[PATCH for-next 0/7] Add support for multicast loopback prevention to mlx4

[PATCH for-next 0/7] Add support for multicast loopback prevention to mlx4

[Eran Ben Elisha]

Hi Doug,

This patch-set adds a new  implementation for multicast loopback prevention for
mlx4 driver.  The current implementation is very limited, especially if link
layer is Ethernet. The new implementation is based on HW feature of dropping
incoming multicast packets if the sender QP counter index is equal to the
receiver counter index.

Patch 0001 extends ib_uverbs_create_qp in order to allow receiving the
multicast loopback flag at create flags.
Patch 0002 adds an infrastructure for the counters' loopback prevention in the
mlx4_core.
Patch 0003 modifies mlx4_en QPs to use the new loopback prevention mode.
Patches 0004-0006 implements this feature for mlx4_ib driver.
Patch 0007 allows setting IB_QP_CREATE_BLOCK_MULTICAST_LOOPBACK in create_flag
field both from uverbs and ib_create_qp.

Thanks,
Eran.

Eran Ben Elisha (5):
  IB/core: Extend ib_uverbs_create_qp
  IB/core: Allow setting create flags in QP init attribute
  IB/mlx4: Add IB counters table
  IB/mlx4: Add counter based implementation for QP multicast loopback
    block
  IB/mlx4: Add support for blocking multicast loopback QP creation user
    flag

Maor Gottlieb (2):
  net/mlx4_core: Add support for filtering multicast loopback
  net/mlx4_en: Implement mcast loopback prevention for ETH qps

 drivers/infiniband/core/uverbs.h                   |   1 +
 drivers/infiniband/core/uverbs_cmd.c               | 256 +++++++++++++++------
 drivers/infiniband/core/uverbs_main.c              |   1 +
 drivers/infiniband/hw/mlx4/mad.c                   |  25 +-
 drivers/infiniband/hw/mlx4/main.c                  |  66 ++++--
 drivers/infiniband/hw/mlx4/mlx4_ib.h               |  10 +-
 drivers/infiniband/hw/mlx4/qp.c                    |  88 ++++++-
 drivers/net/ethernet/mellanox/mlx4/en_main.c       |  22 ++
 drivers/net/ethernet/mellanox/mlx4/en_resources.c  |  25 ++
 drivers/net/ethernet/mellanox/mlx4/fw.c            |   6 +
 drivers/net/ethernet/mellanox/mlx4/mlx4_en.h       |   3 +-
 drivers/net/ethernet/mellanox/mlx4/qp.c            |  19 +-
 .../net/ethernet/mellanox/mlx4/resource_tracker.c  |  30 ++-
 include/linux/mlx4/device.h                        |   2 +
 include/linux/mlx4/qp.h                            |  24 +-
 include/uapi/rdma/ib_user_verbs.h                  |  26 +++
 16 files changed, 496 insertions(+), 108 deletions(-)

-- 
1.8.3.1

--
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 for-next 1/7] IB/core: Extend ib_uverbs_create_qp

[Eran Ben Elisha]

ib_uverbs_ex_create_qp follows the extension verbs
mechanism. New features (for example, QP creation flags
field which is added in a downstream patch) could used
via user-space libraries without breaking the ABI.

Signed-off-by: Eran Ben Elisha <eranbe-VPRAkNaXOzVWk0Htik3J/w@public.gmane.org>
---
 drivers/infiniband/core/uverbs.h      |   1 +
 drivers/infiniband/core/uverbs_cmd.c  | 256 +++++++++++++++++++++++++---------
 drivers/infiniband/core/uverbs_main.c |   1 +
 include/uapi/rdma/ib_user_verbs.h     |  26 ++++
 4 files changed, 220 insertions(+), 64 deletions(-)

diff --git a/drivers/infiniband/core/uverbs.h b/drivers/infiniband/core/uverbs.h
index ba365b6..627cec1 100644
--- a/drivers/infiniband/core/uverbs.h
+++ b/drivers/infiniband/core/uverbs.h
@@ -260,5 +260,6 @@ IB_UVERBS_DECLARE_EX_CMD(create_flow);
 IB_UVERBS_DECLARE_EX_CMD(destroy_flow);
 IB_UVERBS_DECLARE_EX_CMD(query_device);
 IB_UVERBS_DECLARE_EX_CMD(create_cq);
+IB_UVERBS_DECLARE_EX_CMD(create_qp);
 
 #endif /* UVERBS_H */
diff --git a/drivers/infiniband/core/uverbs_cmd.c b/drivers/infiniband/core/uverbs_cmd.c
index bbb02ff..2bac2f1 100644
--- a/drivers/infiniband/core/uverbs_cmd.c
+++ b/drivers/infiniband/core/uverbs_cmd.c
@@ -1721,65 +1721,65 @@ ssize_t ib_uverbs_destroy_cq(struct ib_uverbs_file *file,
 	return in_len;
 }
 
-ssize_t ib_uverbs_create_qp(struct ib_uverbs_file *file,
-			    const char __user *buf, int in_len,
-			    int out_len)
-{
-	struct ib_uverbs_create_qp      cmd;
-	struct ib_uverbs_create_qp_resp resp;
-	struct ib_udata                 udata;
-	struct ib_uqp_object           *obj;
-	struct ib_device	       *device;
-	struct ib_pd                   *pd = NULL;
-	struct ib_xrcd		       *xrcd = NULL;
-	struct ib_uobject	       *uninitialized_var(xrcd_uobj);
-	struct ib_cq                   *scq = NULL, *rcq = NULL;
-	struct ib_srq                  *srq = NULL;
-	struct ib_qp                   *qp;
-	struct ib_qp_init_attr          attr;
-	int ret;
-
-	if (out_len < sizeof resp)
-		return -ENOSPC;
-
-	if (copy_from_user(&cmd, buf, sizeof cmd))
-		return -EFAULT;
+static int create_qp(struct ib_uverbs_file *file,
+		     struct ib_udata *ucore,
+		     struct ib_udata *uhw,
+		     struct ib_uverbs_ex_create_qp *cmd,
+		     size_t cmd_sz,
+		     int (*cb)(struct ib_uverbs_file *file,
+			       struct ib_uverbs_ex_create_qp_resp *resp,
+			       struct ib_udata *udata),
+		     void *context)
+{
+	struct ib_uqp_object		*obj;
+	struct ib_device		*device;
+	struct ib_pd			*pd = NULL;
+	struct ib_xrcd			*xrcd = NULL;
+	struct ib_uobject		*uninitialized_var(xrcd_uobj);
+	struct ib_cq			*scq = NULL, *rcq = NULL;
+	struct ib_srq			*srq = NULL;
+	struct ib_qp			*qp;
+	char				*buf;
+	struct ib_qp_init_attr		attr;
+	struct ib_uverbs_ex_create_qp_resp resp;
+	int				ret;
 
-	if (cmd.qp_type == IB_QPT_RAW_PACKET && !capable(CAP_NET_RAW))
+	if (cmd->qp_type == IB_QPT_RAW_PACKET && !capable(CAP_NET_RAW))
 		return -EPERM;
 
-	INIT_UDATA(&udata, buf + sizeof cmd,
-		   (unsigned long) cmd.response + sizeof resp,
-		   in_len - sizeof cmd, out_len - sizeof resp);
-
 	obj = kzalloc(sizeof *obj, GFP_KERNEL);
 	if (!obj)
 		return -ENOMEM;
 
-	init_uobj(&obj->uevent.uobject, cmd.user_handle, file->ucontext, &qp_lock_class);
+	init_uobj(&obj->uevent.uobject, cmd->user_handle, file->ucontext,
+		  &qp_lock_class);
 	down_write(&obj->uevent.uobject.mutex);
 
-	if (cmd.qp_type == IB_QPT_XRC_TGT) {
-		xrcd = idr_read_xrcd(cmd.pd_handle, file->ucontext, &xrcd_uobj);
+	if (cmd->qp_type == IB_QPT_XRC_TGT) {
+		xrcd = idr_read_xrcd(cmd->pd_handle, file->ucontext,
+				     &xrcd_uobj);
 		if (!xrcd) {
 			ret = -EINVAL;
 			goto err_put;
 		}
 		device = xrcd->device;
 	} else {
-		if (cmd.qp_type == IB_QPT_XRC_INI) {
-			cmd.max_recv_wr = cmd.max_recv_sge = 0;
+		if (cmd->qp_type == IB_QPT_XRC_INI) {
+			cmd->max_recv_wr = 0;
+			cmd->max_recv_sge = 0;
 		} else {
-			if (cmd.is_srq) {
-				srq = idr_read_srq(cmd.srq_handle, file->ucontext);
+			if (cmd->is_srq) {
+				srq = idr_read_srq(cmd->srq_handle,
+						   file->ucontext);
 				if (!srq || srq->srq_type != IB_SRQT_BASIC) {
 					ret = -EINVAL;
 					goto err_put;
 				}
 			}
 
-			if (cmd.recv_cq_handle != cmd.send_cq_handle) {
-				rcq = idr_read_cq(cmd.recv_cq_handle, file->ucontext, 0);
+			if (cmd->recv_cq_handle != cmd->send_cq_handle) {
+				rcq = idr_read_cq(cmd->recv_cq_handle,
+						  file->ucontext, 0);
 				if (!rcq) {
 					ret = -EINVAL;
 					goto err_put;
@@ -1787,9 +1787,9 @@ ssize_t ib_uverbs_create_qp(struct ib_uverbs_file *file,
 			}
 		}
 
-		scq = idr_read_cq(cmd.send_cq_handle, file->ucontext, !!rcq);
+		scq = idr_read_cq(cmd->send_cq_handle, file->ucontext, !!rcq);
 		rcq = rcq ?: scq;
-		pd  = idr_read_pd(cmd.pd_handle, file->ucontext);
+		pd  = idr_read_pd(cmd->pd_handle, file->ucontext);
 		if (!pd || !scq) {
 			ret = -EINVAL;
 			goto err_put;
@@ -1804,31 +1804,54 @@ ssize_t ib_uverbs_create_qp(struct ib_uverbs_file *file,
 	attr.recv_cq       = rcq;
 	attr.srq           = srq;
 	attr.xrcd	   = xrcd;
-	attr.sq_sig_type   = cmd.sq_sig_all ? IB_SIGNAL_ALL_WR : IB_SIGNAL_REQ_WR;
-	attr.qp_type       = cmd.qp_type;
+	attr.sq_sig_type   = cmd->sq_sig_all ? IB_SIGNAL_ALL_WR :
+					      IB_SIGNAL_REQ_WR;
+	attr.qp_type       = cmd->qp_type;
 	attr.create_flags  = 0;
 
-	attr.cap.max_send_wr     = cmd.max_send_wr;
-	attr.cap.max_recv_wr     = cmd.max_recv_wr;
-	attr.cap.max_send_sge    = cmd.max_send_sge;
-	attr.cap.max_recv_sge    = cmd.max_recv_sge;
-	attr.cap.max_inline_data = cmd.max_inline_data;
+	attr.cap.max_send_wr     = cmd->max_send_wr;
+	attr.cap.max_recv_wr     = cmd->max_recv_wr;
+	attr.cap.max_send_sge    = cmd->max_send_sge;
+	attr.cap.max_recv_sge    = cmd->max_recv_sge;
+	attr.cap.max_inline_data = cmd->max_inline_data;
 
 	obj->uevent.events_reported     = 0;
 	INIT_LIST_HEAD(&obj->uevent.event_list);
 	INIT_LIST_HEAD(&obj->mcast_list);
 
-	if (cmd.qp_type == IB_QPT_XRC_TGT)
+	if (cmd_sz >= offsetof(typeof(*cmd), create_flags) +
+		      sizeof(cmd->create_flags))
+		attr.create_flags = cmd->create_flags;
+
+	if (attr.create_flags) {
+		ret = -EINVAL;
+		goto err_put;
+	}
+
+	if (cmd->comp_mask) {
+		ret = -EINVAL;
+		goto err_put;
+	}
+
+	buf = (void *)cmd + sizeof(*cmd);
+	if (cmd_sz > sizeof(*cmd))
+		if (!(buf[0] == 0 && !memcmp(buf, buf + 1,
+					     cmd_sz - sizeof(*cmd) - 1))) {
+			ret = -EINVAL;
+			goto err_put;
+		}
+
+	if (cmd->qp_type == IB_QPT_XRC_TGT)
 		qp = ib_create_qp(pd, &attr);
 	else
-		qp = device->create_qp(pd, &attr, &udata);
+		qp = device->create_qp(pd, &attr, uhw);
 
 	if (IS_ERR(qp)) {
 		ret = PTR_ERR(qp);
 		goto err_put;
 	}
 
-	if (cmd.qp_type != IB_QPT_XRC_TGT) {
+	if (cmd->qp_type != IB_QPT_XRC_TGT) {
 		qp->real_qp	  = qp;
 		qp->device	  = device;
 		qp->pd		  = pd;
@@ -1854,19 +1877,20 @@ ssize_t ib_uverbs_create_qp(struct ib_uverbs_file *file,
 		goto err_destroy;
 
 	memset(&resp, 0, sizeof resp);
-	resp.qpn             = qp->qp_num;
-	resp.qp_handle       = obj->uevent.uobject.id;
-	resp.max_recv_sge    = attr.cap.max_recv_sge;
-	resp.max_send_sge    = attr.cap.max_send_sge;
-	resp.max_recv_wr     = attr.cap.max_recv_wr;
-	resp.max_send_wr     = attr.cap.max_send_wr;
-	resp.max_inline_data = attr.cap.max_inline_data;
+	resp.base.qpn             = qp->qp_num;
+	resp.base.qp_handle       = obj->uevent.uobject.id;
+	resp.base.max_recv_sge    = attr.cap.max_recv_sge;
+	resp.base.max_send_sge    = attr.cap.max_send_sge;
+	resp.base.max_recv_wr     = attr.cap.max_recv_wr;
+	resp.base.max_send_wr     = attr.cap.max_send_wr;
+	resp.base.max_inline_data = attr.cap.max_inline_data;
 
-	if (copy_to_user((void __user *) (unsigned long) cmd.response,
-			 &resp, sizeof resp)) {
-		ret = -EFAULT;
-		goto err_copy;
-	}
+	resp.response_length = offsetof(typeof(resp), response_length) +
+			       sizeof(resp.response_length);
+
+	ret = cb(file, &resp, ucore);
+	if (ret)
+		goto err_cb;
 
 	if (xrcd) {
 		obj->uxrcd = container_of(xrcd_uobj, struct ib_uxrcd_object,
@@ -1892,9 +1916,8 @@ ssize_t ib_uverbs_create_qp(struct ib_uverbs_file *file,
 
 	up_write(&obj->uevent.uobject.mutex);
 
-	return in_len;
-
-err_copy:
+	return 0;
+err_cb:
 	idr_remove_uobj(&ib_uverbs_qp_idr, &obj->uevent.uobject);
 
 err_destroy:
@@ -1916,6 +1939,111 @@ err_put:
 	return ret;
 }
 
+static int ib_uverbs_create_qp_cb(struct ib_uverbs_file *file,
+				  struct ib_uverbs_ex_create_qp_resp *resp,
+				  struct ib_udata *ucore)
+{
+	if (ib_copy_to_udata(ucore, &resp->base, sizeof(resp->base)))
+		return -EFAULT;
+
+	return 0;
+}
+
+ssize_t ib_uverbs_create_qp(struct ib_uverbs_file *file,
+			    const char __user *buf, int in_len,
+			    int out_len)
+{
+	struct ib_uverbs_create_qp      cmd;
+	struct ib_uverbs_ex_create_qp	cmd_ex;
+	struct ib_udata			ucore;
+	struct ib_udata			uhw;
+	ssize_t resp_size = sizeof(struct ib_uverbs_create_qp_resp);
+	int				err;
+
+	if (out_len < resp_size)
+		return -ENOSPC;
+
+	if (copy_from_user(&cmd, buf, sizeof(cmd)))
+		return -EFAULT;
+
+	INIT_UDATA(&ucore, buf, (unsigned long)cmd.response, sizeof(cmd),
+		   resp_size);
+	INIT_UDATA(&uhw, buf + sizeof(cmd),
+		   (unsigned long)cmd.response + resp_size,
+		   in_len - sizeof(cmd), out_len - resp_size);
+
+	memset(&cmd_ex, 0, sizeof(cmd_ex));
+	cmd_ex.user_handle = cmd.user_handle;
+	cmd_ex.pd_handle = cmd.pd_handle;
+	cmd_ex.send_cq_handle = cmd.send_cq_handle;
+	cmd_ex.recv_cq_handle = cmd.recv_cq_handle;
+	cmd_ex.srq_handle = cmd.srq_handle;
+	cmd_ex.max_send_wr = cmd.max_send_wr;
+	cmd_ex.max_recv_wr = cmd.max_recv_wr;
+	cmd_ex.max_send_sge = cmd.max_send_sge;
+	cmd_ex.max_recv_sge = cmd.max_recv_sge;
+	cmd_ex.max_inline_data = cmd.max_inline_data;
+	cmd_ex.sq_sig_all = cmd.sq_sig_all;
+	cmd_ex.qp_type = cmd.qp_type;
+	cmd_ex.is_srq = cmd.is_srq;
+
+	err = create_qp(file, &ucore, &uhw, &cmd_ex,
+			offsetof(typeof(cmd_ex), is_srq) +
+			sizeof(cmd.is_srq), ib_uverbs_create_qp_cb,
+			NULL);
+
+	if (err)
+		return err;
+
+	return in_len;
+}
+
+static int ib_uverbs_ex_create_qp_cb(struct ib_uverbs_file *file,
+				     struct ib_uverbs_ex_create_qp_resp *resp,
+				     struct ib_udata *ucore)
+{
+	if (ib_copy_to_udata(ucore, resp, resp->response_length))
+		return -EFAULT;
+
+	return 0;
+}
+
+int ib_uverbs_ex_create_qp(struct ib_uverbs_file *file,
+			   struct ib_udata *ucore,
+			   struct ib_udata *uhw)
+{
+	struct ib_uverbs_ex_create_qp_resp resp;
+	struct ib_uverbs_ex_create_qp cmd = {0};
+	int err;
+
+	if (ucore->inlen < (offsetof(typeof(cmd), comp_mask) +
+			    sizeof(cmd.comp_mask)))
+		return -EINVAL;
+
+	err = ib_copy_from_udata(&cmd, ucore, min(sizeof(cmd), ucore->inlen));
+	if (err)
+		return err;
+
+	if (cmd.comp_mask)
+		return -EINVAL;
+
+	if (cmd.reserved)
+		return -EINVAL;
+
+	if (ucore->outlen < (offsetof(typeof(resp), response_length) +
+			     sizeof(resp.response_length)))
+		return -ENOSPC;
+
+	err = create_qp(file, ucore, uhw, &cmd,
+			min(ucore->inlen, sizeof(cmd)),
+			ib_uverbs_ex_create_qp_cb, NULL);
+
+	if (err)
+		return err;
+
+	return 0;
+}
+
 ssize_t ib_uverbs_open_qp(struct ib_uverbs_file *file,
 			  const char __user *buf, int in_len, int out_len)
 {
diff --git a/drivers/infiniband/core/uverbs_main.c b/drivers/infiniband/core/uverbs_main.c
index 46c9229..392a05b 100644
--- a/drivers/infiniband/core/uverbs_main.c
+++ b/drivers/infiniband/core/uverbs_main.c
@@ -125,6 +125,7 @@ static int (*uverbs_ex_cmd_table[])(struct ib_uverbs_file *file,
 	[IB_USER_VERBS_EX_CMD_DESTROY_FLOW]	= ib_uverbs_ex_destroy_flow,
 	[IB_USER_VERBS_EX_CMD_QUERY_DEVICE]	= ib_uverbs_ex_query_device,
 	[IB_USER_VERBS_EX_CMD_CREATE_CQ]	= ib_uverbs_ex_create_cq,
+	[IB_USER_VERBS_EX_CMD_CREATE_QP]        = ib_uverbs_ex_create_qp,
 };
 
 static void ib_uverbs_add_one(struct ib_device *device);
diff --git a/include/uapi/rdma/ib_user_verbs.h b/include/uapi/rdma/ib_user_verbs.h
index 978841e..8126c14 100644
--- a/include/uapi/rdma/ib_user_verbs.h
+++ b/include/uapi/rdma/ib_user_verbs.h
@@ -92,6 +92,7 @@ enum {
 enum {
 	IB_USER_VERBS_EX_CMD_QUERY_DEVICE = IB_USER_VERBS_CMD_QUERY_DEVICE,
 	IB_USER_VERBS_EX_CMD_CREATE_CQ = IB_USER_VERBS_CMD_CREATE_CQ,
+	IB_USER_VERBS_EX_CMD_CREATE_QP = IB_USER_VERBS_CMD_CREATE_QP,
 	IB_USER_VERBS_EX_CMD_CREATE_FLOW = IB_USER_VERBS_CMD_THRESHOLD,
 	IB_USER_VERBS_EX_CMD_DESTROY_FLOW,
 };
@@ -516,6 +517,25 @@ struct ib_uverbs_create_qp {
 	__u64 driver_data[0];
 };
 
+struct ib_uverbs_ex_create_qp {
+	__u64 user_handle;
+	__u32 pd_handle;
+	__u32 send_cq_handle;
+	__u32 recv_cq_handle;
+	__u32 srq_handle;
+	__u32 max_send_wr;
+	__u32 max_recv_wr;
+	__u32 max_send_sge;
+	__u32 max_recv_sge;
+	__u32 max_inline_data;
+	__u8  sq_sig_all;
+	__u8  qp_type;
+	__u8  is_srq;
+	__u8 reserved;
+	__u32 comp_mask;
+	__u32 create_flags;
+};
+
 struct ib_uverbs_open_qp {
 	__u64 response;
 	__u64 user_handle;
@@ -538,6 +558,12 @@ struct ib_uverbs_create_qp_resp {
 	__u32 reserved;
 };
 
+struct ib_uverbs_ex_create_qp_resp {
+	struct ib_uverbs_create_qp_resp base;
+	__u32 comp_mask;
+	__u32 response_length;
+};
+
 /*
  * This struct needs to remain a multiple of 8 bytes to keep the
  * alignment of the modify QP parameters.
-- 
1.8.3.1

--
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 for-next 2/7] IB/core: Allow setting create flags in QP init attribute

[Eran Ben Elisha]

Allow setting IB_QP_CREATE_BLOCK_MULTICAST_LOOPBACK at create_flags in
ib_uverbs_create_qp_ex.

Signed-off-by: Eran Ben Elisha <eranbe-VPRAkNaXOzVWk0Htik3J/w@public.gmane.org>
---
 drivers/infiniband/core/uverbs_cmd.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/drivers/infiniband/core/uverbs_cmd.c b/drivers/infiniband/core/uverbs_cmd.c
index 2bac2f1..f3ea3ed 100644
--- a/drivers/infiniband/core/uverbs_cmd.c
+++ b/drivers/infiniband/core/uverbs_cmd.c
@@ -1823,7 +1823,7 @@ static int create_qp(struct ib_uverbs_file *file,
 		      sizeof(cmd->create_flags))
 		attr.create_flags = cmd->create_flags;
 
-	if (attr.create_flags) {
+	if (attr.create_flags & ~IB_QP_CREATE_BLOCK_MULTICAST_LOOPBACK) {
 		ret = -EINVAL;
 		goto err_put;
 	}
-- 
1.8.3.1

--
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 for-next 3/7] net/mlx4_core: Add support for filtering multicast loopback

[Eran Ben Elisha]

From: Maor Gottlieb <maorg-VPRAkNaXOzVWk0Htik3J/w@public.gmane.org>

Update device capabilities regarding HW filtering multicast loopback support.

Add MLX4_UPDATE_QP_ETH_SRC_CHECK_MC_LB attribute to mlx4_update_qp to
enable changing QP context to support filtering incoming multicast
loopback traffic according the sender's counter index.

Set the corresponding bits in QP context to force the loopback source
checks if attribute is given and HW supports it.

Signed-off-by: Maor Gottlieb <maorg-VPRAkNaXOzVWk0Htik3J/w@public.gmane.org>
Signed-off-by: Eran Ben Elisha <eranbe-VPRAkNaXOzVWk0Htik3J/w@public.gmane.org>
---
 drivers/net/ethernet/mellanox/mlx4/fw.c            |  6 +++++
 drivers/net/ethernet/mellanox/mlx4/qp.c            | 19 +++++++++++++-
 .../net/ethernet/mellanox/mlx4/resource_tracker.c  | 30 +++++++++++++++++-----
 include/linux/mlx4/device.h                        |  2 ++
 include/linux/mlx4/qp.h                            | 24 +++++++++++++----
 5 files changed, 68 insertions(+), 13 deletions(-)

diff --git a/drivers/net/ethernet/mellanox/mlx4/fw.c b/drivers/net/ethernet/mellanox/mlx4/fw.c
index e30bf57..5218e1e 100644
--- a/drivers/net/ethernet/mellanox/mlx4/fw.c
+++ b/drivers/net/ethernet/mellanox/mlx4/fw.c
@@ -154,6 +154,8 @@ static void dump_dev_cap_flags2(struct mlx4_dev *dev, u64 flags)
 		[26] = "Port ETS Scheduler support",
 		[27] = "Port beacon support",
 		[28] = "RX-ALL support",
+		[31] = "Modifying loopback source checks using UPDATE_QP support",
+		[32] = "Loopback source checks support",
 	};
 	int i;
 
@@ -946,6 +948,10 @@ int mlx4_QUERY_DEV_CAP(struct mlx4_dev *dev, struct mlx4_dev_cap *dev_cap)
 	MLX4_GET(field32, outbox, QUERY_DEV_CAP_EXT_2_FLAGS_OFFSET);
 	if (field32 & (1 << 16))
 		dev_cap->flags2 |= MLX4_DEV_CAP_FLAG2_UPDATE_QP;
+	if (field32 & (1 << 18))
+		dev_cap->flags2 |= MLX4_DEV_CAP_FLAG2_UPDATE_QP_SRC_CHECK_LB;
+	if (field32 & (1 << 19))
+		dev_cap->flags2 |= MLX4_DEV_CAP_FLAG2_LB_SRC_CHK;
 	if (field32 & (1 << 26))
 		dev_cap->flags2 |= MLX4_DEV_CAP_FLAG2_VLAN_CONTROL;
 	if (field32 & (1 << 20))
diff --git a/drivers/net/ethernet/mellanox/mlx4/qp.c b/drivers/net/ethernet/mellanox/mlx4/qp.c
index 2026863..b162495 100644
--- a/drivers/net/ethernet/mellanox/mlx4/qp.c
+++ b/drivers/net/ethernet/mellanox/mlx4/qp.c
@@ -436,6 +436,23 @@ int mlx4_update_qp(struct mlx4_dev *dev, u32 qpn,
 		cmd->qp_context.pri_path.grh_mylmc = params->smac_index;
 	}
 
+	if (attr & MLX4_UPDATE_QP_ETH_SRC_CHECK_MC_LB) {
+		if (!(dev->caps.flags2
+		      & MLX4_DEV_CAP_FLAG2_UPDATE_QP_SRC_CHECK_LB)) {
+			mlx4_warn(dev,
+				  "Trying to set src check LB, but it isn't supported\n");
+			err = -ENOTSUPP;
+			goto out;
+		}
+		pri_addr_path_mask |=
+			1ULL << MLX4_UPD_QP_PATH_MASK_ETH_SRC_CHECK_MC_LB;
+		if (params->flags &
+		    MLX4_UPDATE_QP_PARAMS_FLAGS_ETH_CHECK_MC_LB) {
+			cmd->qp_context.pri_path.fl |=
+				MLX4_FL_ETH_SRC_CHECK_MC_LB;
+		}
+	}
+
 	if (attr & MLX4_UPDATE_QP_VSD) {
 		qp_mask |= 1ULL << MLX4_UPD_QP_MASK_VSD;
 		if (params->flags & MLX4_UPDATE_QP_PARAMS_FLAGS_VSD_ENABLE)
@@ -458,7 +475,7 @@ int mlx4_update_qp(struct mlx4_dev *dev, u32 qpn,
 	err = mlx4_cmd(dev, mailbox->dma, qpn & 0xffffff, 0,
 		       MLX4_CMD_UPDATE_QP, MLX4_CMD_TIME_CLASS_A,
 		       MLX4_CMD_NATIVE);
-
+out:
 	mlx4_free_cmd_mailbox(dev, mailbox);
 	return err;
 }
diff --git a/drivers/net/ethernet/mellanox/mlx4/resource_tracker.c b/drivers/net/ethernet/mellanox/mlx4/resource_tracker.c
index 731423c..502f335 100644
--- a/drivers/net/ethernet/mellanox/mlx4/resource_tracker.c
+++ b/drivers/net/ethernet/mellanox/mlx4/resource_tracker.c
@@ -770,9 +770,12 @@ static int update_vport_qp_param(struct mlx4_dev *dev,
 			}
 		}
 
+		/* preserve IF_COUNTER flag */
+		qpc->pri_path.vlan_control &=
+			MLX4_CTRL_ETH_SRC_CHECK_IF_COUNTER;
 		if (vp_oper->state.link_state == IFLA_VF_LINK_STATE_DISABLE &&
 		    dev->caps.flags2 & MLX4_DEV_CAP_FLAG2_UPDATE_QP) {
-			qpc->pri_path.vlan_control =
+			qpc->pri_path.vlan_control |=
 				MLX4_VLAN_CTRL_ETH_TX_BLOCK_TAGGED |
 				MLX4_VLAN_CTRL_ETH_TX_BLOCK_PRIO_TAGGED |
 				MLX4_VLAN_CTRL_ETH_TX_BLOCK_UNTAGGED |
@@ -780,12 +783,12 @@ static int update_vport_qp_param(struct mlx4_dev *dev,
 				MLX4_VLAN_CTRL_ETH_RX_BLOCK_UNTAGGED |
 				MLX4_VLAN_CTRL_ETH_RX_BLOCK_TAGGED;
 		} else if (0 != vp_oper->state.default_vlan) {
-			qpc->pri_path.vlan_control =
+			qpc->pri_path.vlan_control |=
 				MLX4_VLAN_CTRL_ETH_TX_BLOCK_TAGGED |
 				MLX4_VLAN_CTRL_ETH_RX_BLOCK_PRIO_TAGGED |
 				MLX4_VLAN_CTRL_ETH_RX_BLOCK_UNTAGGED;
 		} else { /* priority tagged */
-			qpc->pri_path.vlan_control =
+			qpc->pri_path.vlan_control |=
 				MLX4_VLAN_CTRL_ETH_TX_BLOCK_TAGGED |
 				MLX4_VLAN_CTRL_ETH_RX_BLOCK_TAGGED;
 		}
@@ -3762,9 +3765,6 @@ int mlx4_INIT2RTR_QP_wrapper(struct mlx4_dev *dev, int slave,
 	update_gid(dev, inbox, (u8)slave);
 	adjust_proxy_tun_qkey(dev, vhcr, qpc);
 	orig_sched_queue = qpc->pri_path.sched_queue;
-	err = update_vport_qp_param(dev, inbox, slave, qpn);
-	if (err)
-		return err;
 
 	err = get_res(dev, slave, qpn, RES_QP, &qp);
 	if (err)
@@ -3774,6 +3774,10 @@ int mlx4_INIT2RTR_QP_wrapper(struct mlx4_dev *dev, int slave,
 		goto out;
 	}
 
+	err = update_vport_qp_param(dev, inbox, slave, qpn);
+	if (err)
+		goto out;
+
 	err = mlx4_DMA_wrapper(dev, slave, vhcr, inbox, outbox, cmd);
 out:
 	/* if no error, save sched queue value passed in by VF. This is
@@ -4208,7 +4212,9 @@ static int add_eth_header(struct mlx4_dev *dev, int slave,
 
 }
 
-#define MLX4_UPD_QP_PATH_MASK_SUPPORTED (1ULL << MLX4_UPD_QP_PATH_MASK_MAC_INDEX)
+#define MLX4_UPD_QP_PATH_MASK_SUPPORTED      (                                \
+	1ULL << MLX4_UPD_QP_PATH_MASK_MAC_INDEX                     |\
+	1ULL << MLX4_UPD_QP_PATH_MASK_ETH_SRC_CHECK_MC_LB)
 int mlx4_UPDATE_QP_wrapper(struct mlx4_dev *dev, int slave,
 			   struct mlx4_vhcr *vhcr,
 			   struct mlx4_cmd_mailbox *inbox,
@@ -4231,6 +4237,16 @@ int mlx4_UPDATE_QP_wrapper(struct mlx4_dev *dev, int slave,
 	    (pri_addr_path_mask & ~MLX4_UPD_QP_PATH_MASK_SUPPORTED))
 		return -EPERM;
 
+	if ((pri_addr_path_mask &
+	     (1ULL << MLX4_UPD_QP_PATH_MASK_ETH_SRC_CHECK_MC_LB)) &&
+		!(dev->caps.flags2 &
+		  MLX4_DEV_CAP_FLAG2_UPDATE_QP_SRC_CHECK_LB)) {
+			mlx4_warn(dev,
+				  "Src check LB for slave %d isn't supported\n",
+				   slave);
+		return -ENOTSUPP;
+	}
+
 	/* Just change the smac for the QP */
 	err = get_res(dev, slave, qpn, RES_QP, &rqp);
 	if (err) {
diff --git a/include/linux/mlx4/device.h b/include/linux/mlx4/device.h
index 5c7687a..3b9602f 100644
--- a/include/linux/mlx4/device.h
+++ b/include/linux/mlx4/device.h
@@ -212,6 +212,8 @@ enum {
 	MLX4_DEV_CAP_FLAG2_ETS_CFG		= 1LL <<  26,
 	MLX4_DEV_CAP_FLAG2_PORT_BEACON		= 1LL <<  27,
 	MLX4_DEV_CAP_FLAG2_IGNORE_FCS		= 1LL <<  28,
+	MLX4_DEV_CAP_FLAG2_UPDATE_QP_SRC_CHECK_LB = 1ULL << 31,
+	MLX4_DEV_CAP_FLAG2_LB_SRC_CHK           = 1ULL << 32,
 };
 
 enum {
diff --git a/include/linux/mlx4/qp.h b/include/linux/mlx4/qp.h
index 6fed539..62175c2 100644
--- a/include/linux/mlx4/qp.h
+++ b/include/linux/mlx4/qp.h
@@ -135,7 +135,10 @@ struct mlx4_rss_context {
 
 struct mlx4_qp_path {
 	u8			fl;
-	u8			vlan_control;
+	union {
+		u8			vlan_control;
+		u8			control;
+	};
 	u8			disable_pkey_check;
 	u8			pkey_index;
 	u8			counter_index;
@@ -156,9 +159,16 @@ struct mlx4_qp_path {
 };
 
 enum { /* fl */
-	MLX4_FL_CV      = 1 << 6,
-	MLX4_FL_ETH_HIDE_CQE_VLAN       = 1 << 2
+	MLX4_FL_CV	= 1 << 6,
+	MLX4_FL_ETH_HIDE_CQE_VLAN	= 1 << 2,
+	MLX4_FL_ETH_SRC_CHECK_MC_LB	= 1 << 1,
+	MLX4_FL_ETH_SRC_CHECK_UC_LB	= 1 << 0,
 };
+
+enum { /* control */
+	MLX4_CTRL_ETH_SRC_CHECK_IF_COUNTER	= 1 << 7,
+};
+
 enum { /* vlan_control */
 	MLX4_VLAN_CTRL_ETH_TX_BLOCK_TAGGED	= 1 << 6,
 	MLX4_VLAN_CTRL_ETH_TX_BLOCK_PRIO_TAGGED	= 1 << 5, /* 802.1p priority tag */
@@ -254,6 +264,8 @@ enum {
 	MLX4_UPD_QP_PATH_MASK_SCHED_QUEUE		= 14 + 32,
 	MLX4_UPD_QP_PATH_MASK_IF_COUNTER_INDEX		= 15 + 32,
 	MLX4_UPD_QP_PATH_MASK_FVL_RX			= 16 + 32,
+	MLX4_UPD_QP_PATH_MASK_ETH_SRC_CHECK_UC_LB	= 18 + 32,
+	MLX4_UPD_QP_PATH_MASK_ETH_SRC_CHECK_MC_LB	= 19 + 32,
 };
 
 enum { /* param3 */
@@ -435,11 +447,13 @@ enum mlx4_update_qp_attr {
 	MLX4_UPDATE_QP_VSD		= 1 << 1,
 	MLX4_UPDATE_QP_RATE_LIMIT	= 1 << 2,
 	MLX4_UPDATE_QP_QOS_VPORT	= 1 << 3,
-	MLX4_UPDATE_QP_SUPPORTED_ATTRS	= (1 << 4) - 1
+	MLX4_UPDATE_QP_ETH_SRC_CHECK_MC_LB      = 1 << 4,
+	MLX4_UPDATE_QP_SUPPORTED_ATTRS	= (1 << 5) - 1
 };
 
 enum mlx4_update_qp_params_flags {
-	MLX4_UPDATE_QP_PARAMS_FLAGS_VSD_ENABLE		= 1 << 0,
+	MLX4_UPDATE_QP_PARAMS_FLAGS_ETH_CHECK_MC_LB     = 1 << 0,
+	MLX4_UPDATE_QP_PARAMS_FLAGS_VSD_ENABLE		= 1 << 1,
 };
 
 struct mlx4_update_qp_params {
-- 
1.8.3.1

--
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 for-next 4/7] net/mlx4_en: Implement mcast loopback prevention for ETH qps

[Eran Ben Elisha]

From: Maor Gottlieb <maorg-VPRAkNaXOzVWk0Htik3J/w@public.gmane.org>

Set the mcast loopback prevention bit in the QPC for ETH MLX QPs (not
RSS QPs), when the firmware supports this feature. In addition, all rx
ring QPs need to be updated in order not to enforce loopback checks.
This prevents getting packets we sent both from the network stack and
the HCA. Loopback prevention is done by comparing the counter indices of
the sent and receiving QPs. If they're equal, packets aren't
loopback-ed.

Signed-off-by: Maor Gottlieb <maorg-VPRAkNaXOzVWk0Htik3J/w@public.gmane.org>
Signed-off-by: Eran Ben Elisha <eranbe-VPRAkNaXOzVWk0Htik3J/w@public.gmane.org>
---
 drivers/net/ethernet/mellanox/mlx4/en_main.c      | 22 ++++++++++++++++++++
 drivers/net/ethernet/mellanox/mlx4/en_resources.c | 25 +++++++++++++++++++++++
 drivers/net/ethernet/mellanox/mlx4/mlx4_en.h      |  3 ++-
 3 files changed, 49 insertions(+), 1 deletion(-)

diff --git a/drivers/net/ethernet/mellanox/mlx4/en_main.c b/drivers/net/ethernet/mellanox/mlx4/en_main.c
index a946e4b..70e381a 100644
--- a/drivers/net/ethernet/mellanox/mlx4/en_main.c
+++ b/drivers/net/ethernet/mellanox/mlx4/en_main.c
@@ -123,6 +123,28 @@ void mlx4_en_update_loopback_state(struct net_device *dev,
 	 */
 	if (mlx4_is_mfunc(priv->mdev->dev) || priv->validate_loopback)
 		priv->flags |= MLX4_EN_FLAG_ENABLE_HW_LOOPBACK;
+
+	mutex_lock(&priv->mdev->state_lock);
+	if (priv->mdev->dev->caps.flags2 &
+	    MLX4_DEV_CAP_FLAG2_UPDATE_QP_SRC_CHECK_LB &&
+	    priv->rss_map.indir_qp.qpn) {
+		int i;
+		int err = 0;
+
+		for (i = 0; i < priv->rx_ring_num; i++) {
+			int ret;
+
+			ret = mlx4_en_change_mcast_lb(priv,
+						      &priv->rss_map.qps[i],
+						      !!(features &
+							 NETIF_F_LOOPBACK));
+			if (!err)
+				err = ret;
+		}
+		if (err)
+			mlx4_warn(priv->mdev, "failed to change mcast loopback\n");
+	}
+	mutex_unlock(&priv->mdev->state_lock);
 }
 
 static int mlx4_en_get_profile(struct mlx4_en_dev *mdev)
diff --git a/drivers/net/ethernet/mellanox/mlx4/en_resources.c b/drivers/net/ethernet/mellanox/mlx4/en_resources.c
index e482fa1b..12aab5a 100644
--- a/drivers/net/ethernet/mellanox/mlx4/en_resources.c
+++ b/drivers/net/ethernet/mellanox/mlx4/en_resources.c
@@ -69,6 +69,15 @@ void mlx4_en_fill_qp_context(struct mlx4_en_priv *priv, int size, int stride,
 	context->pri_path.counter_index = priv->counter_index;
 	context->cqn_send = cpu_to_be32(cqn);
 	context->cqn_recv = cpu_to_be32(cqn);
+	if (!rss &&
+	    (mdev->dev->caps.flags2 & MLX4_DEV_CAP_FLAG2_LB_SRC_CHK) &&
+	    context->pri_path.counter_index !=
+			    MLX4_SINK_COUNTER_INDEX(mdev->dev)) {
+		/* disable multicast loopback to qp with same counter */
+		if (!(dev->features & NETIF_F_LOOPBACK))
+			context->pri_path.fl |= MLX4_FL_ETH_SRC_CHECK_MC_LB;
+		context->pri_path.control |= MLX4_CTRL_ETH_SRC_CHECK_IF_COUNTER;
+	}
 	context->db_rec_addr = cpu_to_be64(priv->res.db.dma << 2);
 	if (!(dev->features & NETIF_F_HW_VLAN_CTAG_RX))
 		context->param3 |= cpu_to_be32(1 << 30);
@@ -80,6 +89,22 @@ void mlx4_en_fill_qp_context(struct mlx4_en_priv *priv, int size, int stride,
 	}
 }
 
+int mlx4_en_change_mcast_lb(struct mlx4_en_priv *priv, struct mlx4_qp *qp,
+			    int loopback)
+{
+	int ret;
+	struct mlx4_update_qp_params qp_params;
+
+	memset(&qp_params, 0, sizeof(qp_params));
+	if (!loopback)
+		qp_params.flags = MLX4_UPDATE_QP_PARAMS_FLAGS_ETH_CHECK_MC_LB;
+
+	ret = mlx4_update_qp(priv->mdev->dev, qp->qpn,
+			     MLX4_UPDATE_QP_ETH_SRC_CHECK_MC_LB,
+			     &qp_params);
+
+	return ret;
+}
 
 int mlx4_en_map_buffer(struct mlx4_buf *buf)
 {
diff --git a/drivers/net/ethernet/mellanox/mlx4/mlx4_en.h b/drivers/net/ethernet/mellanox/mlx4/mlx4_en.h
index 666d166..7db86d4 100644
--- a/drivers/net/ethernet/mellanox/mlx4/mlx4_en.h
+++ b/drivers/net/ethernet/mellanox/mlx4/mlx4_en.h
@@ -797,7 +797,8 @@ void mlx4_en_fill_qp_context(struct mlx4_en_priv *priv, int size, int stride,
 void mlx4_en_sqp_event(struct mlx4_qp *qp, enum mlx4_event event);
 int mlx4_en_map_buffer(struct mlx4_buf *buf);
 void mlx4_en_unmap_buffer(struct mlx4_buf *buf);
-
+int mlx4_en_change_mcast_lb(struct mlx4_en_priv *priv, struct mlx4_qp *qp,
+			    int loopback);
 void mlx4_en_calc_rx_buf(struct net_device *dev);
 int mlx4_en_config_rss_steer(struct mlx4_en_priv *priv);
 void mlx4_en_release_rss_steer(struct mlx4_en_priv *priv);
-- 
1.8.3.1

--
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 for-next 5/7] IB/mlx4: Add IB counters table

[Eran Ben Elisha]

This is an infrastructure step for allocating and attaching more than
one counter to QPs on the same port. Allocate a counters table and
manage the insertion and removals of the counters in load and unload of
mlx4 IB.

Signed-off-by: Eran Ben Elisha <eranbe-VPRAkNaXOzVWk0Htik3J/w@public.gmane.org>
---
 drivers/infiniband/hw/mlx4/mad.c     | 25 ++++++++++----
 drivers/infiniband/hw/mlx4/main.c    | 63 ++++++++++++++++++++++++++++--------
 drivers/infiniband/hw/mlx4/mlx4_ib.h |  9 +++++-
 drivers/infiniband/hw/mlx4/qp.c      |  8 +++--
 4 files changed, 81 insertions(+), 24 deletions(-)

diff --git a/drivers/infiniband/hw/mlx4/mad.c b/drivers/infiniband/hw/mlx4/mad.c
index 1cd75ff..68f2567 100644
--- a/drivers/infiniband/hw/mlx4/mad.c
+++ b/drivers/infiniband/hw/mlx4/mad.c
@@ -824,18 +824,29 @@ static int iboe_process_mad(struct ib_device *ibdev, int mad_flags, u8 port_num,
 {
 	struct mlx4_counter counter_stats;
 	struct mlx4_ib_dev *dev = to_mdev(ibdev);
-	int err;
+	struct counter_index *tmp_counter;
+	int err = IB_MAD_RESULT_FAILURE, stats_avail = 0;
 
 	if (in_mad->mad_hdr.mgmt_class != IB_MGMT_CLASS_PERF_MGMT)
 		return -EINVAL;
 
 	memset(&counter_stats, 0, sizeof(counter_stats));
-	err = mlx4_get_counter_stats(dev->dev,
-				     dev->counters[port_num - 1].index,
-				     &counter_stats, 0);
-	if (err)
-		err = IB_MAD_RESULT_FAILURE;
-	else {
+	mutex_lock(&dev->counters_table[port_num - 1].mutex);
+	list_for_each_entry(tmp_counter,
+			    &dev->counters_table[port_num - 1].counters_list,
+			    list) {
+		err = mlx4_get_counter_stats(dev->dev,
+					     tmp_counter->index,
+					     &counter_stats, 0);
+		if (err) {
+			err = IB_MAD_RESULT_FAILURE;
+			stats_avail = 0;
+			break;
+		}
+		stats_avail = 1;
+	}
+	mutex_unlock(&dev->counters_table[port_num - 1].mutex);
+	if (stats_avail) {
 		memset(out_mad->data, 0, sizeof out_mad->data);
 		switch (counter_stats.counter_mode & 0xf) {
 		case 0:
diff --git a/drivers/infiniband/hw/mlx4/main.c b/drivers/infiniband/hw/mlx4/main.c
index 1437ed5..d0d5b55 100644
--- a/drivers/infiniband/hw/mlx4/main.c
+++ b/drivers/infiniband/hw/mlx4/main.c
@@ -1113,6 +1113,22 @@ static int add_gid_entry(struct ib_qp *ibqp, union ib_gid *gid)
 	return 0;
 }
 
+static void mlx4_ib_delete_counters_table(struct mlx4_ib_dev *ibdev,
+					  struct mlx4_ib_counters *ctr_table)
+{
+	struct counter_index *counter, *tmp_count;
+
+	mutex_lock(&ctr_table->mutex);
+	list_for_each_entry_safe(counter, tmp_count, &ctr_table->counters_list,
+				 list) {
+		if (counter->allocated)
+			mlx4_counter_free(ibdev->dev, counter->index);
+		list_del(&counter->list);
+		kfree(counter);
+	}
+	mutex_unlock(&ctr_table->mutex);
+}
+
 int mlx4_ib_add_mc(struct mlx4_ib_dev *mdev, struct mlx4_ib_qp *mqp,
 		   union ib_gid *gid)
 {
@@ -1997,6 +2013,7 @@ static void *mlx4_ib_add(struct mlx4_dev *dev)
 	int num_req_counters;
 	int allocated;
 	u32 counter_index;
+	struct counter_index *new_counter_index = NULL;
 
 	pr_info_once("%s", mlx4_ib_version);
 
@@ -2167,6 +2184,11 @@ static void *mlx4_ib_add(struct mlx4_dev *dev)
 	if (init_node_data(ibdev))
 		goto err_map;
 
+	for (i = 0; i < ibdev->num_ports; ++i) {
+		mutex_init(&ibdev->counters_table[i].mutex);
+		INIT_LIST_HEAD(&ibdev->counters_table[i].counters_list);
+	}
+
 	num_req_counters = mlx4_is_bonded(dev) ? 1 : ibdev->num_ports;
 	for (i = 0; i < num_req_counters; ++i) {
 		mutex_init(&ibdev->qp1_proxy_lock[i]);
@@ -2185,15 +2207,34 @@ static void *mlx4_ib_add(struct mlx4_dev *dev)
 			counter_index = mlx4_get_default_counter_index(dev,
 								       i + 1);
 		}
-		ibdev->counters[i].index = counter_index;
-		ibdev->counters[i].allocated = allocated;
+		new_counter_index = kmalloc(sizeof(*new_counter_index),
+					    GFP_KERNEL);
+		if (!new_counter_index) {
+			if (allocated)
+				mlx4_counter_free(ibdev->dev, counter_index);
+			goto err_counter;
+		}
+		new_counter_index->index = counter_index;
+		new_counter_index->allocated = allocated;
+		list_add_tail(&new_counter_index->list,
+			      &ibdev->counters_table[i].counters_list);
+		ibdev->counters_table[i].default_counter = counter_index;
 		pr_info("counter index %d for port %d allocated %d\n",
 			counter_index, i + 1, allocated);
 	}
 	if (mlx4_is_bonded(dev))
 		for (i = 1; i < ibdev->num_ports ; ++i) {
-			ibdev->counters[i].index = ibdev->counters[0].index;
-			ibdev->counters[i].allocated = 0;
+			new_counter_index =
+					kmalloc(sizeof(struct counter_index),
+						GFP_KERNEL);
+			if (!new_counter_index)
+				goto err_counter;
+			new_counter_index->index = counter_index;
+			new_counter_index->allocated = 0;
+			list_add_tail(&new_counter_index->list,
+				      &ibdev->counters_table[i].counters_list);
+			ibdev->counters_table[i].default_counter =
+								counter_index;
 		}
 
 	mlx4_foreach_port(i, dev, MLX4_PORT_TYPE_IB)
@@ -2302,12 +2343,9 @@ err_steer_qp_release:
 		mlx4_qp_release_range(dev, ibdev->steer_qpn_base,
 				      ibdev->steer_qpn_count);
 err_counter:
-	for (i = 0; i < ibdev->num_ports; ++i) {
-		if (ibdev->counters[i].index != -1 &&
-		    ibdev->counters[i].allocated)
-			mlx4_counter_free(ibdev->dev,
-					  ibdev->counters[i].index);
-	}
+	for (i = 0; i < ibdev->num_ports; ++i)
+		mlx4_ib_delete_counters_table(ibdev, &ibdev->counters_table[i]);
+
 err_map:
 	iounmap(ibdev->uar_map);
 
@@ -2411,9 +2449,8 @@ static void mlx4_ib_remove(struct mlx4_dev *dev, void *ibdev_ptr)
 
 	iounmap(ibdev->uar_map);
 	for (p = 0; p < ibdev->num_ports; ++p)
-		if (ibdev->counters[p].index != -1 &&
-		    ibdev->counters[p].allocated)
-			mlx4_counter_free(ibdev->dev, ibdev->counters[p].index);
+		mlx4_ib_delete_counters_table(ibdev, &ibdev->counters_table[p]);
+
 	mlx4_foreach_port(p, dev, MLX4_PORT_TYPE_IB)
 		mlx4_CLOSE_PORT(dev, p);
 
diff --git a/drivers/infiniband/hw/mlx4/mlx4_ib.h b/drivers/infiniband/hw/mlx4/mlx4_ib.h
index fe52ead..2b98459 100644
--- a/drivers/infiniband/hw/mlx4/mlx4_ib.h
+++ b/drivers/infiniband/hw/mlx4/mlx4_ib.h
@@ -515,10 +515,17 @@ struct mlx4_ib_iov_port {
 };
 
 struct counter_index {
+	struct  list_head       list;
 	u32		index;
 	u8		allocated;
 };
 
+struct mlx4_ib_counters {
+	struct list_head        counters_list;
+	struct mutex            mutex; /* mutex for accessing counters list */
+	u32			default_counter;
+};
+
 struct mlx4_ib_dev {
 	struct ib_device	ib_dev;
 	struct mlx4_dev	       *dev;
@@ -537,7 +544,7 @@ struct mlx4_ib_dev {
 	struct mutex		cap_mask_mutex;
 	bool			ib_active;
 	struct mlx4_ib_iboe	iboe;
-	struct counter_index    counters[MLX4_MAX_PORTS];
+	struct mlx4_ib_counters counters_table[MLX4_MAX_PORTS];
 	int		       *eq_table;
 	struct kobject	       *iov_parent;
 	struct kobject	       *ports_parent;
diff --git a/drivers/infiniband/hw/mlx4/qp.c b/drivers/infiniband/hw/mlx4/qp.c
index 4ad9be3..382913e 100644
--- a/drivers/infiniband/hw/mlx4/qp.c
+++ b/drivers/infiniband/hw/mlx4/qp.c
@@ -1460,6 +1460,7 @@ static int __mlx4_ib_modify_qp(struct ib_qp *ibqp,
 	int sqd_event;
 	int steer_qp = 0;
 	int err = -EINVAL;
+	int counter_index;
 
 	/* APM is not supported under RoCE */
 	if (attr_mask & IB_QP_ALT_PATH &&
@@ -1543,9 +1544,10 @@ static int __mlx4_ib_modify_qp(struct ib_qp *ibqp,
 	}
 
 	if (cur_state == IB_QPS_INIT && new_state == IB_QPS_RTR) {
-		if (dev->counters[qp->port - 1].index != -1) {
-			context->pri_path.counter_index =
-					dev->counters[qp->port - 1].index;
+		counter_index =
+			dev->counters_table[qp->port - 1].default_counter;
+		if (counter_index != -1) {
+			context->pri_path.counter_index = counter_index;
 			optpar |= MLX4_QP_OPTPAR_COUNTER_INDEX;
 		} else
 			context->pri_path.counter_index =
-- 
1.8.3.1

--
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 for-next 6/7] IB/mlx4: Add counter based implementation for QP multicast loopback block

[Eran Ben Elisha]

Current implementation for MLX4_IB_QP_BLOCK_MULTICAST_LOOPBACK is not
supported when link layer is Ethernet.

This patch will add counter based implementation for multicast loopback
prevention. HW can drop multicast loopback packets if sender QP counter
index is equal to receiver QP counter index. If qp flag
MLX4_IB_QP_BLOCK_MULTICAST_LOOPBACK is set and link layer is Ethernet,
create a new counter and attach it to the QP so it will continue
receiving multicast loopback traffic but it's own.

The decision if to create a new counter is being made at the qp
modification to RTR after the QP's port is set. When QP is destroyed or
moved back to reset state, delete the counter.

Signed-off-by: Eran Ben Elisha <eranbe-VPRAkNaXOzVWk0Htik3J/w@public.gmane.org>
---
 drivers/infiniband/hw/mlx4/mlx4_ib.h |  1 +
 drivers/infiniband/hw/mlx4/qp.c      | 67 ++++++++++++++++++++++++++++++++++++
 2 files changed, 68 insertions(+)

diff --git a/drivers/infiniband/hw/mlx4/mlx4_ib.h b/drivers/infiniband/hw/mlx4/mlx4_ib.h
index 2b98459..af8ae51 100644
--- a/drivers/infiniband/hw/mlx4/mlx4_ib.h
+++ b/drivers/infiniband/hw/mlx4/mlx4_ib.h
@@ -307,6 +307,7 @@ struct mlx4_ib_qp {
 	struct list_head	qps_list;
 	struct list_head	cq_recv_list;
 	struct list_head	cq_send_list;
+	struct counter_index	*counter_index;
 };
 
 struct mlx4_ib_srq {
diff --git a/drivers/infiniband/hw/mlx4/qp.c b/drivers/infiniband/hw/mlx4/qp.c
index 382913e..2871949 100644
--- a/drivers/infiniband/hw/mlx4/qp.c
+++ b/drivers/infiniband/hw/mlx4/qp.c
@@ -617,6 +617,18 @@ static int qp0_enabled_vf(struct mlx4_dev *dev, int qpn)
 	return 0;
 }
 
+static void mlx4_ib_free_qp_counter(struct mlx4_ib_dev *dev,
+				    struct mlx4_ib_qp *qp)
+{
+	mutex_lock(&dev->counters_table[qp->port - 1].mutex);
+	mlx4_counter_free(dev->dev, qp->counter_index->index);
+	list_del(&qp->counter_index->list);
+	mutex_unlock(&dev->counters_table[qp->port - 1].mutex);
+
+	kfree(qp->counter_index);
+	qp->counter_index = NULL;
+}
+
 static int create_qp_common(struct mlx4_ib_dev *dev, struct ib_pd *pd,
 			    struct ib_qp_init_attr *init_attr,
 			    struct ib_udata *udata, int sqpn, struct mlx4_ib_qp **caller_qp,
@@ -1189,6 +1201,9 @@ int mlx4_ib_destroy_qp(struct ib_qp *qp)
 		mutex_unlock(&dev->qp1_proxy_lock[mqp->port - 1]);
 	}
 
+	if (mqp->counter_index)
+		mlx4_ib_free_qp_counter(dev, mqp);
+
 	pd = get_pd(mqp);
 	destroy_qp_common(dev, mqp, !!pd->ibpd.uobject);
 
@@ -1447,6 +1462,40 @@ static int handle_eth_ud_smac_index(struct mlx4_ib_dev *dev, struct mlx4_ib_qp *
 	return 0;
 }
 
+static int create_qp_lb_counter(struct mlx4_ib_dev *dev, struct mlx4_ib_qp *qp)
+{
+	struct counter_index *new_counter_index;
+	int err;
+	u32 tmp_idx;
+
+	if (rdma_port_get_link_layer(&dev->ib_dev, qp->port) !=
+	    IB_LINK_LAYER_ETHERNET ||
+	    !(qp->flags & MLX4_IB_QP_BLOCK_MULTICAST_LOOPBACK) ||
+	    !(dev->dev->caps.flags2 & MLX4_DEV_CAP_FLAG2_LB_SRC_CHK))
+		return 0;
+
+	err = mlx4_counter_alloc(dev->dev, &tmp_idx);
+	if (err)
+		return err;
+
+	new_counter_index = kmalloc(sizeof(*new_counter_index), GFP_KERNEL);
+	if (!new_counter_index) {
+		mlx4_counter_free(dev->dev, tmp_idx);
+		return -ENOMEM;
+	}
+
+	new_counter_index->index = tmp_idx;
+	new_counter_index->allocated = 1;
+	qp->counter_index = new_counter_index;
+
+	mutex_lock(&dev->counters_table[qp->port - 1].mutex);
+	list_add_tail(&new_counter_index->list,
+		      &dev->counters_table[qp->port - 1].counters_list);
+	mutex_unlock(&dev->counters_table[qp->port - 1].mutex);
+
+	return 0;
+}
+
 static int __mlx4_ib_modify_qp(struct ib_qp *ibqp,
 			       const struct ib_qp_attr *attr, int attr_mask,
 			       enum ib_qp_state cur_state, enum ib_qp_state new_state)
@@ -1520,6 +1569,9 @@ static int __mlx4_ib_modify_qp(struct ib_qp *ibqp,
 		context->sq_size_stride = ilog2(qp->sq.wqe_cnt) << 3;
 	context->sq_size_stride |= qp->sq.wqe_shift - 4;
 
+	if (new_state == IB_QPS_RESET && qp->counter_index)
+		mlx4_ib_free_qp_counter(dev, qp);
+
 	if (cur_state == IB_QPS_RESET && new_state == IB_QPS_INIT) {
 		context->sq_size_stride |= !!qp->sq_no_prefetch << 7;
 		context->xrcd = cpu_to_be32((u32) qp->xrcdn);
@@ -1544,11 +1596,24 @@ static int __mlx4_ib_modify_qp(struct ib_qp *ibqp,
 	}
 
 	if (cur_state == IB_QPS_INIT && new_state == IB_QPS_RTR) {
+		err = create_qp_lb_counter(dev, qp);
+		if (err)
+			goto out;
+
 		counter_index =
 			dev->counters_table[qp->port - 1].default_counter;
+		if (qp->counter_index)
+			counter_index = qp->counter_index->index;
+
 		if (counter_index != -1) {
 			context->pri_path.counter_index = counter_index;
 			optpar |= MLX4_QP_OPTPAR_COUNTER_INDEX;
+			if (qp->counter_index) {
+				context->pri_path.fl |=
+					MLX4_FL_ETH_SRC_CHECK_MC_LB;
+				context->pri_path.vlan_control |=
+					MLX4_CTRL_ETH_SRC_CHECK_IF_COUNTER;
+			}
 		} else
 			context->pri_path.counter_index =
 				MLX4_SINK_COUNTER_INDEX(dev->dev);
@@ -1850,6 +1915,8 @@ static int __mlx4_ib_modify_qp(struct ib_qp *ibqp,
 		}
 	}
 out:
+	if (err && qp->counter_index)
+		mlx4_ib_free_qp_counter(dev, qp);
 	if (err && steer_qp)
 		mlx4_ib_steer_qp_reg(dev, qp, 0);
 	kfree(context);
-- 
1.8.3.1

--
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 for-next 7/7] IB/mlx4: Add support for blocking multicast loopback QP creation user flag

[Eran Ben Elisha]

MLX4_IB_QP_BLOCK_MULTICAST_LOOPBACK is now supported downstream.

In addition, this flag was supported only for IB_QPT_UD, now, with the
new implementation it is supported for all QP types.

Support IB_USER_VERBS_EX_CMD_CREATE_QP in order to get the flag from
user space using the extension create qp command.

Signed-off-by: Eran Ben Elisha <eranbe-VPRAkNaXOzVWk0Htik3J/w@public.gmane.org>
---
 drivers/infiniband/hw/mlx4/main.c |  3 ++-
 drivers/infiniband/hw/mlx4/qp.c   | 13 ++++++++-----
 2 files changed, 10 insertions(+), 6 deletions(-)

diff --git a/drivers/infiniband/hw/mlx4/main.c b/drivers/infiniband/hw/mlx4/main.c
index d0d5b55..4c56659 100644
--- a/drivers/infiniband/hw/mlx4/main.c
+++ b/drivers/infiniband/hw/mlx4/main.c
@@ -2175,7 +2175,8 @@ static void *mlx4_ib_add(struct mlx4_dev *dev)
 
 	ibdev->ib_dev.uverbs_ex_cmd_mask |=
 		(1ull << IB_USER_VERBS_EX_CMD_QUERY_DEVICE) |
-		(1ull << IB_USER_VERBS_EX_CMD_CREATE_CQ);
+		(1ull << IB_USER_VERBS_EX_CMD_CREATE_CQ) |
+		(1ull << IB_USER_VERBS_EX_CMD_CREATE_QP);
 
 	mlx4_ib_alloc_eqs(dev, ibdev);
 
diff --git a/drivers/infiniband/hw/mlx4/qp.c b/drivers/infiniband/hw/mlx4/qp.c
index 2871949..ba25a1b 100644
--- a/drivers/infiniband/hw/mlx4/qp.c
+++ b/drivers/infiniband/hw/mlx4/qp.c
@@ -758,9 +758,6 @@ static int create_qp_common(struct mlx4_ib_dev *dev, struct ib_pd *pd,
 	} else {
 		qp->sq_no_prefetch = 0;
 
-		if (init_attr->create_flags & IB_QP_CREATE_BLOCK_MULTICAST_LOOPBACK)
-			qp->flags |= MLX4_IB_QP_BLOCK_MULTICAST_LOOPBACK;
-
 		if (init_attr->create_flags & IB_QP_CREATE_IPOIB_UD_LSO)
 			qp->flags |= MLX4_IB_QP_LSO;
 
@@ -834,6 +831,9 @@ static int create_qp_common(struct mlx4_ib_dev *dev, struct ib_pd *pd,
 			goto err_proxy;
 	}
 
+	if (init_attr->create_flags & IB_QP_CREATE_BLOCK_MULTICAST_LOOPBACK)
+		qp->flags |= MLX4_IB_QP_BLOCK_MULTICAST_LOOPBACK;
+
 	err = mlx4_qp_alloc(dev->dev, qpn, &qp->mqp, gfp);
 	if (err)
 		goto err_qpn;
@@ -1098,6 +1098,7 @@ struct ib_qp *mlx4_ib_create_qp(struct ib_pd *pd,
 {
 	struct mlx4_ib_qp *qp = NULL;
 	int err;
+	int sup_u_create_flags = MLX4_IB_QP_BLOCK_MULTICAST_LOOPBACK;
 	u16 xrcdn = 0;
 	gfp_t gfp;
 
@@ -1121,8 +1122,10 @@ struct ib_qp *mlx4_ib_create_qp(struct ib_pd *pd,
 	}
 
 	if (init_attr->create_flags &&
-	    (udata ||
-	     ((init_attr->create_flags & ~(MLX4_IB_SRIOV_SQP | MLX4_IB_QP_CREATE_USE_GFP_NOIO)) &&
+	    ((udata && init_attr->create_flags & ~(sup_u_create_flags)) ||
+	     ((init_attr->create_flags & ~(MLX4_IB_SRIOV_SQP |
+					   MLX4_IB_QP_CREATE_USE_GFP_NOIO |
+					   MLX4_IB_QP_BLOCK_MULTICAST_LOOPBACK)) &&
 	      init_attr->qp_type != IB_QPT_UD) ||
 	     ((init_attr->create_flags & MLX4_IB_SRIOV_SQP) &&
 	      init_attr->qp_type > IB_QPT_GSI)))
-- 
1.8.3.1

--
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 for-next 4/7] net/mlx4_en: Implement mcast loopback prevention for ETH qps

[Yuval Shaia]

On Sun, Aug 16, 2015 at 07:05:30PM +0300, Eran Ben Elisha wrote:
> From: Maor Gottlieb <maorg-VPRAkNaXOzVWk0Htik3J/w@public.gmane.org>
> 
> Set the mcast loopback prevention bit in the QPC for ETH MLX QPs (not
> RSS QPs), when the firmware supports this feature. In addition, all rx
> ring QPs need to be updated in order not to enforce loopback checks.
> This prevents getting packets we sent both from the network stack and
> the HCA. Loopback prevention is done by comparing the counter indices of
> the sent and receiving QPs. If they're equal, packets aren't
> loopback-ed.
> 
> Signed-off-by: Maor Gottlieb <maorg-VPRAkNaXOzVWk0Htik3J/w@public.gmane.org>
> Signed-off-by: Eran Ben Elisha <eranbe-VPRAkNaXOzVWk0Htik3J/w@public.gmane.org>
> ---
>  drivers/net/ethernet/mellanox/mlx4/en_main.c      | 22 ++++++++++++++++++++
>  drivers/net/ethernet/mellanox/mlx4/en_resources.c | 25 +++++++++++++++++++++++
>  drivers/net/ethernet/mellanox/mlx4/mlx4_en.h      |  3 ++-
>  3 files changed, 49 insertions(+), 1 deletion(-)
> 
> diff --git a/drivers/net/ethernet/mellanox/mlx4/en_main.c b/drivers/net/ethernet/mellanox/mlx4/en_main.c
> index a946e4b..70e381a 100644
> --- a/drivers/net/ethernet/mellanox/mlx4/en_main.c
> +++ b/drivers/net/ethernet/mellanox/mlx4/en_main.c
> @@ -123,6 +123,28 @@ void mlx4_en_update_loopback_state(struct net_device *dev,
>  	 */
>  	if (mlx4_is_mfunc(priv->mdev->dev) || priv->validate_loopback)
>  		priv->flags |= MLX4_EN_FLAG_ENABLE_HW_LOOPBACK;
> +
> +	mutex_lock(&priv->mdev->state_lock);
> +	if (priv->mdev->dev->caps.flags2 &
> +	    MLX4_DEV_CAP_FLAG2_UPDATE_QP_SRC_CHECK_LB &&
> +	    priv->rss_map.indir_qp.qpn) {
> +		int i;
> +		int err = 0;
> +
> +		for (i = 0; i < priv->rx_ring_num; i++) {
> +			int ret;
> +
> +			ret = mlx4_en_change_mcast_lb(priv,
> +						      &priv->rss_map.qps[i],
> +						      !!(features &
> +							 NETIF_F_LOOPBACK));
If "feature" is not changing inside the loop can we take it out just for
the code to look more readable?
> +			if (!err)
> +				err = ret;
> +		}
> +		if (err)
> +			mlx4_warn(priv->mdev, "failed to change mcast loopback\n");
> +	}
> +	mutex_unlock(&priv->mdev->state_lock);
>  }
>  
>  static int mlx4_en_get_profile(struct mlx4_en_dev *mdev)
> diff --git a/drivers/net/ethernet/mellanox/mlx4/en_resources.c b/drivers/net/ethernet/mellanox/mlx4/en_resources.c
> index e482fa1b..12aab5a 100644
> --- a/drivers/net/ethernet/mellanox/mlx4/en_resources.c
> +++ b/drivers/net/ethernet/mellanox/mlx4/en_resources.c
> @@ -69,6 +69,15 @@ void mlx4_en_fill_qp_context(struct mlx4_en_priv *priv, int size, int stride,
>  	context->pri_path.counter_index = priv->counter_index;
>  	context->cqn_send = cpu_to_be32(cqn);
>  	context->cqn_recv = cpu_to_be32(cqn);
> +	if (!rss &&
> +	    (mdev->dev->caps.flags2 & MLX4_DEV_CAP_FLAG2_LB_SRC_CHK) &&
> +	    context->pri_path.counter_index !=
> +			    MLX4_SINK_COUNTER_INDEX(mdev->dev)) {
> +		/* disable multicast loopback to qp with same counter */
> +		if (!(dev->features & NETIF_F_LOOPBACK))
> +			context->pri_path.fl |= MLX4_FL_ETH_SRC_CHECK_MC_LB;
> +		context->pri_path.control |= MLX4_CTRL_ETH_SRC_CHECK_IF_COUNTER;
> +	}
>  	context->db_rec_addr = cpu_to_be64(priv->res.db.dma << 2);
>  	if (!(dev->features & NETIF_F_HW_VLAN_CTAG_RX))
>  		context->param3 |= cpu_to_be32(1 << 30);
> @@ -80,6 +89,22 @@ void mlx4_en_fill_qp_context(struct mlx4_en_priv *priv, int size, int stride,
>  	}
>  }
>  
> +int mlx4_en_change_mcast_lb(struct mlx4_en_priv *priv, struct mlx4_qp *qp,
> +			    int loopback)
> +{
> +	int ret;
> +	struct mlx4_update_qp_params qp_params;
> +
> +	memset(&qp_params, 0, sizeof(qp_params));
> +	if (!loopback)
> +		qp_params.flags = MLX4_UPDATE_QP_PARAMS_FLAGS_ETH_CHECK_MC_LB;
> +
> +	ret = mlx4_update_qp(priv->mdev->dev, qp->qpn,
> +			     MLX4_UPDATE_QP_ETH_SRC_CHECK_MC_LB,
> +			     &qp_params);
> +
> +	return ret;
> +}
>  
>  int mlx4_en_map_buffer(struct mlx4_buf *buf)
>  {
> diff --git a/drivers/net/ethernet/mellanox/mlx4/mlx4_en.h b/drivers/net/ethernet/mellanox/mlx4/mlx4_en.h
> index 666d166..7db86d4 100644
> --- a/drivers/net/ethernet/mellanox/mlx4/mlx4_en.h
> +++ b/drivers/net/ethernet/mellanox/mlx4/mlx4_en.h
> @@ -797,7 +797,8 @@ void mlx4_en_fill_qp_context(struct mlx4_en_priv *priv, int size, int stride,
>  void mlx4_en_sqp_event(struct mlx4_qp *qp, enum mlx4_event event);
>  int mlx4_en_map_buffer(struct mlx4_buf *buf);
>  void mlx4_en_unmap_buffer(struct mlx4_buf *buf);
> -
> +int mlx4_en_change_mcast_lb(struct mlx4_en_priv *priv, struct mlx4_qp *qp,
> +			    int loopback);
>  void mlx4_en_calc_rx_buf(struct net_device *dev);
>  int mlx4_en_config_rss_steer(struct mlx4_en_priv *priv);
>  void mlx4_en_release_rss_steer(struct mlx4_en_priv *priv);
> -- 
> 1.8.3.1
> 
> --
> 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
--
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 for-next 4/7] net/mlx4_en: Implement mcast loopback prevention for ETH qps

[eran ben elisha]

On Sun, Aug 16, 2015 at 10:19 PM, Yuval Shaia <yuval.shaia-QHcLZuEGTsvQT0dZR+AlfA@public.gmane.org> wrote:
>
> On Sun, Aug 16, 2015 at 07:05:30PM +0300, Eran Ben Elisha wrote:
> > From: Maor Gottlieb <maorg-VPRAkNaXOzVWk0Htik3J/w@public.gmane.org>
> >
> > Set the mcast loopback prevention bit in the QPC for ETH MLX QPs (not
> > RSS QPs), when the firmware supports this feature. In addition, all rx
> > ring QPs need to be updated in order not to enforce loopback checks.
> > This prevents getting packets we sent both from the network stack and
> > the HCA. Loopback prevention is done by comparing the counter indices of
> > the sent and receiving QPs. If they're equal, packets aren't
> > loopback-ed.
> >
> > Signed-off-by: Maor Gottlieb <maorg-VPRAkNaXOzVWk0Htik3J/w@public.gmane.org>
> > Signed-off-by: Eran Ben Elisha <eranbe-VPRAkNaXOzVWk0Htik3J/w@public.gmane.org>
> > ---
> >  drivers/net/ethernet/mellanox/mlx4/en_main.c      | 22 ++++++++++++++++++++
> >  drivers/net/ethernet/mellanox/mlx4/en_resources.c | 25 +++++++++++++++++++++++
> >  drivers/net/ethernet/mellanox/mlx4/mlx4_en.h      |  3 ++-
> >  3 files changed, 49 insertions(+), 1 deletion(-)
> >
> > diff --git a/drivers/net/ethernet/mellanox/mlx4/en_main.c b/drivers/net/ethernet/mellanox/mlx4/en_main.c
> > index a946e4b..70e381a 100644
> > --- a/drivers/net/ethernet/mellanox/mlx4/en_main.c
> > +++ b/drivers/net/ethernet/mellanox/mlx4/en_main.c
> > @@ -123,6 +123,28 @@ void mlx4_en_update_loopback_state(struct net_device *dev,
> >        */
> >       if (mlx4_is_mfunc(priv->mdev->dev) || priv->validate_loopback)
> >               priv->flags |= MLX4_EN_FLAG_ENABLE_HW_LOOPBACK;
> > +
> > +     mutex_lock(&priv->mdev->state_lock);
> > +     if (priv->mdev->dev->caps.flags2 &
> > +         MLX4_DEV_CAP_FLAG2_UPDATE_QP_SRC_CHECK_LB &&
> > +         priv->rss_map.indir_qp.qpn) {
> > +             int i;
> > +             int err = 0;
> > +
> > +             for (i = 0; i < priv->rx_ring_num; i++) {
> > +                     int ret;
> > +
> > +                     ret = mlx4_en_change_mcast_lb(priv,
> > +                                                   &priv->rss_map.qps[i],
> > +                                                   !!(features &
> > +                                                      NETIF_F_LOOPBACK));
> If "feature" is not changing inside the loop can we take it out just for
> the code to look more readable?

Thanks, I will change it in V1.

>
> > +                     if (!err)
> > +                             err = ret;
> > +             }
> > +             if (err)
> > +                     mlx4_warn(priv->mdev, "failed to change mcast loopback\n");
> > +     }
> > +     mutex_unlock(&priv->mdev->state_lock);
> >  }
> >
> >  static int mlx4_en_get_profile(struct mlx4_en_dev *mdev)
> > diff --git a/drivers/net/ethernet/mellanox/mlx4/en_resources.c b/drivers/net/ethernet/mellanox/mlx4/en_resources.c
> > index e482fa1b..12aab5a 100644
> > --- a/drivers/net/ethernet/mellanox/mlx4/en_resources.c
> > +++ b/drivers/net/ethernet/mellanox/mlx4/en_resources.c
> > @@ -69,6 +69,15 @@ void mlx4_en_fill_qp_context(struct mlx4_en_priv *priv, int size, int stride,
> >       context->pri_path.counter_index = priv->counter_index;
> >       context->cqn_send = cpu_to_be32(cqn);
> >       context->cqn_recv = cpu_to_be32(cqn);
> > +     if (!rss &&
> > +         (mdev->dev->caps.flags2 & MLX4_DEV_CAP_FLAG2_LB_SRC_CHK) &&
> > +         context->pri_path.counter_index !=
> > +                         MLX4_SINK_COUNTER_INDEX(mdev->dev)) {
> > +             /* disable multicast loopback to qp with same counter */
> > +             if (!(dev->features & NETIF_F_LOOPBACK))
> > +                     context->pri_path.fl |= MLX4_FL_ETH_SRC_CHECK_MC_LB;
> > +             context->pri_path.control |= MLX4_CTRL_ETH_SRC_CHECK_IF_COUNTER;
> > +     }
> >       context->db_rec_addr = cpu_to_be64(priv->res.db.dma << 2);
> >       if (!(dev->features & NETIF_F_HW_VLAN_CTAG_RX))
> >               context->param3 |= cpu_to_be32(1 << 30);
> > @@ -80,6 +89,22 @@ void mlx4_en_fill_qp_context(struct mlx4_en_priv *priv, int size, int stride,
> >       }
> >  }
> >
> > +int mlx4_en_change_mcast_lb(struct mlx4_en_priv *priv, struct mlx4_qp *qp,
> > +                         int loopback)
> > +{
> > +     int ret;
> > +     struct mlx4_update_qp_params qp_params;
> > +
> > +     memset(&qp_params, 0, sizeof(qp_params));
> > +     if (!loopback)
> > +             qp_params.flags = MLX4_UPDATE_QP_PARAMS_FLAGS_ETH_CHECK_MC_LB;
> > +
> > +     ret = mlx4_update_qp(priv->mdev->dev, qp->qpn,
> > +                          MLX4_UPDATE_QP_ETH_SRC_CHECK_MC_LB,
> > +                          &qp_params);
> > +
> > +     return ret;
> > +}
> >
> >  int mlx4_en_map_buffer(struct mlx4_buf *buf)
> >  {
> > diff --git a/drivers/net/ethernet/mellanox/mlx4/mlx4_en.h b/drivers/net/ethernet/mellanox/mlx4/mlx4_en.h
> > index 666d166..7db86d4 100644
> > --- a/drivers/net/ethernet/mellanox/mlx4/mlx4_en.h
> > +++ b/drivers/net/ethernet/mellanox/mlx4/mlx4_en.h
> > @@ -797,7 +797,8 @@ void mlx4_en_fill_qp_context(struct mlx4_en_priv *priv, int size, int stride,
> >  void mlx4_en_sqp_event(struct mlx4_qp *qp, enum mlx4_event event);
> >  int mlx4_en_map_buffer(struct mlx4_buf *buf);
> >  void mlx4_en_unmap_buffer(struct mlx4_buf *buf);
> > -
> > +int mlx4_en_change_mcast_lb(struct mlx4_en_priv *priv, struct mlx4_qp *qp,
> > +                         int loopback);
> >  void mlx4_en_calc_rx_buf(struct net_device *dev);
> >  int mlx4_en_config_rss_steer(struct mlx4_en_priv *priv);
> >  void mlx4_en_release_rss_steer(struct mlx4_en_priv *priv);
> > --
> > 1.8.3.1
> >
> > --
> > 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
> --
> 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
--
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 for-next 2/7] IB/core: Allow setting create flags in QP init attribute

[Jason Gunthorpe]

On Sun, Aug 16, 2015 at 07:05:28PM +0300, Eran Ben Elisha wrote:

> -	if (attr.create_flags) {
> +	if (attr.create_flags & ~IB_QP_CREATE_BLOCK_MULTICAST_LOOPBACK) {
>  		ret = -EINVAL;
>  		goto err_put;

Shouldn't EINVAL still happen if the driver doesn't support this
feature?

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 for-next 2/7] IB/core: Allow setting create flags in QP init attribute

[eran ben elisha]

On Mon, Aug 17, 2015 at 7:47 PM, Jason Gunthorpe
<jgunthorpe-ePGOBjL8dl3ta4EC/59zMFaTQe2KTcn/@public.gmane.org> wrote:
> On Sun, Aug 16, 2015 at 07:05:28PM +0300, Eran Ben Elisha wrote:
>
>> -     if (attr.create_flags) {
>> +     if (attr.create_flags & ~IB_QP_CREATE_BLOCK_MULTICAST_LOOPBACK) {
>>               ret = -EINVAL;
>>               goto err_put;
>
> Shouldn't EINVAL still happen if the driver doesn't support this
> feature?
>
> Jason

I expect the same behavior as if ib_create_qp was called with
IB_QP_CREATE_BLOCK_MULTICAST_LOOPBACK in the create_flags.
This is not a new flag, and driver should fail the command if not supported.

Eran.

> --
> 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
--
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 for-next 0/7] Add support for multicast loopback prevention to mlx4

[Christoph Lameter]

We ran this through our tests and this works both for Infiniband and
Ethernet.

Tested-by: Christoph Lameter <cl-vYTEC60ixJUAvxtiuMwx3w@public.gmane.org>

--
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