin_lock_irqsave() when possible
  IB/sa: Make sure SA queries use default P_Key
  IPoIB: Recycle loopback skbs instead of freeing and reallocating
  IB/mthca: Replace memset(<addr>, 0, PAGE_SIZE) with clear_page(<addr>)
  IPoIB/cm: Fix warning if IPV6 is not enabled
  IB/core: Take sizeof the correct pointer when calling kmalloc()
  IB/ehca: Improve latency by unlocking after triggering the hardware
  IB/ehca: Notify consumers of LID/PKEY/SM changes after nondisruptive events
  IB/ehca: Return QP pointer in poll_cq()
  IB/ehca: Change idr spinlocks into rwlocks
  IB/ehca: Refactor sync between completions and destroy_cq using atomic_t
  IB/ehca: Lock renaming, static initializers
  IB/ehca: Report RDMA atomic attributes in query_qp()
  ...

88 files changed, 2738 insertions, 1052 deletions

diff --git a/drivers/infiniband/Kconfig b/drivers/infiniband/Kconfig
index 994decc7bcf2..a193dfbf99d2 100644
--- a/drivers/infiniband/Kconfig
+++ b/drivers/infiniband/Kconfig
@@ -1,14 +1,14 @@
-menu "InfiniBand support"
-	depends on HAS_IOMEM
-
-config INFINIBAND
-	depends on PCI || BROKEN
+menuconfig INFINIBAND
 	tristate "InfiniBand support"
+	depends on PCI || BROKEN
+	depends on HAS_IOMEM
 	---help---
 	  Core support for InfiniBand (IB).  Make sure to also select
 	  any protocols you wish to use as well as drivers for your
 	  InfiniBand hardware.
 
+if INFINIBAND
+
 config INFINIBAND_USER_MAD
 	tristate "InfiniBand userspace MAD support"
 	depends on INFINIBAND
@@ -20,7 +20,6 @@ config INFINIBAND_USER_MAD
 
 config INFINIBAND_USER_ACCESS
 	tristate "InfiniBand userspace access (verbs and CM)"
-	depends on INFINIBAND
 	---help---
 	  Userspace InfiniBand access support.  This enables the
 	  kernel side of userspace verbs and the userspace
@@ -37,7 +36,7 @@ config INFINIBAND_USER_MEM
 
 config INFINIBAND_ADDR_TRANS
 	bool
-	depends on INFINIBAND && INET
+	depends on INET
 	default y
 
 source "drivers/infiniband/hw/mthca/Kconfig"
@@ -54,4 +53,4 @@ source "drivers/infiniband/ulp/srp/Kconfig"
 
 source "drivers/infiniband/ulp/iser/Kconfig"
 
-endmenu
+endif # INFINIBAND
diff --git a/drivers/infiniband/core/agent.c b/drivers/infiniband/core/agent.c
index ecd1a3057c61..db2633e4aae6 100644
--- a/drivers/infiniband/core/agent.c
+++ b/drivers/infiniband/core/agent.c
@@ -3,7 +3,7 @@
  * Copyright (c) 2004, 2005 Infinicon Corporation.  All rights reserved.
  * Copyright (c) 2004, 2005 Intel Corporation.  All rights reserved.
  * Copyright (c) 2004, 2005 Topspin Corporation.  All rights reserved.
- * Copyright (c) 2004, 2005 Voltaire Corporation.  All rights reserved.
+ * Copyright (c) 2004-2007 Voltaire Corporation.  All rights reserved.
  * Copyright (c) 2005 Sun Microsystems, Inc. All rights reserved.
  *
  * This software is available to you under a choice of one of two
@@ -34,7 +34,6 @@
  * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
  * SOFTWARE.
  *
- * $Id: agent.c 1389 2004-12-27 22:56:47Z roland $
  */
 
 #include <linux/slab.h>
@@ -42,6 +41,7 @@
 
 #include "agent.h"
 #include "smi.h"
+#include "mad_priv.h"
 
 #define SPFX "ib_agent: "
 
@@ -87,8 +87,13 @@ int agent_send_response(struct ib_mad *mad, struct ib_grh *grh,
 	struct ib_mad_send_buf *send_buf;
 	struct ib_ah *ah;
 	int ret;
+	struct ib_mad_send_wr_private *mad_send_wr;
+
+	if (device->node_type == RDMA_NODE_IB_SWITCH)
+		port_priv = ib_get_agent_port(device, 0);
+	else
+		port_priv = ib_get_agent_port(device, port_num);
 
-	port_priv = ib_get_agent_port(device, port_num);
 	if (!port_priv) {
 		printk(KERN_ERR SPFX "Unable to find port agent\n");
 		return -ENODEV;
@@ -113,6 +118,14 @@ int agent_send_response(struct ib_mad *mad, struct ib_grh *grh,
 
 	memcpy(send_buf->mad, mad, sizeof *mad);
 	send_buf->ah = ah;
+
+	if (device->node_type == RDMA_NODE_IB_SWITCH) {
+		mad_send_wr = container_of(send_buf,
+					   struct ib_mad_send_wr_private,
+					   send_buf);
+		mad_send_wr->send_wr.wr.ud.port_num = port_num;
+	}
+
 	if ((ret = ib_post_send_mad(send_buf, NULL))) {
 		printk(KERN_ERR SPFX "ib_post_send_mad error:%d\n", ret);
 		goto err2;
diff --git a/drivers/infiniband/core/cm.c b/drivers/infiniband/core/cm.c
index 40c004a2697e..9820c67ba47d 100644
--- a/drivers/infiniband/core/cm.c
+++ b/drivers/infiniband/core/cm.c
@@ -87,6 +87,7 @@ struct cm_port {
 struct cm_device {
 	struct list_head list;
 	struct ib_device *device;
+	u8 ack_delay;
 	struct cm_port port[0];
 };
 
@@ -95,7 +96,7 @@ struct cm_av {
 	union ib_gid dgid;
 	struct ib_ah_attr ah_attr;
 	u16 pkey_index;
-	u8 packet_life_time;
+	u8 timeout;
 };
 
 struct cm_work {
@@ -154,6 +155,7 @@ struct cm_id_private {
 	u8 retry_count;
 	u8 rnr_retry_count;
 	u8 service_timeout;
+	u8 target_ack_delay;
 
 	struct list_head work_list;
 	atomic_t work_count;
@@ -293,7 +295,7 @@ static int cm_init_av_by_path(struct ib_sa_path_rec *path, struct cm_av *av)
 	av->port = port;
 	ib_init_ah_from_path(cm_dev->device, port->port_num, path,
 			     &av->ah_attr);
-	av->packet_life_time = path->packet_life_time;
+	av->timeout = path->packet_life_time + 1;
 	return 0;
 }
 
@@ -318,12 +320,10 @@ static int cm_alloc_id(struct cm_id_private *cm_id_priv)
 
 static void cm_free_id(__be32 local_id)
 {
-	unsigned long flags;
-
-	spin_lock_irqsave(&cm.lock, flags);
+	spin_lock_irq(&cm.lock);
 	idr_remove(&cm.local_id_table,
 		   (__force int) (local_id ^ cm.random_id_operand));
-	spin_unlock_irqrestore(&cm.lock, flags);
+	spin_unlock_irq(&cm.lock);
 }
 
 static struct cm_id_private * cm_get_id(__be32 local_id, __be32 remote_id)
@@ -345,11 +345,10 @@ static struct cm_id_private * cm_get_id(__be32 local_id, __be32 remote_id)
 static struct cm_id_private * cm_acquire_id(__be32 local_id, __be32 remote_id)
 {
 	struct cm_id_private *cm_id_priv;
-	unsigned long flags;
 
-	spin_lock_irqsave(&cm.lock, flags);
+	spin_lock_irq(&cm.lock);
 	cm_id_priv = cm_get_id(local_id, remote_id);
-	spin_unlock_irqrestore(&cm.lock, flags);
+	spin_unlock_irq(&cm.lock);
 
 	return cm_id_priv;
 }
@@ -646,6 +645,25 @@ static inline int cm_convert_to_ms(int iba_time)
 	return 1 << max(iba_time - 8, 0);
 }
 
+/*
+ * calculate: 4.096x2^ack_timeout = 4.096x2^ack_delay + 2x4.096x2^life_time
+ * Because of how ack_timeout is stored, adding one doubles the timeout.
+ * To avoid large timeouts, select the max(ack_delay, life_time + 1), and
+ * increment it (round up) only if the other is within 50%.
+ */
+static u8 cm_ack_timeout(u8 ca_ack_delay, u8 packet_life_time)
+{
+	int ack_timeout = packet_life_time + 1;
+
+	if (ack_timeout >= ca_ack_delay)
+		ack_timeout += (ca_ack_delay >= (ack_timeout - 1));
+	else
+		ack_timeout = ca_ack_delay +
+			      (ack_timeout >= (ca_ack_delay - 1));
+
+	return min(31, ack_timeout);
+}
+
 static void cm_cleanup_timewait(struct cm_timewait_info *timewait_info)
 {
 	if (timewait_info->inserted_remote_id) {
@@ -689,7 +707,7 @@ static void cm_enter_timewait(struct cm_id_private *cm_id_priv)
 	 * timewait before notifying the user that we've exited timewait.
 	 */
 	cm_id_priv->id.state = IB_CM_TIMEWAIT;
-	wait_time = cm_convert_to_ms(cm_id_priv->av.packet_life_time + 1);
+	wait_time = cm_convert_to_ms(cm_id_priv->av.timeout);
 	queue_delayed_work(cm.wq, &cm_id_priv->timewait_info->work.work,
 			   msecs_to_jiffies(wait_time));
 	cm_id_priv->timewait_info = NULL;
@@ -713,31 +731,30 @@ static void cm_destroy_id(struct ib_cm_id *cm_id, int err)
 {
 	struct cm_id_private *cm_id_priv;
 	struct cm_work *work;
-	unsigned long flags;
 
 	cm_id_priv = container_of(cm_id, struct cm_id_private, id);
 retest:
-	spin_lock_irqsave(&cm_id_priv->lock, flags);
+	spin_lock_irq(&cm_id_priv->lock);
 	switch (cm_id->state) {
 	case IB_CM_LISTEN:
 		cm_id->state = IB_CM_IDLE;
-		spin_unlock_irqrestore(&cm_id_priv->lock, flags);
-		spin_lock_irqsave(&cm.lock, flags);
+		spin_unlock_irq(&cm_id_priv->lock);
+		spin_lock_irq(&cm.lock);
 		rb_erase(&cm_id_priv->service_node, &cm.listen_service_table);
-		spin_unlock_irqrestore(&cm.lock, flags);
+		spin_unlock_irq(&cm.lock);
 		break;
 	case IB_CM_SIDR_REQ_SENT:
 		cm_id->state = IB_CM_IDLE;
 		ib_cancel_mad(cm_id_priv->av.port->mad_agent, cm_id_priv->msg);
-		spin_unlock_irqrestore(&cm_id_priv->lock, flags);
+		spin_unlock_irq(&cm_id_priv->lock);
 		break;
 	case IB_CM_SIDR_REQ_RCVD:
-		spin_unlock_irqrestore(&cm_id_priv->lock, flags);
+		spin_unlock_irq(&cm_id_priv->lock);
 		cm_reject_sidr_req(cm_id_priv, IB_SIDR_REJECT);
 		break;
 	case IB_CM_REQ_SENT:
 		ib_cancel_mad(cm_id_priv->av.port->mad_agent, cm_id_priv->msg);
-		spin_unlock_irqrestore(&cm_id_priv->lock, flags);
+		spin_unlock_irq(&cm_id_priv->lock);
 		ib_send_cm_rej(cm_id, IB_CM_REJ_TIMEOUT,
 			       &cm_id_priv->id.device->node_guid,
 			       sizeof cm_id_priv->id.device->node_guid,
@@ -747,9 +764,9 @@ retest:
 		if (err == -ENOMEM) {
 			/* Do not reject to allow future retries. */
 			cm_reset_to_idle(cm_id_priv);
-			spin_unlock_irqrestore(&cm_id_priv->lock, flags);
+			spin_unlock_irq(&cm_id_priv->lock);
 		} else {
-			spin_unlock_irqrestore(&cm_id_priv->lock, flags);
+			spin_unlock_irq(&cm_id_priv->lock);
 			ib_send_cm_rej(cm_id, IB_CM_REJ_CONSUMER_DEFINED,
 				       NULL, 0, NULL, 0);
 		}
@@ -762,25 +779,25 @@ retest:
 	case IB_CM_MRA_REQ_SENT:
 	case IB_CM_REP_RCVD:
 	case IB_CM_MRA_REP_SENT:
-		spin_unlock_irqrestore(&cm_id_priv->lock, flags);
+		spin_unlock_irq(&cm_id_priv->lock);
 		ib_send_cm_rej(cm_id, IB_CM_REJ_CONSUMER_DEFINED,
 			       NULL, 0, NULL, 0);
 		break;
 	case IB_CM_ESTABLISHED:
-		spin_unlock_irqrestore(&cm_id_priv->lock, flags);
+		spin_unlock_irq(&cm_id_priv->lock);
 		ib_send_cm_dreq(cm_id, NULL, 0);
 		goto retest;
 	case IB_CM_DREQ_SENT:
 		ib_cancel_mad(cm_id_priv->av.port->mad_agent, cm_id_priv->msg);
 		cm_enter_timewait(cm_id_priv);
-		spin_unlock_irqrestore(&cm_id_priv->lock, flags);
+		spin_unlock_irq(&cm_id_priv->lock);
 		break;
 	case IB_CM_DREQ_RCVD:
-		spin_unlock_irqrestore(&cm_id_priv->lock, flags);
+		spin_unlock_irq(&cm_id_priv->lock);
 		ib_send_cm_drep(cm_id, NULL, 0);
 		break;
 	default:
-		spin_unlock_irqrestore(&cm_id_priv->lock, flags);
+		spin_unlock_irq(&cm_id_priv->lock);
 		break;
 	}
 
@@ -912,7 +929,8 @@ static void cm_format_req(struct cm_req_msg *req_msg,
 	cm_req_set_primary_sl(req_msg, param->primary_path->sl);
 	cm_req_set_primary_subnet_local(req_msg, 1); /* local only... */
 	cm_req_set_primary_local_ack_timeout(req_msg,
-		min(31, param->primary_path->packet_life_time + 1));
+		cm_ack_timeout(cm_id_priv->av.port->cm_dev->ack_delay,
+			       param->primary_path->packet_life_time));
 
 	if (param->alternate_path) {
 		req_msg