[PATCH RESEND 0/5] ib/mad: Add ability for user space applications to control RMPP directly

[PATCH RESEND 0/5] ib/mad: Add ability for user space applications to control RMPP directly

[ira.weiny-ral2JQCrhuEAvxtiuMwx3w]

My apologies if anyone receives duplicates of these emails.  Apparently the
series did not make it to linux-rdma the first time.

The first 3 patches update the umad and mad modules to use pr_* style print
messages in preparation for adding pr_notice messages to the existing and new
registration functions.

The final 2 patches add a new registration mechanism which allows for flags to
be specified at ib_mad registration time.  The first such flag is to allow user
control of RMPP transactions.

Existing registration and RMPP transactions are unaffected by this change.

[PATCH 1/5] ib/umad: update module to pr_* style print messages
[PATCH 2/5] ib/mad: update module to pr_* style print messages
[PATCH 3/5] ib/mad: Add pr_notice messages for various umad/mad
[PATCH 4/5] ib/mad: add new ioctl to user space to support new
[PATCH 5/5] ib/mad: Add user space RMPP support

--
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

[PATCH 1/5] ib/umad: update module to pr_* style print messages

[ira.weiny-ral2JQCrhuEAvxtiuMwx3w]

From: Ira Weiny <ira.weiny-ral2JQCrhuEAvxtiuMwx3w@public.gmane.org>

Also combine previously line broken user-visible strings as per
Documentation/CodingStyle

"However, never break user-visible strings such as printk messages, because
that breaks the ability to grep for them."

Reviewed-by: Sean Hefty <sean.hefty-ral2JQCrhuEAvxtiuMwx3w@public.gmane.org>
Reviewed-by: Hal Rosenstock <hal-LDSdmyG8hGV8YrgS2mwiifqBs+8SCbDb@public.gmane.org>
Signed-off-by: Ira Weiny <ira.weiny-ral2JQCrhuEAvxtiuMwx3w@public.gmane.org>
---
 drivers/infiniband/core/user_mad.c | 19 ++++++++++---------
 1 file changed, 10 insertions(+), 9 deletions(-)

diff --git a/drivers/infiniband/core/user_mad.c b/drivers/infiniband/core/user_mad.c
index f0d588f..0df292a 100644
--- a/drivers/infiniband/core/user_mad.c
+++ b/drivers/infiniband/core/user_mad.c
@@ -33,6 +33,8 @@
  * SOFTWARE.
  */
 
+#define pr_fmt(fmt) "user_mad: " fmt
+
 #include <linux/module.h>
 #include <linux/init.h>
 #include <linux/device.h>
@@ -669,10 +671,9 @@ found:
 	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");
+			pr_warn("process %s did not enable P_Key index support.\n",
+				current->comm);
+			pr_warn("   Documentation/infiniband/user_mad.txt has info on the new ABI.\n");
 		}
 	}
 
@@ -982,7 +983,7 @@ static int find_overflow_devnum(void)
 		ret = alloc_chrdev_region(&overflow_maj, 0, IB_UMAD_MAX_PORTS * 2,
 					  "infiniband_mad");
 		if (ret) {
-			printk(KERN_ERR "user_mad: couldn't register dynamic device number\n");
+			pr_err("couldn't register dynamic device number\n");
 			return ret;
 		}
 	}
@@ -1187,14 +1188,14 @@ static int __init ib_umad_init(void)
 	ret = register_chrdev_region(base_dev, IB_UMAD_MAX_PORTS * 2,
 				     "infiniband_mad");
 	if (ret) {
-		printk(KERN_ERR "user_mad: couldn't register device number\n");
+		pr_err("couldn't register device number\n");
 		goto out;
 	}
 
 	umad_class = class_create(THIS_MODULE, "infiniband_mad");
 	if (IS_ERR(umad_class)) {
 		ret = PTR_ERR(umad_class);
-		printk(KERN_ERR "user_mad: couldn't create class infiniband_mad\n");
+		pr_err("couldn't create class infiniband_mad\n");
 		goto out_chrdev;
 	}
 
@@ -1202,13 +1203,13 @@ static int __init ib_umad_init(void)
 
 	ret = class_create_file(umad_class, &class_attr_abi_version.attr);
 	if (ret) {
-		printk(KERN_ERR "user_mad: couldn't create abi_version attribute\n");
+		pr_err("couldn't create abi_version attribute\n");
 		goto out_class;
 	}
 
 	ret = ib_register_client(&umad_client);
 	if (ret) {
-		printk(KERN_ERR "user_mad: couldn't register ib_umad client\n");
+		pr_err("couldn't register ib_umad client\n");
 		goto out_class;
 	}
 
-- 
1.8.2

--
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

[PATCH 2/5] ib/mad: update module to pr_* style print messages

[ira.weiny-ral2JQCrhuEAvxtiuMwx3w]

From: Ira Weiny <ira.weiny-ral2JQCrhuEAvxtiuMwx3w@public.gmane.org>

Also combine previously line broken user-visible strings as per
Documentation/CodingStyle

"However, never break user-visible strings such as printk messages, because
that breaks the ability to grep for them."

Reviewed-by: Sean Hefty <sean.hefty-ral2JQCrhuEAvxtiuMwx3w@public.gmane.org>
Reviewed-by: Hal Rosenstock <hal-LDSdmyG8hGV8YrgS2mwiifqBs+8SCbDb@public.gmane.org>
Signed-off-by: Ira Weiny <ira.weiny-ral2JQCrhuEAvxtiuMwx3w@public.gmane.org>
---
 drivers/infiniband/core/agent.c |  14 +++---
 drivers/infiniband/core/mad.c   | 109 +++++++++++++++++++---------------------
 2 files changed, 59 insertions(+), 64 deletions(-)

diff --git a/drivers/infiniband/core/agent.c b/drivers/infiniband/core/agent.c
index 2bc7f5a..e1f08b4 100644
--- a/drivers/infiniband/core/agent.c
+++ b/drivers/infiniband/core/agent.c
@@ -36,6 +36,8 @@
  *
  */
 
+#define pr_fmt(fmt) SPFX fmt
+
 #include <linux/slab.h>
 #include <linux/string.h>
 
@@ -94,14 +96,14 @@ void agent_send_response(struct ib_mad *mad, struct ib_grh *grh,
 		port_priv = ib_get_agent_port(device, port_num);
 
 	if (!port_priv) {
-		printk(KERN_ERR SPFX "Unable to find port agent\n");
+		pr_err("Unable to find port agent\n");
 		return;
 	}
 
 	agent = port_priv->agent[qpn];
 	ah = ib_create_ah_from_wc(agent->qp->pd, wc, grh, port_num);
 	if (IS_ERR(ah)) {
-		printk(KERN_ERR SPFX "ib_create_ah_from_wc error %ld\n",
+		pr_err("ib_create_ah_from_wc error %ld\n",
 			PTR_ERR(ah));
 		return;
 	}
@@ -110,7 +112,7 @@ void agent_send_response(struct ib_mad *mad, struct ib_grh *grh,
 				      IB_MGMT_MAD_HDR, IB_MGMT_MAD_DATA,
 				      GFP_KERNEL);
 	if (IS_ERR(send_buf)) {
-		printk(KERN_ERR SPFX "ib_create_send_mad error\n");
+		pr_err("ib_create_send_mad error\n");
 		goto err1;
 	}
 
@@ -125,7 +127,7 @@ void agent_send_response(struct ib_mad *mad, struct ib_grh *grh,
 	}
 
 	if (ib_post_send_mad(send_buf, NULL)) {
-		printk(KERN_ERR SPFX "ib_post_send_mad error\n");
+		pr_err("ib_post_send_mad error\n");
 		goto err2;
 	}
 	return;
@@ -151,7 +153,7 @@ int ib_agent_port_open(struct ib_device *device, int port_num)
 	/* Create new device info */
 	port_priv = kzalloc(sizeof *port_priv, GFP_KERNEL);
 	if (!port_priv) {
-		printk(KERN_ERR SPFX "No memory for ib_agent_port_private\n");
+		pr_err("No memory for ib_agent_port_private\n");
 		ret = -ENOMEM;
 		goto error1;
 	}
@@ -202,7 +204,7 @@ int ib_agent_port_close(struct ib_device *device, int port_num)
 	port_priv = __ib_get_agent_port(device, port_num);
 	if (port_priv == NULL) {
 		spin_unlock_irqrestore(&ib_agent_port_list_lock, flags);
-		printk(KERN_ERR SPFX "Port %d not found\n", port_num);
+		pr_err("Port %d not found\n", port_num);
 		return -ENODEV;
 	}
 	list_del(&port_priv->port_list);
diff --git a/drivers/infiniband/core/mad.c b/drivers/infiniband/core/mad.c
index ab31f13..7395e23 100644
--- a/drivers/infiniband/core/mad.c
+++ b/drivers/infiniband/core/mad.c
@@ -33,6 +33,9 @@
  * SOFTWARE.
  *
  */
+
+#define pr_fmt(fmt) PFX fmt
+
 #include <linux/dma-mapping.h>
 #include <linux/slab.h>
 #include <linux/module.h>
@@ -706,7 +709,7 @@ static int handle_outgoing_dr_smp(struct ib_mad_agent_private *mad_agent_priv,
 	     smi_handle_dr_smp_send(smp, device->node_type, port_num) ==
 	     IB_SMI_DISCARD) {
 		ret = -EINVAL;
-		printk(KERN_ERR PFX "Invalid directed route\n");
+		pr_err("Invalid directed route\n");
 		goto out;
 	}
 
@@ -718,7 +721,7 @@ static int handle_outgoing_dr_smp(struct ib_mad_agent_private *mad_agent_priv,
 	local = kmalloc(sizeof *local, GFP_ATOMIC);
 	if (!local) {
 		ret = -ENOMEM;
-		printk(KERN_ERR PFX "No memory for ib_mad_local_private\n");
+		pr_err("No memory for ib_mad_local_private\n");
 		goto out;
 	}
 	local->mad_priv = NULL;
@@ -726,7 +729,7 @@ static int handle_outgoing_dr_smp(struct ib_mad_agent_private *mad_agent_priv,
 	mad_priv = kmem_cache_alloc(ib_mad_cache, GFP_ATOMIC);
 	if (!mad_priv) {
 		ret = -ENOMEM;
-		printk(KERN_ERR PFX "No memory for local response MAD\n");
+		pr_err("No memory for local response MAD\n");
 		kfree(local);
 		goto out;
 	}
@@ -837,8 +840,7 @@ static int alloc_send_rmpp_list(struct ib_mad_send_wr_private *send_wr,
 	for (left = send_buf->data_len + pad; left > 0; left -= seg_size) {
 		seg = kmalloc(sizeof (*seg) + seg_size, gfp_mask);
 		if (!seg) {
-			printk(KERN_ERR "alloc_send_rmpp_segs: RMPP mem "
-			       "alloc failed for len %zd, gfp %#x\n",
+			pr_err("alloc_send_rmpp_segs: RMPP mem alloc failed for len %zd, gfp %#x\n",
 			       sizeof (*seg) + seg_size, gfp_mask);
 			free_send_rmpp_list(send_wr);
 			return -ENOMEM;
@@ -1199,7 +1201,7 @@ EXPORT_SYMBOL(ib_redirect_mad_qp);
 int ib_process_mad_wc(struct ib_mad_agent *mad_agent,
 		      struct ib_wc *wc)
 {
-	printk(KERN_ERR PFX "ib_process_mad_wc() not implemented yet\n");
+	pr_err("ib_process_mad_wc() not implemented yet\n");
 	return 0;
 }
 EXPORT_SYMBOL(ib_process_mad_wc);
@@ -1211,7 +1213,7 @@ static int method_in_use(struct ib_mad_mgmt_method_table **method,
 
 	for_each_set_bit(i, mad_reg_req->method_mask, IB_MGMT_MAX_METHODS) {
 		if ((*method)->agent[i]) {
-			printk(KERN_ERR PFX "Method %d already in use\n", i);
+			pr_err("Method %d already in use\n", i);
 			return -EINVAL;
 		}
 	}
@@ -1223,8 +1225,7 @@ static int allocate_method_table(struct ib_mad_mgmt_method_table **method)
 	/* Allocate management method table */
 	*method = kzalloc(sizeof **method, GFP_ATOMIC);
 	if (!*method) {
-		printk(KERN_ERR PFX "No memory for "
-		       "ib_mad_mgmt_method_table\n");
+		pr_err("No memory for ib_mad_mgmt_method_table\n");
 		return -ENOMEM;
 	}
 
@@ -1319,8 +1320,7 @@ static int add_nonoui_reg_req(struct ib_mad_reg_req *mad_reg_req,
 		/* Allocate management class table for "new" class version */
 		*class = kzalloc(sizeof **class, GFP_ATOMIC);
 		if (!*class) {
-			printk(KERN_ERR PFX "No memory for "
-			       "ib_mad_mgmt_class_table\n");
+			pr_err("No memory for ib_mad_mgmt_class_table\n");
 			ret = -ENOMEM;
 			goto error1;
 		}
@@ -1386,8 +1386,7 @@ static int add_oui_reg_req(struct ib_mad_reg_req *mad_reg_req,
 		/* Allocate mgmt vendor class table for "new" class version */
 		vendor = kzalloc(sizeof *vendor, GFP_ATOMIC);
 		if (!vendor) {
-			printk(KERN_ERR PFX "No memory for "
-			       "ib_mad_mgmt_vendor_class_table\n");
+			pr_err("No memory for ib_mad_mgmt_vendor_class_table\n");
 			goto error1;
 		}
 
@@ -1397,8 +1396,7 @@ static int add_oui_reg_req(struct ib_mad_reg_req *mad_reg_req,
 		/* Allocate table for this management vendor class */
 		vendor_class = kzalloc(sizeof *vendor_class, GFP_ATOMIC);
 		if (!vendor_class) {
-			printk(KERN_ERR PFX "No memory for "
-			       "ib_mad_mgmt_vendor_class\n");
+			pr_err("No memory for ib_mad_mgmt_vendor_class\n");
 			goto error2;
 		}
 
@@ -1429,7 +1427,7 @@ static int add_oui_reg_req(struct ib_mad_reg_req *mad_reg_req,
 			goto check_in_use;
 		}
 	}
-	printk(KERN_ERR PFX "All OUI slots in use\n");
+	pr_err("All OUI slots in use\n");
 	goto error3;
 
 check_in_use:
@@ -1640,8 +1638,7 @@ find_mad_agent(struct ib_mad_port_private *port_priv,
 		if (mad_agent->agent.recv_handler)
 			atomic_inc(&mad_agent->refcount);
 		else {
-			printk(KERN_NOTICE PFX "No receive handler for client "
-			       "%p on port %d\n",
+			pr_notice("No receive handler for client %p on port %d\n",
 			       &mad_agent->agent, port_priv->port_num);
 			mad_agent = NULL;
 		}
@@ -1658,8 +1655,8 @@ static int validate_mad(struct ib_mad *mad, u32 qp_num)
 
 	/* Make sure MAD base version is understood */
 	if (mad->mad_hdr.base_version != IB_MGMT_BASE_VERSION) {
-		printk(KERN_ERR PFX "MAD received with unsupported base "
-		       "version %d\n", mad->mad_hdr.base_version);
+		pr_err("MAD received with unsupported base version %d\n",
+			mad->mad_hdr.base_version);
 		goto out;
 	}
 
@@ -1911,8 +1908,7 @@ static void ib_mad_recv_done_handler(struct ib_mad_port_private *port_priv,
 
 	response = kmem_cache_alloc(ib_mad_cache, GFP_KERNEL);
 	if (!response) {
-		printk(KERN_ERR PFX "ib_mad_recv_done_handler no memory "
-		       "for response buffer\n");
+		pr_err("ib_mad_recv_done_handler no memory for response buffer\n");
 		goto out;
 	}
 
@@ -2176,7 +2172,7 @@ retry:
 		ret = ib_post_send(qp_info->qp, &queued_send_wr->send_wr,
 				   &bad_send_wr);
 		if (ret) {
-			printk(KERN_ERR PFX "ib_post_send failed: %d\n", ret);
+			pr_err("ib_post_send failed: %d\n", ret);
 			mad_send_wr = queued_send_wr;
 			wc->status = IB_WC_LOC_QP_OP_ERR;
 			goto retry;
@@ -2248,8 +2244,8 @@ static void mad_error_handler(struct ib_mad_port_private *port_priv,
 					   IB_QP_STATE | IB_QP_CUR_STATE);
 			kfree(attr);
 			if (ret)
-				printk(KERN_ERR PFX "mad_error_handler - "
-				       "ib_modify_qp to RTS : %d\n", ret);
+				pr_err("mad_error_handler - ib_modify_qp to RTS : %d\n",
+					ret);
 			else
 				mark_sends_for_retry(qp_info);
 		}
@@ -2408,7 +2404,7 @@ static void local_completions(struct work_struct *work)
 		if (local->mad_priv) {
 			recv_mad_agent = local->recv_mad_agent;
 			if (!recv_mad_agent) {
-				printk(KERN_ERR PFX "No receive MAD agent for local completion\n");
+				pr_err("No receive MAD agent for local completion\n");
 				free_mad = 1;
 				goto local_send_completion;
 			}
@@ -2589,7 +2585,7 @@ static int ib_mad_post_receive_mads(struct ib_mad_qp_info *qp_info,
 		} else {
 			mad_priv = kmem_cache_alloc(ib_mad_cache, GFP_KERNEL);
 			if (!mad_priv) {
-				printk(KERN_ERR PFX "No memory for receive buffer\n");
+				pr_err("No memory for receive buffer\n");
 				ret = -ENOMEM;
 				break;
 			}
@@ -2625,7 +2621,7 @@ static int ib_mad_post_receive_mads(struct ib_mad_qp_info *qp_info,
 					      sizeof mad_priv->header,
 					    DMA_FROM_DEVICE);
 			kmem_cache_free(ib_mad_cache, mad_priv);
-			printk(KERN_ERR PFX "ib_post_recv failed: %d\n", ret);
+			pr_err("ib_post_recv failed: %d\n", ret);
 			break;
 		}
 	} while (post);
@@ -2681,7 +2677,7 @@ static int ib_mad_port_start(struct ib_mad_port_private *port_priv)
 
 	attr = kmalloc(sizeof *attr, GFP_KERNEL);
 	if (!attr) {
-		printk(KERN_ERR PFX "Couldn't kmalloc ib_qp_attr\n");
+		pr_err("Couldn't kmalloc ib_qp_attr\n");
 		return -ENOMEM;
 	}
 
@@ -2705,16 +2701,16 @@ static int ib_mad_port_start(struct ib_mad_port_private *port_priv)
 		ret = ib_modify_qp(qp, attr, IB_QP_STATE |
 					     IB_QP_PKEY_INDEX | IB_QP_QKEY);
 		if (ret) {
-			printk(KERN_ERR PFX "Couldn't change QP%d state to "
-			       "INIT: %d\n", i, ret);
+			pr_err("Couldn't change QP%d state to INIT: %d\n",
+				i, ret);
 			goto out;
 		}
 
 		attr->qp_state = IB_QPS_RTR;
 		ret = ib_modify_qp(qp, attr, IB_QP_STATE);
 		if (ret) {
-			printk(KERN_ERR PFX "Couldn't change QP%d state to "
-			       "RTR: %d\n", i, ret);
+			pr_err("Couldn't change QP%d state to RTR: %d\n",
+				i, ret);
 			goto out;
 		}
 
@@ -2722,16 +2718,16 @@ static int ib_mad_port_start(struct ib_mad_port_private *port_priv)
 		attr->sq_psn = IB_MAD_SEND_Q_PSN;
 		ret = ib_modify_qp(qp, attr, IB_QP_STATE | IB_QP_SQ_PSN);
 		if (ret) {
-			printk(KERN_ERR PFX "Couldn't change QP%d state to "
-			       "RTS: %d\n", i, ret);
+			pr_err("Couldn't change QP%d state to RTS: %d\n",
+				i, ret);
 			goto out;
 		}
 	}
 
 	ret = ib_req_notify_cq(port_priv->cq, IB_CQ_NEXT_COMP);
 	if (ret) {
-		printk(KERN_ERR PFX "Failed to request completion "
-		       "notification: %d\n", ret);
+		pr_err("Failed to request completion notification: %d\n",
+			ret);
 		goto out;
 	}
 
@@ -2741,7 +2737,7 @@ static int ib_mad_port_start(struct ib_mad_port_private *port_priv)
 
 		ret = ib_mad_post_receive_mads(&port_priv->qp_info[i], NULL);
 		if (ret) {
-			printk(KERN_ERR PFX "Couldn't post receive WRs\n");
+			pr_err("Couldn't post receive WRs\n");
 			goto out;
 		}
 	}
@@ -2755,7 +2751,7 @@ static void qp_event_handler(struct ib_event *event, void *qp_context)
 	struct ib_mad_qp_info	*qp_info = qp_context;
 
 	/* It's worse than that! He's dead, Jim! */
-	printk(KERN_ERR PFX "Fatal error (%d) on MAD QP (%d)\n",
+	pr_err("Fatal error (%d) on MAD QP (%d)\n",
 		event->event, qp_info->qp->qp_num);
 }
 
@@ -2801,7 +2797,7 @@ static int create_mad_qp(struct ib_mad_qp_info *qp_info,
 	qp_init_attr.event_handler = qp_event_handler;
 	qp_info->qp = ib_create_qp(qp_info->port_priv->pd, &qp_init_attr);
 	if (IS_ERR(qp_info->qp)) {
-		printk(KERN_ERR PFX "Couldn't create ib_mad QP%d\n",
+		pr_err("Couldn't create ib_mad QP%d\n",
 		       get_spl_qp_index(qp_type));
 		ret = PTR_ERR(qp_info->qp);
 		goto error;
@@ -2840,7 +2836,7 @@ static int ib_mad_port_open(struct ib_device *device,
 	/* Create new device info */
 	port_priv = kzalloc(sizeof *port_priv, GFP_KERNEL);
 	if (!port_priv) {
-		printk(KERN_ERR PFX "No memory for ib_mad_port_private\n");
+		pr_err("No memory for ib_mad_port_private\n");
 		return -ENOMEM;
 	}
 
@@ -2860,21 +2856,21 @@ static int ib_mad_port_open(struct ib_device *device,
 				     ib_mad_thread_completion_handler,
 				     NULL, port_priv, cq_size, 0);
 	if (IS_ERR(port_priv->cq)) {
-		printk(KERN_ERR PFX "Couldn't create ib_mad CQ\n");
+		pr_err("Couldn't create ib_mad CQ\n");
 		ret = PTR_ERR(port_priv->cq);
 		goto error3;
 	}
 
 	port_priv->pd = ib_alloc_pd(device);
 	if (IS_ERR(port_priv->pd)) {
-		printk(KERN_ERR PFX "Couldn't create ib_mad PD\n");
+		pr_err("Couldn't create ib_mad PD\n");
 		ret = PTR_ERR(port_priv->pd);
 		goto error4;
 	}
 
 	port_priv->mr = ib_get_dma_mr(port_priv->pd, IB_ACCESS_LOCAL_WRITE);
 	if (IS_ERR(port_priv->mr)) {
-		printk(KERN_ERR PFX "Couldn't get ib_mad DMA MR\n");
+		pr_err("Couldn't get ib_mad DMA MR\n");
 		ret = PTR_ERR(port_priv->mr);
 		goto error5;
 	}
@@ -2902,7 +2898,7 @@ static int ib_mad_port_open(struct ib_device *device,
 
 	ret = ib_mad_port_start(port_priv);
 	if (ret) {
-		printk(KERN_ERR PFX "Couldn't start port\n");
+		pr_err("Couldn't start port\n");
 		goto error9;
 	}
 
@@ -2946,7 +2942,7 @@ static int ib_mad_port_close(struct ib_device *device, int port_num)
 	port_priv = __ib_get_mad_port(device, port_num);
 	if (port_priv == NULL) {
 		spin_unlock_irqrestore(&ib_mad_port_list_lock, flags);
-		printk(KERN_ERR PFX "Port %d not found\n", port_num);
+		pr_err("Port %d not found\n", port_num);
 		return -ENODEV;
 	}
 	list_del_init(&port_priv->port_list);
@@ -2984,13 +2980,12 @@ static void ib_mad_init_device(struct ib_device *device)
 
 	for (i = start; i <= end; i++) {
 		if (ib_mad_port_open(device, i)) {
-			printk(KERN_ERR PFX "Couldn't open %s port %d\n",
+			pr_err("Couldn't open %s port %d\n",
 			       device->name, i);
 			goto error;
 		}
 		if (ib_agent_port_open(device, i)) {
-			printk(KERN_ERR PFX "Couldn't open %s port %d "
-			       "for agents\n",
+			pr_err("Couldn't open %s port %d for agents\n",
 			       device->name, i);
 			goto error_agent;
 		}
@@ -2999,7 +2994,7 @@ static void ib_mad_init_device(struct ib_device *device)
 
 error_agent:
 	if (ib_mad_port_close(device, i))
-		printk(KERN_ERR PFX "Couldn't close %s port %d\n",
+		pr_err("Couldn't close %s port %d\n",
 		       device->name, i);
 
 error:
@@ -3007,11 +3002,10 @@ error:
 
 	while (i >= start) {
 		if (ib_agent_port_close(device, i))
-			printk(KERN_ERR PFX "Couldn't close %s port %d "
-			       "for agents\n",
+			pr_err("Couldn't close %s port %d for agents\n",
 			       device->name, i);
 		if (ib_mad_port_close(device, i))
-			printk(KERN_ERR PFX "Couldn't close %s port %d\n",
+			pr_err("Couldn't close %s port %d\n",
 			       device->name, i);
 		i--;
 	}
@@ -3033,11 +3027,10 @@ static void ib_mad_remove_device(struct ib_device *device)
 	}
 	for (i = 0; i < num_ports; i++, cur_port++) {
 		if (ib_agent_port_close(device, cur_port))
-			printk(KERN_ERR PFX "Couldn't close %s port %d "
-			       "for agents\n",
+			pr_err("Couldn't close %s port %d for agents\n",
 			       device->name, cur_port);
 		if (ib_mad_port_close(device, cur_port))
-			printk(KERN_ERR PFX "Couldn't close %s port %d\n",
+			pr_err("Couldn't close %s port %d\n",
 			       device->name, cur_port);
 	}
 }
@@ -3064,7 +3057,7 @@ static int __init ib_mad_init_module(void)
 					 SLAB_HWCACHE_ALIGN,
 					 NULL);
 	if (!ib_mad_cache) {
-		printk(KERN_ERR PFX "Couldn't create ib_mad cache\n");
+		pr_err("Couldn't create ib_mad cache\n");
 		ret = -ENOMEM;
 		goto error1;
 	}
@@ -3072,7 +3065,7 @@ static int __init ib_mad_init_module(void)
 	INIT_LIST_HEAD(&ib_mad_port_list);
 
 	if (ib_register_client(&mad_client)) {
-		printk(KERN_ERR PFX "Couldn't register ib_mad client\n");
+		pr_err("Couldn't register ib_mad client\n");
 		ret = -EINVAL;
 		goto error2;
 	}
-- 
1.8.2

--
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

[PATCH 3/5] ib/mad: Add pr_notice messages for various umad/mad registration failures

[ira.weiny-ral2JQCrhuEAvxtiuMwx3w]

From: Ira Weiny <ira.weiny-ral2JQCrhuEAvxtiuMwx3w@public.gmane.org>

Reviewed-by: Sean Hefty <sean.hefty-ral2JQCrhuEAvxtiuMwx3w@public.gmane.org>
Reviewed-by: Hal Rosenstock <hal-LDSdmyG8hGV8YrgS2mwiifqBs+8SCbDb@public.gmane.org>
Signed-off-by: Ira Weiny <ira.weiny-ral2JQCrhuEAvxtiuMwx3w@public.gmane.org>
---
 drivers/infiniband/core/mad.c      | 47 ++++++++++++++++++++++++++++++--------
 drivers/infiniband/core/user_mad.c |  5 ++++
 2 files changed, 43 insertions(+), 9 deletions(-)

diff --git a/drivers/infiniband/core/mad.c b/drivers/infiniband/core/mad.c
index 7395e23..c48e9a8 100644
--- a/drivers/infiniband/core/mad.c
+++ b/drivers/infiniband/core/mad.c
@@ -214,58 +214,85 @@ struct ib_mad_agent *ib_register_mad_agent(struct ib_device *device,
 
 	/* Validate parameters */
 	qpn = get_spl_qp_index(qp_type);
-	if (qpn == -1)
+	if (qpn == -1) {
+		pr_notice("ib_register_mad_agent: invalid QP Type %d\n",
+			qp_type);
 		goto error1;
+	}
 
-	if (rmpp_version && rmpp_version != IB_MGMT_RMPP_VERSION)
+	if (rmpp_version && rmpp_version != IB_MGMT_RMPP_VERSION) {
+		pr_notice("ib_register_mad_agent: invalid RMPP Version %u\n",
+			rmpp_version);
 		goto error1;
+	}
 
 	/* Validate MAD registration request if supplied */
 	if (mad_reg_req) {
-		if (mad_reg_req->mgmt_class_version >= MAX_MGMT_VERSION)
+		if (mad_reg_req->mgmt_class_version >= MAX_MGMT_VERSION) {
+			pr_notice("ib_register_mad_agent: invalid Class Version %u\n",
+				mad_reg_req->mgmt_class_version);
 			goto error1;
-		if (!recv_handler)
+		}
+		if (!recv_handler) {
+			pr_notice("ib_register_mad_agent: no recv_handler\n");
 			goto error1;
+		}
 		if (mad_reg_req->mgmt_class >= MAX_MGMT_CLASS) {
 			/*
 			 * IB_MGMT_CLASS_SUBN_DIRECTED_ROUTE is the only
 			 * one in this range currently allowed
 			 */
 			if (mad_reg_req->mgmt_class !=
-			    IB_MGMT_CLASS_SUBN_DIRECTED_ROUTE)
+			    IB_MGMT_CLASS_SUBN_DIRECTED_ROUTE) {
+				pr_notice("ib_register_mad_agent: Invalid Mgmt Class 0x%x\n",
+					mad_reg_req->mgmt_class);
 				goto error1;
+			}
 		} else if (mad_reg_req->mgmt_class == 0) {
 			/*
 			 * Class 0 is reserved in IBA and is used for
 			 * aliasing of IB_MGMT_CLASS_SUBN_DIRECTED_ROUTE
 			 */
+			pr_notice("ib_register_mad_agent: Invalid Mgmt Class 0\n");
 			goto error1;
 		} else if (is_vendor_class(mad_reg_req->mgmt_class)) {
 			/*
 			 * If class is in "new" vendor range,
 			 * ensure supplied OUI is not zero
 			 */
-			if (!is_vendor_oui(mad_reg_req->oui))
+			if (!is_vendor_oui(mad_reg_req->oui)) {
+				pr_notice("ib_register_mad_agent: No OUI specified for class 0x%x\n",
+					mad_reg_req->mgmt_class);
 				goto error1;
+			}
 		}
 		/* Make sure class supplied is consistent with RMPP */
 		if (!ib_is_mad_class_rmpp(mad_reg_req->mgmt_class)) {
-			if (rmpp_version)
+			if (rmpp_version) {
+				pr_notice("ib_register_mad_agent: RMPP version for non-RMPP class 0x%x\n",
+					mad_reg_req->mgmt_class);
 				goto error1;
+			}
 		}
 		/* Make sure class supplied is consistent with QP type */
 		if (qp_type == IB_QPT_SMI) {
 			if ((mad_reg_req->mgmt_class !=
 					IB_MGMT_CLASS_SUBN_LID_ROUTED) &&
 			    (mad_reg_req->mgmt_class !=
-					IB_MGMT_CLASS_SUBN_DIRECTED_ROUTE))
+					IB_MGMT_CLASS_SUBN_DIRECTED_ROUTE)) {
+				pr_notice("ib_register_mad_agent: Invalid SM QP type: class 0x%x\n",
+					mad_reg_req->mgmt_class);
 				goto error1;
+			}
 		} else {
 			if ((mad_reg_req->mgmt_class ==
 					IB_MGMT_CLASS_SUBN_LID_ROUTED) ||
 			    (mad_reg_req->mgmt_class ==
-					IB_MGMT_CLASS_SUBN_DIRECTED_ROUTE))
+					IB_MGMT_CLASS_SUBN_DIRECTED_ROUTE)) {
+				pr_notice("ib_register_mad_agent: Invalid GS QP type: class 0x%x\n",
+					mad_reg_req->mgmt_class);
 				goto error1;
+			}
 		}
 	} else {
 		/* No registration request supplied */
@@ -276,6 +303,7 @@ struct ib_mad_agent *ib_register_mad_agent(struct ib_device *device,
 	/* Validate device and port */
 	port_priv = ib_get_mad_port(device, port_num);
 	if (!port_priv) {
+		pr_notice("ib_register_mad_agent: Invalid port\n");
 		ret = ERR_PTR(-ENODEV);
 		goto error1;
 	}
@@ -283,6 +311,7 @@ struct ib_mad_agent *ib_register_mad_agent(struct ib_device *device,
 	/* Verify the QP requested is supported.  For example, Ethernet devices
 	 * will not have QP0 */
 	if (!port_priv->qp_info[qpn].qp) {
+		pr_notice("ib_register_mad_agent: QP %d not supported\n", qpn);
 		ret = ERR_PTR(-EPROTONOSUPPORT);
 		goto error1;
 	}
diff --git a/drivers/infiniband/core/user_mad.c b/drivers/infiniband/core/user_mad.c
index 0df292a..1341282 100644
--- a/drivers/infiniband/core/user_mad.c
+++ b/drivers/infiniband/core/user_mad.c
@@ -612,6 +612,7 @@ static int ib_umad_reg_agent(struct ib_umad_file *file, void __user *arg,
 	mutex_lock(&file->mutex);
 
 	if (!file->port->ib_dev) {
+		pr_notice("ib_umad_reg_agent: invalid device\n");
 		ret = -EPIPE;
 		goto out;
 	}
@@ -622,6 +623,8 @@ static int ib_umad_reg_agent(struct ib_umad_file *file, void __user *arg,
 	}
 
 	if (ureq.qpn != 0 && ureq.qpn != 1) {
+		pr_notice("ib_umad_reg_agent: invalid QPN %d specified\n",
+			ureq.qpn);
 		ret = -EINVAL;
 		goto out;
 	}
@@ -630,6 +633,8 @@ static int ib_umad_reg_agent(struct ib_umad_file *file, void __user *arg,
 		if (!__get_agent(file, agent_id))
 			goto found;
 
+	pr_notice("ib_umad_reg_agent: Max Agents (%u) reached\n",
+		IB_UMAD_MAX_AGENTS);
 	ret = -ENOMEM;
 	goto out;
 
-- 
1.8.2

--
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

[PATCH 4/5] ib/mad: add new ioctl to user space to support new registration options

[ira.weiny-ral2JQCrhuEAvxtiuMwx3w]

From: Ira Weiny <ira.weiny-ral2JQCrhuEAvxtiuMwx3w@public.gmane.org>

Registrations options are specified through flags.  Definitions of flags will
be in subsequent patches.

Reviewed-by: Sean Hefty <sean.hefty-ral2JQCrhuEAvxtiuMwx3w@public.gmane.org>
Reviewed-by: Hal Rosenstock <hal-LDSdmyG8hGV8YrgS2mwiifqBs+8SCbDb@public.gmane.org>
Signed-off-by: Ira Weiny <ira.weiny-ral2JQCrhuEAvxtiuMwx3w@public.gmane.org>
---
 Documentation/infiniband/user_mad.txt   |  13 ++--
 drivers/infiniband/core/agent.c         |   4 +-
 drivers/infiniband/core/cm.c            |   5 +-
 drivers/infiniband/core/mad.c           |   4 +-
 drivers/infiniband/core/sa_query.c      |   2 +-
 drivers/infiniband/core/user_mad.c      | 107 +++++++++++++++++++++++++++++++-
 drivers/infiniband/hw/mlx4/mad.c        |   2 +-
 drivers/infiniband/hw/mthca/mthca_mad.c |   2 +-
 drivers/infiniband/hw/qib/qib_mad.c     |   2 +-
 drivers/infiniband/ulp/srpt/ib_srpt.c   |   2 +-
 include/rdma/ib_mad.h                   |   7 ++-
 include/uapi/rdma/ib_user_mad.h         |  38 ++++++++++++
 12 files changed, 172 insertions(+), 16 deletions(-)

diff --git a/Documentation/infiniband/user_mad.txt b/Documentation/infiniband/user_mad.txt
index 8a36695..bf24958 100644
--- a/Documentation/infiniband/user_mad.txt
+++ b/Documentation/infiniband/user_mad.txt
@@ -26,6 +26,11 @@ Creating MAD agents
   ioctl.  Also, all agents registered through a file descriptor will
   be unregistered when the descriptor is closed.
 
+  2013 -- a new registration ioctl is now provided which allows additional
+       fields to be provided during registration.
+       Users of this registration call are implicitly setting the use of
+       pkey_index (see below).
+
 Receiving MADs
 
   MADs are received using read().  The receive side now supports
@@ -104,10 +109,10 @@ 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.
+  with a pkey_index member has been defined; however, to preserve binary
+  compatibility with older applications, this new layout will not be used
+  unless one of IB_USER_MAD_ENABLE_PKEY or IB_USER_MAD_REGISTER_AGENT2 ioctl's
+  are 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
diff --git a/drivers/infiniband/core/agent.c b/drivers/infiniband/core/agent.c
index e1f08b4..bed8ee4 100644
--- a/drivers/infiniband/core/agent.c
+++ b/drivers/infiniband/core/agent.c
@@ -163,7 +163,7 @@ int ib_agent_port_open(struct ib_device *device, int port_num)
 		port_priv->agent[0] = ib_register_mad_agent(device, port_num,
 							    IB_QPT_SMI, NULL, 0,
 							    &agent_send_handler,
-							    NULL, NULL);
+							    NULL, NULL, 0);
 		if (IS_ERR(port_priv->agent[0])) {
 			ret = PTR_ERR(port_priv->agent[0]);
 			goto error2;
@@ -174,7 +174,7 @@ int ib_agent_port_open(struct ib_device *device, int port_num)
 	port_priv->agent[1] = ib_register_mad_agent(device, port_num,
 						    IB_QPT_GSI, NULL, 0,
 						    &agent_send_handler,
-						    NULL, NULL);
+						    NULL, NULL, 0);
 	if (IS_ERR(port_priv->agent[1])) {
 		ret = PTR_ERR(port_priv->agent[1]);
 		goto error3;
diff --git a/drivers/infiniband/core/cm.c b/drivers/infiniband/core/cm.c
index c323917..e28a494 100644
--- a/drivers/infiniband/core/cm.c
+++ b/drivers/infiniband/core/cm.c
@@ -3753,7 +3753,7 @@ static void cm_add_one(struct ib_device *ib_device)
 	struct cm_port *port;
 	struct ib_mad_reg_req reg_req = {
 		.mgmt_class = IB_MGMT_CLASS_CM,
-		.mgmt_class_version = IB_CM_CLASS_VERSION
+		.mgmt_class_version = IB_CM_CLASS_VERSION,
 	};
 	struct ib_port_modify port_modify = {
 		.set_port_cap_mask = IB_PORT_CM_SUP
@@ -3801,7 +3801,8 @@ static void cm_add_one(struct ib_device *ib_device)
 							0,
 							cm_send_handler,
 							cm_recv_handler,
-							port);
+							port,
+							0);
 		if (IS_ERR(port->mad_agent))
 			goto error2;
 
diff --git a/drivers/infiniband/core/mad.c b/drivers/infiniband/core/mad.c
index c48e9a8..a26baa9 100644
--- a/drivers/infiniband/core/mad.c
+++ b/drivers/infiniband/core/mad.c
@@ -198,7 +198,8 @@ struct ib_mad_agent *ib_register_mad_agent(struct ib_device *device,
 					   u8 rmpp_version,
 					   ib_mad_send_handler send_handler,
 					   ib_mad_recv_handler recv_handler,
-					   void *context)
+					   void *context,
+					   u32 registration_flags)
 {
 	struct ib_mad_port_private *port_priv;
 	struct ib_mad_agent *ret = ERR_PTR(-EINVAL);
@@ -348,6 +349,7 @@ struct ib_mad_agent *ib_register_mad_agent(struct ib_device *device,
 	mad_agent_priv->agent.context = context;
 	mad_agent_priv->agent.qp = port_priv->qp_info[qpn].qp;
 	mad_agent_priv->agent.port_num = port_num;
+	mad_agent_priv->agent.flags = registration_flags;
 	spin_lock_init(&mad_agent_priv->lock);
 	INIT_LIST_HEAD(&mad_agent_priv->send_list);
 	INIT_LIST_HEAD(&mad_agent_priv->wait_list);
diff --git a/drivers/infiniband/core/sa_query.c b/drivers/infiniband/core/sa_query.c
index f820958..d079bc5 100644
--- a/drivers/infiniband/core/sa_query.c
+++ b/drivers/infiniband/core/sa_query.c
@@ -1184,7 +1184,7 @@ static void ib_sa_add_one(struct ib_device *device)
 		sa_dev->port[i].agent =
 			ib_register_mad_agent(device, i + s, IB_QPT_GSI,
 					      NULL, 0, send_handler,
-					      recv_handler, sa_dev);
+					      recv_handler, sa_dev, 0);
 		if (IS_ERR(sa_dev->port[i].agent))
 			goto err;
 
diff --git a/drivers/infiniband/core/user_mad.c b/drivers/infiniband/core/user_mad.c
index 1341282..571e9d8b 100644
--- a/drivers/infiniband/core/user_mad.c
+++ b/drivers/infiniband/core/user_mad.c
@@ -640,6 +640,7 @@ static int ib_umad_reg_agent(struct ib_umad_file *file, void __user *arg,
 
 found:
 	if (ureq.mgmt_class) {
+		memset(&req, 0, sizeof(req));
 		req.mgmt_class         = ureq.mgmt_class;
 		req.mgmt_class_version = ureq.mgmt_class_version;
 		memcpy(req.oui, ureq.oui, sizeof req.oui);
@@ -660,7 +661,7 @@ found:
 				      ureq.qpn ? IB_QPT_GSI : IB_QPT_SMI,
 				      ureq.mgmt_class ? &req : NULL,
 				      ureq.rmpp_version,
-				      send_handler, recv_handler, file);
+				      send_handler, recv_handler, file, 0);
 	if (IS_ERR(agent)) {
 		ret = PTR_ERR(agent);
 		agent = NULL;
@@ -696,6 +697,106 @@ out:
 	return ret;
 }
 
+static int ib_umad_reg_agent2(struct ib_umad_file *file, void __user *arg)
+{
+	struct ib_user_mad_reg_req2 ureq;
+	struct ib_mad_reg_req req;
+	struct ib_mad_agent *agent = NULL;
+	int agent_id;
+	int ret;
+
+	mutex_lock(&file->port->file_mutex);
+	mutex_lock(&file->mutex);
+
+	if (!file->port->ib_dev) {
+		pr_notice("ib_umad_reg_agent2: invalid device\n");
+		ret = -EPIPE;
+		goto out;
+	}
+
+	if (copy_from_user(&ureq, arg, sizeof(ureq))) {
+		ret = -EFAULT;
+		goto out;
+	}
+
+	if (ureq.qpn != 0 && ureq.qpn != 1) {
+		pr_notice("ib_umad_reg_agent2: invalid QPN %d specified\n",
+			ureq.qpn);
+		ret = -EINVAL;
+		goto out;
+	}
+
+	if (ureq.flags & ~IB_USER_MAD_REG_FLAGS_CAP) {
+		pr_notice("ib_umad_reg_agent2 failed: invalid registration flags specified 0x%x; supported 0x%x\n",
+			ureq.flags, IB_USER_MAD_REG_FLAGS_CAP);
+		ret = -EINVAL;
+
+		if (put_user((u32)IB_USER_MAD_REG_FLAGS_CAP,
+				(u32 __user *) (arg + offsetof(struct
+				ib_user_mad_reg_req2, flags))))
+			ret = -EFAULT;
+
+		goto out;
+	}
+
+	for (agent_id = 0; agent_id < IB_UMAD_MAX_AGENTS; ++agent_id)
+		if (!__get_agent(file, agent_id))
+			goto found;
+
+	pr_notice("ib_umad_reg_agent2: Max Agents (%u) reached\n",
+		IB_UMAD_MAX_AGENTS);
+	ret = -ENOMEM;
+	goto out;
+
+found:
+	if (ureq.mgmt_class) {
+		memset(&req, 0, sizeof(req));
+		req.mgmt_class         = ureq.mgmt_class;
+		req.mgmt_class_version = ureq.mgmt_class_version;
+		memcpy(req.oui, ureq.oui, sizeof(req.oui));
+		memcpy(req.method_mask, ureq.method_mask,
+			sizeof(req.method_mask));
+	}
+
+	agent = ib_register_mad_agent(file->port->ib_dev, file->port->port_num,
+				      ureq.qpn ? IB_QPT_GSI : IB_QPT_SMI,
+				      ureq.mgmt_class ? &req : NULL,
+				      ureq.rmpp_version,
+				      send_handler, recv_handler, file,
+				      ureq.flags);
+	if (IS_ERR(agent)) {
+		ret = PTR_ERR(agent);
+		agent = NULL;
+		goto out;
+	}
+
+	if (put_user(agent_id,
+		     (u32 __user *)(arg +
+				offsetof(struct ib_user_mad_reg_req2, id)))) {
+		ret = -EFAULT;
+		goto out;
+	}
+
+	if (!file->already_used) {
+		file->already_used = 1;
+		file->use_pkey_index = 1;
+	}
+
+	file->agent[agent_id] = agent;
+	ret = 0;
+
+out:
+	mutex_unlock(&file->mutex);
+
+	if (ret && agent)
+		ib_unregister_mad_agent(agent);
+
+	mutex_unlock(&file->port->file_mutex);
+
+	return ret;
+}
+
+
 static int ib_umad_unreg_agent(struct ib_umad_file *file, u32 __user *arg)
 {
 	struct ib_mad_agent *agent = NULL;
@@ -751,6 +852,8 @@ static long ib_umad_ioctl(struct file *filp, unsigned int cmd,
 		return ib_umad_unreg_agent(filp->private_data, (__u32 __user *) arg);
 	case IB_USER_MAD_ENABLE_PKEY:
 		return ib_umad_enable_pkey(filp->private_data);
+	case IB_USER_MAD_REGISTER_AGENT2:
+		return ib_umad_reg_agent2(filp->private_data, (void __user *) arg);
 	default:
 		return -ENOIOCTLCMD;
 	}
@@ -767,6 +870,8 @@ static long ib_umad_compat_ioctl(struct file *filp, unsigned int cmd,
 		return ib_umad_unreg_agent(filp->private_data, compat_ptr(arg));
 	case IB_USER_MAD_ENABLE_PKEY:
 		return ib_umad_enable_pkey(filp->private_data);
+	case IB_USER_MAD_REGISTER_AGENT2:
+		return ib_umad_reg_agent2(filp->private_data, compat_ptr(arg));
 	default:
 		return -ENOIOCTLCMD;
 	}
diff --git a/drivers/infiniband/hw/mlx4/mad.c b/drivers/infiniband/hw/mlx4/mad.c
index fd36ec6..f0ef7c84 100644
--- a/drivers/infiniband/hw/mlx4/mad.c
+++ b/drivers/infiniband/hw/mlx4/mad.c
@@ -880,7 +880,7 @@ int mlx4_ib_mad_init(struct mlx4_ib_dev *dev)
 				agent = ib_register_mad_agent(&dev->ib_dev, p + 1,
 							      q ? IB_QPT_GSI : IB_QPT_SMI,
 							      NULL, 0, send_handler,
-							      NULL, NULL);
+							      NULL, NULL, 0);
 				if (IS_ERR(agent)) {
 					ret = PTR_ERR(agent);
 					goto err;
diff --git a/drivers/infiniband/hw/mthca/mthca_mad.c b/drivers/infiniband/hw/mthca/mthca_mad.c
index b6f7f45..8881fa3 100644
--- a/drivers/infiniband/hw/mthca/mthca_mad.c
+++ b/drivers/infiniband/hw/mthca/mthca_mad.c
@@ -294,7 +294,7 @@ int mthca_create_agents(struct mthca_dev *dev)
 			agent = ib_register_mad_agent(&dev->ib_dev, p + 1,
 						      q ? IB_QPT_GSI : IB_QPT_SMI,
 						      NULL, 0, send_handler,
-						      NULL, NULL);
+						      NULL, NULL, 0);
 			if (IS_ERR(agent)) {
 				ret = PTR_ERR(agent);
 				goto err;
diff --git a/drivers/infiniband/hw/qib/qib_mad.c b/drivers/infiniband/hw/qib/qib_mad.c
index edad991..e87a0f6 100644
--- a/drivers/infiniband/hw/qib/qib_mad.c
+++ b/drivers/infiniband/hw/qib/qib_mad.c
@@ -2476,7 +2476,7 @@ int qib_create_agents(struct qib_ibdev *dev)
 		ibp = &dd->pport[p].ibport_data;
 		agent = ib_register_mad_agent(&dev->ibdev, p + 1, IB_QPT_SMI,
 					      NULL, 0, send_handler,
-					      NULL, NULL);
+					      NULL, NULL, 0);
 		if (IS_ERR(agent)) {
 			ret = PTR_ERR(agent);
 			goto err;
diff --git a/drivers/infiniband/ulp/srpt/ib_srpt.c b/drivers/infiniband/ulp/srpt/ib_srpt.c
index fe09f27..8a8311e 100644
--- a/drivers/infiniband/ulp/srpt/ib_srpt.c
+++ b/drivers/infiniband/ulp/srpt/ib_srpt.c
@@ -563,7 +563,7 @@ static int srpt_refresh_port(struct srpt_port *sport)
 							 &reg_req, 0,
 							 srpt_mad_send_handler,
 							 srpt_mad_recv_handler,
-							 sport);
+							 sport, 0);
 		if (IS_ERR(sport->mad_agent)) {
 			ret = PTR_ERR(sport->mad_agent);
 			sport->mad_agent = NULL;
diff --git a/include/rdma/ib_mad.h b/include/rdma/ib_mad.h
index 3d81b90..876f497 100644
--- a/include/rdma/ib_mad.h
+++ b/include/rdma/ib_mad.h
@@ -355,6 +355,7 @@ typedef void (*ib_mad_recv_handler)(struct ib_mad_agent *mad_agent,
  * @hi_tid: Access layer assigned transaction ID for this client.
  *   Unsolicited MADs sent by this client will have the upper 32-bits
  *   of their TID set to this value.
+ * @flags: registration flags
  * @port_num: Port number on which QP is registered
  * @rmpp_version: If set, indicates the RMPP version used by this agent.
  */
@@ -367,6 +368,7 @@ struct ib_mad_agent {
 	ib_mad_snoop_handler	snoop_handler;
 	void			*context;
 	u32			hi_tid;
+	u32			flags;
 	u8			port_num;
 	u8			rmpp_version;
 };
@@ -426,6 +428,7 @@ struct ib_mad_recv_wc {
  *   in the range from 0x30 to 0x4f. Otherwise not used.
  * @method_mask: The caller will receive unsolicited MADs for any method
  *   where @method_mask = 1.
+ *
  */
 struct ib_mad_reg_req {
 	u8	mgmt_class;
@@ -451,6 +454,7 @@ struct ib_mad_reg_req {
  * @recv_handler: The completion callback routine invoked for a received
  *   MAD.
  * @context: User specified context associated with the registration.
+ * @registration_flags: Registration flags to set for this agent
  */
 struct ib_mad_agent *ib_register_mad_agent(struct ib_device *device,
 					   u8 port_num,
@@ -459,7 +463,8 @@ struct ib_mad_agent *ib_register_mad_agent(struct ib_device *device,
 					   u8 rmpp_version,
 					   ib_mad_send_handler send_handler,
 					   ib_mad_recv_handler recv_handler,
-					   void *context);
+					   void *context,
+					   u32 registration_flags);
 
 enum ib_mad_snoop_flags {
 	/*IB_MAD_SNOOP_POSTED_SENDS	   = 1,*/
diff --git a/include/uapi/rdma/ib_user_mad.h b/include/uapi/rdma/ib_user_mad.h
index d6fce1c..51c67ec 100644
--- a/include/uapi/rdma/ib_user_mad.h
+++ b/include/uapi/rdma/ib_user_mad.h
@@ -191,6 +191,41 @@ struct ib_user_mad_reg_req {
 	__u8	rmpp_version;
 };
 
+/**
+ * ib_user_mad_reg_req2 - MAD registration request
+ *
+ * @id                 - Set by the _kernel_; used by userspace to identify the
+ *                       registered agent in future requests.
+ * @qpn                - Queue pair number; must be 0 or 1.
+ * @mgmt_class         - Indicates which management class of MADs should be
+ *                       receive by the caller.  This field is only required if
+ *                       the user wishes to receive unsolicited MADs, otherwise
+ *                       it should be 0.
+ * @mgmt_class_version - Indicates which version of MADs for the given
+ *                       management class to receive.
+ * @res                - Ignored.
+ * @flags              - additional registration flags; Must be in the set of
+ *                       flags defined in IB_USER_MAD_REG_FLAGS_CAP
+ * @method_mask        - The caller wishes to receive unsolicited MADs for the
+ *                       methods whose bit(s) is(are) set.
+ * @oui                - Indicates IEEE OUI to use when mgmt_class is a vendor
+ *                       class in the range from 0x30 to 0x4f. Otherwise not
+ *                       used.
+ * @rmpp_version       - If set, indicates the RMPP version to use.
+ */
+#define IB_USER_MAD_REG_FLAGS_CAP (0)
+struct ib_user_mad_reg_req2 {
+	__u32	id;
+	__u32	qpn;
+	__u8	mgmt_class;
+	__u8	mgmt_class_version;
+	__u16   res;
+	__u32   flags;
+	__u64   method_mask[2];
+	__u8    oui[3]; /* network order */
+	__u8	rmpp_version;
+};
+
 #define IB_IOCTL_MAGIC		0x1b
 
 #define IB_USER_MAD_REGISTER_AGENT	_IOWR(IB_IOCTL_MAGIC, 1, \
@@ -200,4 +235,7 @@ struct ib_user_mad_reg_req {
 
 #define IB_USER_MAD_ENABLE_PKEY		_IO(IB_IOCTL_MAGIC, 3)
 
+#define IB_USER_MAD_REGISTER_AGENT2     _IOWR(IB_IOCTL_MAGIC, 4, \
+					      struct ib_user_mad_reg_req2)
+
 #endif /* IB_USER_MAD_H */
-- 
1.8.2

--
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

[PATCH 5/5] ib/mad: Add user space RMPP support

[ira.weiny-ral2JQCrhuEAvxtiuMwx3w]

From: Ira Weiny <ira.weiny-ral2JQCrhuEAvxtiuMwx3w@public.gmane.org>

Using the new registration mechanism define a flag which indicates the user
wishes to process RMPP messages in user space rather than have the kernel
process them.

Reviewed-by: Sean Hefty <sean.hefty-ral2JQCrhuEAvxtiuMwx3w@public.gmane.org>
Reviewed-by: Hal Rosenstock <hal-LDSdmyG8hGV8YrgS2mwiifqBs+8SCbDb@public.gmane.org>
Signed-off-by: Ira Weiny <ira.weiny-ral2JQCrhuEAvxtiuMwx3w@public.gmane.org>
---
 drivers/infiniband/core/mad.c      | 74 ++++++++++++++++++++++++++------------
 drivers/infiniband/core/user_mad.c | 34 +++++++++++-------
 include/rdma/ib_mad.h              | 11 ++++++
 include/uapi/rdma/ib_user_mad.h    |  5 ++-
 4 files changed, 88 insertions(+), 36 deletions(-)

diff --git a/drivers/infiniband/core/mad.c b/drivers/infiniband/core/mad.c
index a26baa9..b7b19a5 100644
--- a/drivers/infiniband/core/mad.c
+++ b/drivers/infiniband/core/mad.c
@@ -275,6 +275,7 @@ struct ib_mad_agent *ib_register_mad_agent(struct ib_device *device,
 				goto error1;
 			}
 		}
+
 		/* Make sure class supplied is consistent with QP type */
 		if (qp_type == IB_QPT_SMI) {
 			if ((mad_reg_req->mgmt_class !=
@@ -299,6 +300,8 @@ struct ib_mad_agent *ib_register_mad_agent(struct ib_device *device,
 		/* No registration request supplied */
 		if (!send_handler)
 			goto error1;
+		if (registration_flags & IB_MAD_USER_RMPP)
+			goto error1;
 	}
 
 	/* Validate device and port */
@@ -895,6 +898,12 @@ static int alloc_send_rmpp_list(struct ib_mad_send_wr_private *send_wr,
 	return 0;
 }
 
+int ib_mad_kernel_rmpp_agent(struct ib_mad_agent *agent)
+{
+	return agent->rmpp_version && !(agent->flags & IB_MAD_USER_RMPP);
+}
+EXPORT_SYMBOL(ib_mad_kernel_rmpp_agent);
+
 struct ib_mad_send_buf * ib_create_send_mad(struct ib_mad_agent *mad_agent,
 					    u32 remote_qpn, u16 pkey_index,
 					    int rmpp_active,
@@ -911,10 +920,12 @@ struct ib_mad_send_buf * ib_create_send_mad(struct ib_mad_agent *mad_agent,
 	pad = get_pad_size(hdr_len, data_len);
 	message_size = hdr_len + data_len + pad;
 
-	if ((!mad_agent->rmpp_version &&
-	     (rmpp_active || message_size > sizeof(struct ib_mad))) ||
-	    (!rmpp_active && message_size > sizeof(struct ib_mad)))
-		return ERR_PTR(-EINVAL);
+	if (ib_mad_kernel_rmpp_agent(mad_agent)) {
+		if (!rmpp_active && message_size > sizeof(struct ib_mad))
+			return ERR_PTR(-EINVAL);
+	} else
+		if (rmpp_active || message_size > sizeof(struct ib_mad))
+			return ERR_PTR(-EINVAL);
 
 	size = rmpp_active ? hdr_len : sizeof(struct ib_mad);
 	buf = kzalloc(sizeof *mad_send_wr + size, gfp_mask);
@@ -1168,7 +1179,7 @@ int ib_post_send_mad(struct ib_mad_send_buf *send_buf,
 			      &mad_agent_priv->send_list);
 		spin_unlock_irqrestore(&mad_agent_priv->lock, flags);
 
-		if (mad_agent_priv->agent.rmpp_version) {
+		if (ib_mad_kernel_rmpp_agent(&mad_agent_priv->agent)) {
 			ret = ib_send_rmpp_mad(mad_send_wr);
 			if (ret >= 0 && ret != IB_RMPP_RESULT_CONSUMED)
 				ret = ib_send_mad(mad_send_wr);
@@ -1713,6 +1724,7 @@ static int is_data_mad(struct ib_mad_agent_private *mad_agent_priv,
 
 	rmpp_mad = (struct ib_rmpp_mad *)mad_hdr;
 	return !mad_agent_priv->agent.rmpp_version ||
+		!ib_mad_kernel_rmpp_agent(&mad_agent_priv->agent) ||
 		!(ib_get_rmpp_flags(&rmpp_mad->rmpp_hdr) &
 				    IB_MGMT_RMPP_FLAG_ACTIVE) ||
 		(rmpp_mad->rmpp_hdr.rmpp_type == IB_MGMT_RMPP_TYPE_DATA);
@@ -1840,7 +1852,7 @@ static void ib_mad_complete_recv(struct ib_mad_agent_private *mad_agent_priv,
 
 	INIT_LIST_HEAD(&mad_recv_wc->rmpp_list);
 	list_add(&mad_recv_wc->recv_buf.list, &mad_recv_wc->rmpp_list);
-	if (mad_agent_priv->agent.rmpp_version) {
+	if (ib_mad_kernel_rmpp_agent(&mad_agent_priv->agent)) {
 		mad_recv_wc = ib_process_rmpp_recv_wc(mad_agent_priv,
 						      mad_recv_wc);
 		if (!mad_recv_wc) {
@@ -1855,23 +1867,39 @@ static void ib_mad_complete_recv(struct ib_mad_agent_private *mad_agent_priv,
 		mad_send_wr = ib_find_send_mad(mad_agent_priv, mad_recv_wc);
 		if (!mad_send_wr) {
 			spin_unlock_irqrestore(&mad_agent_priv->lock, flags);
-			ib_free_recv_mad(mad_recv_wc);
-			deref_mad_agent(mad_agent_priv);
-			return;
-		}
-		ib_mark_mad_done(mad_send_wr);
-		spin_unlock_irqrestore(&mad_agent_priv->lock, flags);
+			if (!ib_mad_kernel_rmpp_agent(&mad_agent_priv->agent)
+			   && ib_is_mad_class_rmpp(mad_recv_wc->recv_buf.mad->mad_hdr.mgmt_class)
+			   && (ib_get_rmpp_flags(&((struct ib_rmpp_mad *)mad_recv_wc->recv_buf.mad)->rmpp_hdr)
+					& IB_MGMT_RMPP_FLAG_ACTIVE)) {
+				/* user rmpp is in effect
+				 * and this is an active RMPP MAD
+				 */
+				mad_recv_wc->wc->wr_id = 0;
+				mad_agent_priv->agent.recv_handler(&mad_agent_priv->agent,
+								   mad_recv_wc);
+				atomic_dec(&mad_agent_priv->refcount);
+			} else {
+				/* not user rmpp, revert to normal behavior and
+				 * drop the mad */
+				ib_free_recv_mad(mad_recv_wc);
+				deref_mad_agent(mad_agent_priv);
+				return;
+			}
+		} else {
+			ib_mark_mad_done(mad_send_wr);
+			spin_unlock_irqrestore(&mad_agent_priv->lock, flags);
 
-		/* Defined behavior is to complete response before request */
-		mad_recv_wc->wc->wr_id = (unsigned long) &mad_send_wr->send_buf;
-		mad_agent_priv->agent.recv_handler(&mad_agent_priv->agent,
-						   mad_recv_wc);
-		atomic_dec(&mad_agent_priv->refcount);
+			/* Defined behavior is to complete response before request */
+			mad_recv_wc->wc->wr_id = (unsigned long) &mad_send_wr->send_buf;
+			mad_agent_priv->agent.recv_handler(&mad_agent_priv->agent,
+							   mad_recv_wc);
+			atomic_dec(&mad_agent_priv->refcount);
 
-		mad_send_wc.status = IB_WC_SUCCESS;
-		mad_send_wc.vendor_err = 0;
-		mad_send_wc.send_buf = &mad_send_wr->send_buf;
-		ib_mad_complete_send_wr(mad_send_wr, &mad_send_wc);
+			mad_send_wc.status = IB_WC_SUCCESS;
+			mad_send_wc.vendor_err = 0;
+			mad_send_wc.send_buf = &mad_send_wr->send_buf;
+			ib_mad_complete_send_wr(mad_send_wr, &mad_send_wc);
+		}
 	} else {
 		mad_agent_priv->agent.recv_handler(&mad_agent_priv->agent,
 						   mad_recv_wc);
@@ -2110,7 +2138,7 @@ void ib_mad_complete_send_wr(struct ib_mad_send_wr_private *mad_send_wr,
 
 	mad_agent_priv = mad_send_wr->mad_agent_priv;
 	spin_lock_irqsave(&mad_agent_priv->lock, flags);
-	if (mad_agent_priv->agent.rmpp_version) {
+	if (ib_mad_kernel_rmpp_agent(&mad_agent_priv->agent)) {
 		ret = ib_process_rmpp_send_wc(mad_send_wr, mad_send_wc);
 		if (ret == IB_RMPP_RESULT_CONSUMED)
 			goto done;
@@ -2503,7 +2531,7 @@ static int retry_send(struct ib_mad_send_wr_private *mad_send_wr)
 
 	mad_send_wr->timeout = msecs_to_jiffies(mad_send_wr->send_buf.timeout_ms);
 
-	if (mad_send_wr->mad_agent_priv->agent.rmpp_version) {
+	if (ib_mad_kernel_rmpp_agent(&mad_send_wr->mad_agent_priv->agent)) {
 		ret = ib_retry_rmpp(mad_send_wr);
 		switch (ret) {
 		case IB_RMPP_RESULT_UNHANDLED:
diff --git a/drivers/infiniband/core/user_mad.c b/drivers/infiniband/core/user_mad.c
index 571e9d8b..d4261d1 100644
--- a/drivers/infiniband/core/user_mad.c
+++ b/drivers/infiniband/core/user_mad.c
@@ -502,13 +502,15 @@ static ssize_t ib_umad_write(struct file *filp, const char __user *buf,
 
 	rmpp_mad = (struct ib_rmpp_mad *) packet->mad.data;
 	hdr_len = ib_get_mad_data_offset(rmpp_mad->mad_hdr.mgmt_class);
-	if (!ib_is_mad_class_rmpp(rmpp_mad->mad_hdr.mgmt_class)) {
-		copy_offset = IB_MGMT_MAD_HDR;
-		rmpp_active = 0;
-	} else {
+
+	if (ib_is_mad_class_rmpp(rmpp_mad->mad_hdr.mgmt_class)
+	    && ib_mad_kernel_rmpp_agent(agent)) {
 		copy_offset = IB_MGMT_RMPP_HDR;
 		rmpp_active = ib_get_rmpp_flags(&rmpp_mad->rmpp_hdr) &
-			      IB_MGMT_RMPP_FLAG_ACTIVE;
+						IB_MGMT_RMPP_FLAG_ACTIVE;
+	} else {
+		copy_offset = IB_MGMT_MAD_HDR;
+		rmpp_active = 0;
 	}
 
 	data_len = count - hdr_size(file) - hdr_len;
@@ -554,14 +556,22 @@ static ssize_t ib_umad_write(struct file *filp, const char __user *buf,
 		rmpp_mad->mad_hdr.tid = *tid;
 	}
 
-	spin_lock_irq(&file->send_lock);
-	ret = is_duplicate(file, packet);
-	if (!ret)
+	if (!ib_mad_kernel_rmpp_agent(agent)
+	   && ib_is_mad_class_rmpp(rmpp_mad->mad_hdr.mgmt_class)
+	   && (ib_get_rmpp_flags(&rmpp_mad->rmpp_hdr) & IB_MGMT_RMPP_FLAG_ACTIVE)) {
+		spin_lock_irq(&file->send_lock);
 		list_add_tail(&packet->list, &file->send_list);
-	spin_unlock_irq(&file->send_lock);
-	if (ret) {
-		ret = -EINVAL;
-		goto err_msg;
+		spin_unlock_irq(&file->send_lock);
+	} else {
+		spin_lock_irq(&file->send_lock);
+		ret = is_duplicate(file, packet);
+		if (!ret)
+			list_add_tail(&packet->list, &file->send_list);
+		spin_unlock_irq(&file->send_lock);
+		if (ret) {
+			ret = -EINVAL;
+			goto err_msg;
+		}
 	}
 
 	ret = ib_post_send_mad(packet->msg, NULL);
diff --git a/include/rdma/ib_mad.h b/include/rdma/ib_mad.h
index 876f497..9bb99e9 100644
--- a/include/rdma/ib_mad.h
+++ b/include/rdma/ib_mad.h
@@ -40,6 +40,7 @@
 #include <linux/list.h>
 
 #include <rdma/ib_verbs.h>
+#include <uapi/rdma/ib_user_mad.h>
 
 /* Management base version */
 #define IB_MGMT_BASE_VERSION			1
@@ -359,6 +360,9 @@ typedef void (*ib_mad_recv_handler)(struct ib_mad_agent *mad_agent,
  * @port_num: Port number on which QP is registered
  * @rmpp_version: If set, indicates the RMPP version used by this agent.
  */
+enum {
+	IB_MAD_USER_RMPP = IB_USER_MAD_USER_RMPP,
+};
 struct ib_mad_agent {
 	struct ib_device	*device;
 	struct ib_qp		*qp;
@@ -666,4 +670,11 @@ void *ib_get_rmpp_segment(struct ib_mad_send_buf *send_buf, int seg_num);
  */
 void ib_free_send_mad(struct ib_mad_send_buf *send_buf);
 
+/**
+ * ib_mad_kernel_rmpp_agent - Returns if the agent is performing RMPP.
+ * @agent: the agent in question
+ * @return: true if agent is performing rmpp, false otherwise.
+ */
+int ib_mad_kernel_rmpp_agent(struct ib_mad_agent *agent);
+
 #endif /* IB_MAD_H */
diff --git a/include/uapi/rdma/ib_user_mad.h b/include/uapi/rdma/ib_user_mad.h
index 51c67ec..d3bc09e 100644
--- a/include/uapi/rdma/ib_user_mad.h
+++ b/include/uapi/rdma/ib_user_mad.h
@@ -213,7 +213,10 @@ struct ib_user_mad_reg_req {
  *                       used.
  * @rmpp_version       - If set, indicates the RMPP version to use.
  */
-#define IB_USER_MAD_REG_FLAGS_CAP (0)
+enum {
+	IB_USER_MAD_USER_RMPP = (1 << 0),
+};
+#define IB_USER_MAD_REG_FLAGS_CAP (IB_USER_MAD_USER_RMPP)
 struct ib_user_mad_reg_req2 {
 	__u32	id;
 	__u32	qpn;
-- 
1.8.2

--
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

Re: [PATCH 2/5] ib/mad: update module to pr_* style print messages

[Jason Gunthorpe]

On Mon, May 05, 2014 at 09:44:47PM -0400, ira.weiny-ral2JQCrhuEAvxtiuMwx3w@public.gmane.org wrote:
> From: Ira Weiny <ira.weiny-ral2JQCrhuEAvxtiuMwx3w@public.gmane.org>
> 
> Also combine previously line broken user-visible strings as per
> Documentation/CodingStyle

These seem like nice cleanups, however it would be much nicer to use
the dev_* versions when a struct device is available, I think many of
these cases can reach the /dev/umadX struct device..

Jason
--
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

Re: [PATCH 3/5] ib/mad: Add pr_notice messages for various umad/mad registration failures

[Jason Gunthorpe]

On Mon, May 05, 2014 at 09:44:48PM -0400, ira.weiny-ral2JQCrhuEAvxtiuMwx3w@public.gmane.org wrote:
> From: Ira Weiny <ira.weiny-ral2JQCrhuEAvxtiuMwx3w@public.gmane.org>

An empty commit message is not good enough for the mainline kernel,
please describe motivation at least.

I am worried about log DOS potential here, it looks like many of these
messages could be triggered by userspace? Have you thought about this?

Generally moving this stuff to a pr_debug might be more appropriate,
it is very unusual to have logging like this for malformed syscall
arguments.

Jason
--
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

Re: [PATCH 4/5] ib/mad: add new ioctl to user space to support new registration options

[Jason Gunthorpe]

On Mon, May 05, 2014 at 09:44:49PM -0400, ira.weiny-ral2JQCrhuEAvxtiuMwx3w@public.gmane.org wrote:

> +struct ib_user_mad_reg_req2 {
> +	__u32	id;
> +	__u32	qpn;
> +	__u8	mgmt_class;
> +	__u8	mgmt_class_version;
> +	__u16   res;
> +	__u32   flags;
> +	__u64   method_mask[2];
> +	__u8    oui[3]; /* network order */
> +	__u8	rmpp_version;
> +};

Please make the padding on 64 bit explicit so we don't have weird
problems mixing 32 and 64 bit code.

Similar comment as for userspace using a host order uint for the oui.

It would also be good to verify that a 32 bit binary can successfully
execute the ioctl on a 64 bit kernel.

Jason
--
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

RE: [PATCH 2/5] ib/mad: update module to pr_* style print messages

[Weiny, Ira]

> 
> On Mon, May 05, 2014 at 09:44:47PM -0400, ira.weiny-ral2JQCrhuEAvxtiuMwx3w@public.gmane.org wrote:
> > From: Ira Weiny <ira.weiny-ral2JQCrhuEAvxtiuMwx3w@public.gmane.org>
> >
> > Also combine previously line broken user-visible strings as per
> > Documentation/CodingStyle
> 
> These seem like nice cleanups, however it would be much nicer to use the
> dev_* versions when a struct device is available, I think many of these cases
> can reach the /dev/umadX struct device..

Ok, this works nice for ib_umad:

infiniband_mad umad1: ib_umad_reg_agent2 failed: invalid registration flags specified 0x2; supported 0x1
vs.
user_mad: ib_umad_reg_agent2 failed: invalid registration flags specified 0x2; supported 0x1


However in ib_mad I'm not so sure.  The error now implies it is the driver which is flagging the error, _not_ ib_mad.

For example here is what you get when you change pr_notice to dev_notice in ib_mad.

ib_qib 0000:01:00.0: ib_register_mad_agent: invalid RMPP Version 2

OR (There are 2 "struct device" in the struct ib_device passed to the register function: dma_device vs dev)

infiniband qib0: ib_register_mad_agent: invalid RMPP Version 2

vs.

ib_mad: ib_register_mad_agent: invalid RMPP Version 2

Either message implies it is the HCA driver which is flagging the error.  Which I think is misleading but maybe this is more correct and is the "kernel way"?

Ira

> 
> Jason
--
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

RE: [PATCH 3/5] ib/mad: Add pr_notice messages for various umad/mad registration failures

[Weiny, Ira]

> 
> On Mon, May 05, 2014 at 09:44:48PM -0400, ira.weiny-ral2JQCrhuEAvxtiuMwx3w@public.gmane.org wrote:
> > From: Ira Weiny <ira.weiny-ral2JQCrhuEAvxtiuMwx3w@public.gmane.org>
> 
> An empty commit message is not good enough for the mainline kernel,
> please describe motivation at least.

Will do.

> 
> I am worried about log DOS potential here, it looks like many of these
> messages could be triggered by userspace? Have you thought about this?

Yes I did think about it, see below.

> 
> Generally moving this stuff to a pr_debug might be more appropriate, it is
> very unusual to have logging like this for malformed syscall arguments.

I debated using debug.  The problem with debug is that it requires a kernel recompile.  This may not be acceptable for some users.  Ib_umad devices should  only be accessible from a privileged user which I felt was sufficient protection against a DOS attack.  But if it would be more acceptable I can change to [pr|dev]_debug.

Ira

> 
> Jason
--
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

RE: [PATCH 4/5] ib/mad: add new ioctl to user space to support new registration options

[Weiny, Ira]

> 
> On Mon, May 05, 2014 at 09:44:49PM -0400, ira.weiny-ral2JQCrhuEAvxtiuMwx3w@public.gmane.org wrote:
> 
> > +struct ib_user_mad_reg_req2 {
> > +	__u32	id;
> > +	__u32	qpn;
> > +	__u8	mgmt_class;
> > +	__u8	mgmt_class_version;
> > +	__u16   res;
> > +	__u32   flags;
> > +	__u64   method_mask[2];
> > +	__u8    oui[3]; /* network order */
> > +	__u8	rmpp_version;
> > +};
> 
> Please make the padding on 64 bit explicit so we don't have weird problems
> mixing 32 and 64 bit code.

Just to be clear you mean to add a __u32 reserved field to the end of this struct?

> 
> Similar comment as for userspace using a host order uint for the oui.

Ok would you prefer changing this to __u32 and specifying in host order?  Then adjust the padding, like this:

__u32 oui;
_u8 rmpp_version;
_u8 res[3];

To complete the changes?

> 
> It would also be good to verify that a 32 bit binary can successfully execute
> the ioctl on a 64 bit kernel.

I'll try to do this once we agree to the above changes.

Thanks,
Ira

> 
> Jason
--
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

Re: [PATCH 4/5] ib/mad: add new ioctl to user space to support new registration options

[Jason Gunthorpe]

On Wed, May 07, 2014 at 02:39:26PM +0000, Weiny, Ira wrote:
> > On Mon, May 05, 2014 at 09:44:49PM -0400, ira.weiny-ral2JQCrhuEAvxtiuMwx3w@public.gmane.org wrote:
> > 
> > > +struct ib_user_mad_reg_req2 {
> > > +	__u32	id;
> > > +	__u32	qpn;
> > > +	__u8	mgmt_class;
> > > +	__u8	mgmt_class_version;
> > > +	__u16   res;
> > > +	__u32   flags;
> > > +	__u64   method_mask[2];
> > > +	__u8    oui[3]; /* network order */
> > > +	__u8	rmpp_version;
> > > +};
> > 
> > Please make the padding on 64 bit explicit so we don't have weird problems
> > mixing 32 and 64 bit code.
> 
> Just to be clear you mean to add a __u32 reserved field to the end of this struct?

I think that is enough, yes. Yann has been using pahole to investigate
these issues.


> > Similar comment as for userspace using a host order uint for the oui.
> 
> Ok would you prefer changing this to __u32 and specifying in host order?  Then adjust the padding, like this:
> 
> __u32 oui;
> _u8 rmpp_version;
> _u8 res[3];
> 
> To complete the changes?

s/res/reserved/

Sure, that seems simpler for all involved. 

Jason
--
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

Re: [PATCH 2/5] ib/mad: update module to pr_* style print messages

[Jason Gunthorpe]

On Wed, May 07, 2014 at 02:30:04PM +0000, Weiny, Ira wrote:
> > 
> > On Mon, May 05, 2014 at 09:44:47PM -0400, ira.weiny-ral2JQCrhuEAvxtiuMwx3w@public.gmane.org wrote:
> > > From: Ira Weiny <ira.weiny-ral2JQCrhuEAvxtiuMwx3w@public.gmane.org>
> > >
> > > Also combine previously line broken user-visible strings as per
> > > Documentation/CodingStyle
> > 
> > These seem like nice cleanups, however it would be much nicer to use the
> > dev_* versions when a struct device is available, I think many of these cases
> > can reach the /dev/umadX struct device..
> 
> Ok, this works nice for ib_umad:
> 
> infiniband_mad umad1: ib_umad_reg_agent2 failed: invalid registration flags specified 0x2; supported 0x1
> vs.
> user_mad: ib_umad_reg_agent2 failed: invalid registration flags specified 0x2; supported 0x1

Right..

> However in ib_mad I'm not so sure.  The error now implies it is the
> driver which is flagging the error, _not_ ib_mad.

Does it matter? ib_mad is a kernel implementation detail, and a kernel
consumer could trigger the debugging print as well. The function name
is a big enough clue that the problem is ib_mad related, IMHO.

Jason
--
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