From d21c95c569c462da20d491b75d0a45bd70ddc1bf Mon Sep 17 00:00:00 2001
From: Jonathan Corbet <corbet@lwn.net>
Date: Fri, 16 May 2008 13:40:30 -0600
Subject: Add "no BKL needed" comments to several drivers

This documents the fact that somebody looked at the relevant open()
functions and concluded that, due to their trivial nature, no locking was
needed.

Signed-off-by: Jonathan Corbet <corbet@lwn.net>
---
 drivers/infiniband/hw/ipath/ipath_file_ops.c | 1 +
 1 file changed, 1 insertion(+)

(limited to 'drivers/infiniband/hw')

diff --git a/drivers/infiniband/hw/ipath/ipath_file_ops.c b/drivers/infiniband/hw/ipath/ipath_file_ops.c
index b472b15637f..e80cfbd4f3f 100644
--- a/drivers/infiniband/hw/ipath/ipath_file_ops.c
+++ b/drivers/infiniband/hw/ipath/ipath_file_ops.c
@@ -1812,6 +1812,7 @@ done:
 	return ret;
 }
 
+/* No BKL needed here */
 static int ipath_open(struct inode *in, struct file *fp)
 {
 	/* The real work is performed later in ipath_assign_port() */
-- 
cgit v1.2.3


From f2b9857eee17797541b845782ade4d7a9d50f843 Mon Sep 17 00:00:00 2001
From: Jonathan Corbet <corbet@lwn.net>
Date: Sun, 18 May 2008 15:32:43 -0600
Subject: Add a bunch of cycle_kernel_lock() calls

All of the open() functions which don't need the BKL on their face may
still depend on its acquisition to serialize opens against driver
initialization.  So make those functions acquire then release the BKL to be
on the safe side.

Signed-off-by: Jonathan Corbet <corbet@lwn.net>
---
 drivers/infiniband/hw/ipath/ipath_file_ops.c | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)

(limited to 'drivers/infiniband/hw')

diff --git a/drivers/infiniband/hw/ipath/ipath_file_ops.c b/drivers/infiniband/hw/ipath/ipath_file_ops.c
index e80cfbd4f3f..35f301c88b5 100644
--- a/drivers/infiniband/hw/ipath/ipath_file_ops.c
+++ b/drivers/infiniband/hw/ipath/ipath_file_ops.c
@@ -39,6 +39,7 @@
 #include <linux/highmem.h>
 #include <linux/io.h>
 #include <linux/jiffies.h>
+#include <linux/smp_lock.h>
 #include <asm/pgtable.h>
 
 #include "ipath_kernel.h"
@@ -1812,10 +1813,10 @@ done:
 	return ret;
 }
 
-/* No BKL needed here */
 static int ipath_open(struct inode *in, struct file *fp)
 {
 	/* The real work is performed later in ipath_assign_port() */
+	cycle_kernel_lock();
 	fp->private_data = kzalloc(sizeof(struct ipath_filedata), GFP_KERNEL);
 	return fp->private_data ? 0 : -ENOMEM;
 }
-- 
cgit v1.2.3


From 5e19cf663be534c7c15a35a86fa7ddc9f797e4f4 Mon Sep 17 00:00:00 2001
From: Steve Wise <swise@opengridcomputing.com>
Date: Tue, 8 Jul 2008 14:40:05 -0700
Subject: RDMA/cxgb3: Fix regression caused by class_device -> device
 conversion

The change to iwch_provider.c in commit f4e91eb4 ("IB: convert struct
class_device to struct device") undid the fix done in commit 7f049f2f
("RDMA/cxgb3: Hold rtnl_lock() around ethtool get_drvinfo call").  It
removed the calls to rtnl_lock() that serialized the iw_cxgb3 ethtool
ops calls into the cxgb3 driver.  This locking is needed to avoid
messing up the internal state of the cxgb3 driver.

Signed-off-by: Steve Wise <swise@opengridcomputing.com>
Signed-off-by: Roland Dreier <rolandd@cisco.com>
---
 drivers/infiniband/hw/cxgb3/iwch_provider.c | 4 ++++
 1 file changed, 4 insertions(+)

(limited to 'drivers/infiniband/hw')

diff --git a/drivers/infiniband/hw/cxgb3/iwch_provider.c b/drivers/infiniband/hw/cxgb3/iwch_provider.c
index 8934178a23e..95f82cfb6c5 100644
--- a/drivers/infiniband/hw/cxgb3/iwch_provider.c
+++ b/drivers/infiniband/hw/cxgb3/iwch_provider.c
@@ -1096,7 +1096,9 @@ static ssize_t show_fw_ver(struct device *dev, struct device_attribute *attr, ch
 	struct net_device *lldev = iwch_dev->rdev.t3cdev_p->lldev;
 
 	PDBG("%s dev 0x%p\n", __func__, dev);
+	rtnl_lock();
 	lldev->ethtool_ops->get_drvinfo(lldev, &info);
+	rtnl_unlock();
 	return sprintf(buf, "%s\n", info.fw_version);
 }
 
@@ -1109,7 +1111,9 @@ static ssize_t show_hca(struct device *dev, struct device_attribute *attr,
 	struct net_device *lldev = iwch_dev->rdev.t3cdev_p->lldev;
 
 	PDBG("%s dev 0x%p\n", __func__, dev);
+	rtnl_lock();
 	lldev->ethtool_ops->get_drvinfo(lldev, &info);
+	rtnl_unlock();
 	return sprintf(buf, "%s\n", info.driver);
 }
 
-- 
cgit v1.2.3


From 929555a2baed9b0b050d03532655bfd721a43c44 Mon Sep 17 00:00:00 2001
From: Christophe Jaillet <christophe.jaillet@wanadoo.fr>
Date: Mon, 14 Jul 2008 23:48:43 -0700
Subject: RDMA/nes: Remove unnecessary memset()

Remove an explicit memset(..., 0, ...) of a 'listener' structure
allocated with kzalloc().

Signed-off-by: Christophe Jaillet <christophe.jaillet@wanadoo.fr>
Acked-by: Faisal Latif <faisal@neteffect.com>
Signed-off-by: Roland Dreier <rolandd@cisco.com>
---
 drivers/infiniband/hw/nes/nes_cm.c | 1 -
 1 file changed, 1 deletion(-)

(limited to 'drivers/infiniband/hw')

diff --git a/drivers/infiniband/hw/nes/nes_cm.c b/drivers/infiniband/hw/nes/nes_cm.c
index 9a4b40fae40..6aa531d5276 100644
--- a/drivers/infiniband/hw/nes/nes_cm.c
+++ b/drivers/infiniband/hw/nes/nes_cm.c
@@ -1603,7 +1603,6 @@ static struct nes_cm_listener *mini_cm_listen(struct nes_cm_core *cm_core,
 			return NULL;
 		}
 
-		memset(listener, 0, sizeof(struct nes_cm_listener));
 		listener->loc_addr = htonl(cm_info->loc_addr);
 		listener->loc_port = htons(cm_info->loc_port);
 		listener->reused_node = 0;
-- 
cgit v1.2.3


From 9670e553915e67fb68f13258644342c68dc26b84 Mon Sep 17 00:00:00 2001
From: Eli Cohen <eli@mellanox.co.il>
Date: Mon, 14 Jul 2008 23:48:44 -0700
Subject: IB/mlx4: Optimize QP stamping

The idea is that for QPs with fixed size work requests (eg selective
signaling QPs), before stamping the WQE, we read the value of the DS
field, which gives the effective size of the descriptor as used in the
previous post.  Then we stamp only that area, since the rest of the
descriptor is already stamped.

When initializing the send queue buffer, make sure the DS field is
initialized to the max descriptor size so that the subsequent stamping
will be done on the entire descriptor area.

Signed-off-by: Eli Cohen <eli@mellanox.co.il>
Signed-off-by: Roland Dreier <rolandd@cisco.com>
---
 drivers/infiniband/hw/mlx4/qp.c | 8 ++++++--
 1 file changed, 6 insertions(+), 2 deletions(-)

(limited to 'drivers/infiniband/hw')

diff --git a/drivers/infiniband/hw/mlx4/qp.c b/drivers/infiniband/hw/mlx4/qp.c
index a80df22deae..4b0ac5d68c4 100644
--- a/drivers/infiniband/hw/mlx4/qp.c
+++ b/drivers/infiniband/hw/mlx4/qp.c
@@ -129,9 +129,10 @@ static void stamp_send_wqe(struct mlx4_ib_qp *qp, int n, int size)
 	int ind;
 	void *buf;
 	__be32 stamp;
+	struct mlx4_wqe_ctrl_seg *ctrl;
 
-	s = roundup(size, 1U << qp->sq.wqe_shift);
 	if (qp->sq_max_wqes_per_wr > 1) {
+		s = roundup(size, 1U << qp->sq.wqe_shift);
 		for (i = 0; i < s; i += 64) {
 			ind = (i >> qp->sq.wqe_shift) + n;
 			stamp = ind & qp->sq.wqe_cnt ? cpu_to_be32(0x7fffffff) :
@@ -141,7 +142,8 @@ static void stamp_send_wqe(struct mlx4_ib_qp *qp, int n, int size)
 			*wqe = stamp;
 		}
 	} else {
-		buf = get_send_wqe(qp, n & (qp->sq.wqe_cnt - 1));
+		ctrl = buf = get_send_wqe(qp, n & (qp->sq.wqe_cnt - 1));
+		s = (ctrl->fence_size & 0x3f) << 4;
 		for (i = 64; i < s; i += 64) {
 			wqe = buf + i;
 			*wqe = cpu_to_be32(0xffffffff);
@@ -1063,6 +1065,8 @@ static int __mlx4_ib_modify_qp(struct ib_qp *ibqp,
 		for (i = 0; i < qp->sq.wqe_cnt; ++i) {
 			ctrl = get_send_wqe(qp, i);
 			ctrl->owner_opcode = cpu_to_be32(1 << 31);
+			if (qp->sq_max_wqes_per_wr == 1)
+				ctrl->fence_size = 1 << (qp->sq.wqe_shift - 4);
 
 			stamp_send_wqe(qp, i, 1 << qp->sq.wqe_shift);
 		}
-- 
cgit v1.2.3


From fd91b1bf1bb6fb443cb8c7600c7314f093b31f40 Mon Sep 17 00:00:00 2001
From: "Robert P. J. Day" <rpjday@crashcourse.ca>
Date: Mon, 14 Jul 2008 23:48:44 -0700
Subject: IB/ipath: Simplify code using ARRAY_SIZE() macro

Signed-off-by: Robert P. J. Day <rpjday@crashcourse.ca>
Signed-off-by: Roland Dreier <rolandd@cisco.com>
---
 drivers/infiniband/hw/ipath/ipath_iba7220.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

(limited to 'drivers/infiniband/hw')

diff --git a/drivers/infiniband/hw/ipath/ipath_iba7220.c b/drivers/infiniband/hw/ipath/ipath_iba7220.c
index 8eee7830f04..fb70712ac85 100644
--- a/drivers/infiniband/hw/ipath/ipath_iba7220.c
+++ b/drivers/infiniband/hw/ipath/ipath_iba7220.c
@@ -2228,8 +2228,8 @@ static void ipath_autoneg_send(struct ipath_devdata *dd, int which)
 		0xffffffff, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0,
 		0x40000001, 0x1388, 0x15e, /* rest 0's */
 		};
-	dcnt = sizeof(madpayload_start)/sizeof(madpayload_start[0]);
-	hcnt = sizeof(hdr)/sizeof(hdr[0]);
+	dcnt = ARRAY_SIZE(madpayload_start);
+	hcnt = ARRAY_SIZE(hdr);
 	if (!swapped) {
 		/* for maintainability, do it at runtime */
 		for (i = 0; i < hcnt; i++) {
-- 
cgit v1.2.3


From f3781d2e89f12dd5afa046dc56032af6e39bd116 Mon Sep 17 00:00:00 2001
From: Roland Dreier <rolandd@cisco.com>
Date: Mon, 14 Jul 2008 23:48:44 -0700
Subject: RDMA: Remove subversion $Id tags

They don't get updated by git and so they're worse than useless.

Signed-off-by: Roland Dreier <rolandd@cisco.com>
---
 drivers/infiniband/hw/mthca/mthca_allocator.c  | 2 --
 drivers/infiniband/hw/mthca/mthca_av.c         | 2 --
 drivers/infiniband/hw/mthca/mthca_catas.c      | 2 --
 drivers/infiniband/hw/mthca/mthca_cmd.c        | 2 --
 drivers/infiniband/hw/mthca/mthca_cmd.h        | 2 --
 drivers/infiniband/hw/mthca/mthca_config_reg.h | 2 --
 drivers/infiniband/hw/mthca/mthca_cq.c         | 2 --
 drivers/infiniband/hw/mthca/mthca_dev.h        | 2 --
 drivers/infiniband/hw/mthca/mthca_doorbell.h   | 2 --
 drivers/infiniband/hw/mthca/mthca_eq.c         | 2 --
 drivers/infiniband/hw/mthca/mthca_mad.c        | 2 --
 drivers/infiniband/hw/mthca/mthca_main.c       | 2 --
 drivers/infiniband/hw/mthca/mthca_mcg.c        | 2 --
 drivers/infiniband/hw/mthca/mthca_memfree.c    | 2 --
 drivers/infiniband/hw/mthca/mthca_memfree.h    | 2 --
 drivers/infiniband/hw/mthca/mthca_mr.c         | 2 --
 drivers/infiniband/hw/mthca/mthca_pd.c         | 2 --
 drivers/infiniband/hw/mthca/mthca_profile.c    | 2 --
 drivers/infiniband/hw/mthca/mthca_profile.h    | 2 --
 drivers/infiniband/hw/mthca/mthca_provider.c   | 2 --
 drivers/infiniband/hw/mthca/mthca_provider.h   | 2 --
 drivers/infiniband/hw/mthca/mthca_qp.c         | 2 --
 drivers/infiniband/hw/mthca/mthca_reset.c      | 2 --
 drivers/infiniband/hw/mthca/mthca_srq.c        | 2 --
 drivers/infiniband/hw/mthca/mthca_uar.c        | 2 --
 drivers/infiniband/hw/mthca/mthca_user.h       | 1 -
 drivers/infiniband/hw/mthca/mthca_wqe.h        | 2 --
 27 files changed, 53 deletions(-)

(limited to 'drivers/infiniband/hw')

diff --git a/drivers/infiniband/hw/mthca/mthca_allocator.c b/drivers/infiniband/hw/mthca/mthca_allocator.c
index a7630670961..c5ccc2daab6 100644
--- a/drivers/infiniband/hw/mthca/mthca_allocator.c
+++ b/drivers/infiniband/hw/mthca/mthca_allocator.c
@@ -28,8 +28,6 @@
  * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
  * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
  * SOFTWARE.
- *
- * $Id: mthca_allocator.c 1349 2004-12-16 21:09:43Z roland $
  */
 
 #include <linux/errno.h>
diff --git a/drivers/infiniband/hw/mthca/mthca_av.c b/drivers/infiniband/hw/mthca/mthca_av.c
index 4b111a852ff..32f6c631545 100644
--- a/drivers/infiniband/hw/mthca/mthca_av.c
+++ b/drivers/infiniband/hw/mthca/mthca_av.c
@@ -29,8 +29,6 @@
  * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
  * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
  * SOFTWARE.
- *
- * $Id: mthca_av.c 1349 2004-12-16 21:09:43Z roland $
  */
 
 #include <linux/string.h>
diff --git a/drivers/infiniband/hw/mthca/mthca_catas.c b/drivers/infiniband/hw/mthca/mthca_catas.c
index e948158a28d..40573e48439 100644
--- a/drivers/infiniband/hw/mthca/mthca_catas.c
+++ b/drivers/infiniband/hw/mthca/mthca_catas.c
@@ -28,8 +28,6 @@
  * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
  * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
  * SOFTWARE.
- *
- * $Id$
  */
 
 #include <linux/jiffies.h>
diff --git a/drivers/infiniband/hw/mthca/mthca_cmd.c b/drivers/infiniband/hw/mthca/mthca_cmd.c
index 54d230ee7d6..c33e1c53c79 100644
--- a/drivers/infiniband/hw/mthca/mthca_cmd.c
+++ b/drivers/infiniband/hw/mthca/mthca_cmd.c
@@ -30,8 +30,6 @@
  * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
  * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
  * SOFTWARE.
- *
- * $Id: mthca_cmd.c 1349 2004-12-16 21:09:43Z roland $
  */
 
 #include <linux/completion.h>
diff --git a/drivers/infiniband/hw/mthca/mthca_cmd.h b/drivers/infiniband/hw/mthca/mthca_cmd.h
index 8928ca4a932..6efd3265f24 100644
--- a/drivers/infiniband/hw/mthca/mthca_cmd.h
+++ b/drivers/infiniband/hw/mthca/mthca_cmd.h
@@ -30,8 +30,6 @@
  * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
  * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
  * SOFTWARE.
- *
- * $Id: mthca_cmd.h 1349 2004-12-16 21:09:43Z roland $
  */
 
 #ifndef MTHCA_CMD_H
diff --git a/drivers/infiniband/hw/mthca/mthca_config_reg.h b/drivers/infiniband/hw/mthca/mthca_config_reg.h
index afa56bfaab2..75671f75cac 100644
--- a/drivers/infiniband/hw/mthca/mthca_config_reg.h
+++ b/drivers/infiniband/hw/mthca/mthca_config_reg.h
@@ -29,8 +29,6 @@
  * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
  * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
  * SOFTWARE.
- *
- * $Id: mthca_config_reg.h 1349 2004-12-16 21:09:43Z roland $
  */
 
 #ifndef MTHCA_CONFIG_REG_H
diff --git a/drivers/infiniband/hw/mthca/mthca_cq.c b/drivers/infiniband/hw/mthca/mthca_cq.c
index 20401d2ba6b..f788fce71ac 100644
--- a/drivers/infiniband/hw/mthca/mthca_cq.c
+++ b/drivers/infiniband/hw/mthca/mthca_cq.c
@@ -32,8 +32,6 @@
  * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
  * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
  * SOFTWARE.
- *
- * $Id: mthca_cq.c 1369 2004-12-20 16:17:07Z roland $
  */
 
 #include <linux/hardirq.h>
diff --git a/drivers/infiniband/hw/mthca/mthca_dev.h b/drivers/infiniband/hw/mthca/mthca_dev.h
index 7bc32f8e377..2997d8d564e 100644
--- a/drivers/infiniband/hw/mthca/mthca_dev.h
+++ b/drivers/infiniband/hw/mthca/mthca_dev.h
@@ -32,8 +32,6 @@
  * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
  * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
  * SOFTWARE.
- *
- * $Id: mthca_dev.h 1349 2004-12-16 21:09:43Z roland $
  */
 
 #ifndef MTHCA_DEV_H
diff --git a/drivers/infiniband/hw/mthca/mthca_doorbell.h b/drivers/infiniband/hw/mthca/mthca_doorbell.h
index b374dc395be..14f51ef97d7 100644
--- a/drivers/infiniband/hw/mthca/mthca_doorbell.h
+++ b/drivers/infiniband/hw/mthca/mthca_doorbell.h
@@ -30,8 +30,6 @@
  * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
  * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
  * SOFTWARE.
- *
- * $Id: mthca_doorbell.h 1349 2004-12-16 21:09:43Z roland $
  */
 
 #include <linux/types.h>
diff --git a/drivers/infiniband/hw/mthca/mthca_eq.c b/drivers/infiniband/hw/mthca/mthca_eq.c
index 8bde7f98e58..4e36aa7cb3d 100644
--- a/drivers/infiniband/hw/mthca/mthca_eq.c
+++ b/drivers/infiniband/hw/mthca/mthca_eq.c
@@ -29,8 +29,6 @@
  * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
  * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
  * SOFTWARE.
- *
- * $Id: mthca_eq.c 1382 2004-12-24 02:21:02Z roland $
  */
 
 #include <linux/errno.h>
diff --git a/drivers/infiniband/hw/mthca/mthca_mad.c b/drivers/infiniband/hw/mthca/mthca_mad.c
index 8b7e83e6e88..640449582ab 100644
--- a/drivers/infiniband/hw/mthca/mthca_mad.c
+++ b/drivers/infiniband/hw/mthca/mthca_mad.c
@@ -30,8 +30,6 @@
  * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
  * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
  * SOFTWARE.
- *
- * $Id: mthca_mad.c 1349 2004-12-16 21:09:43Z roland $
  */
 
 #include <linux/string.h>
diff --git a/drivers/infiniband/hw/mthca/mthca_main.c b/drivers/infiniband/hw/mthca/mthca_main.c
index 200cf13fc9b..fb9f91b60f3 100644
--- a/drivers/infiniband/hw/mthca/mthca_main.c
+++ b/drivers/infiniband/hw/mthca/mthca_main.c
@@ -30,8 +30,6 @@
  * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
  * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
  * SOFTWARE.
- *
- * $Id: mthca_main.c 1396 2004-12-28 04:10:27Z roland $
  */
 
 #include <linux/module.h>
diff --git a/drivers/infiniband/hw/mthca/mthca_mcg.c b/drivers/infiniband/hw/mthca/mthca_mcg.c
index a8ad072be07..3f5f9487920 100644
--- a/drivers/infiniband/hw/mthca/mthca_mcg.c
+++ b/drivers/infiniband/hw/mthca/mthca_mcg.c
@@ -28,8 +28,6 @@
  * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
  * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
  * SOFTWARE.
- *
- * $Id: mthca_mcg.c 1349 2004-12-16 21:09:43Z roland $
  */
 
 #include <linux/string.h>
diff --git a/drivers/infiniband/hw/mthca/mthca_memfree.c b/drivers/infiniband/hw/mthca/mthca_memfree.c
index d5862e5d99a..1f7d1a29d2a 100644
--- a/drivers/infiniband/hw/mthca/mthca_memfree.c
+++ b/drivers/infiniband/hw/mthca/mthca_memfree.c
@@ -30,8 +30,6 @@
  * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
  * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
  * SOFTWARE.
- *
- * $Id$
  */
 
 #include <linux/mm.h>
diff --git a/drivers/infiniband/hw/mthca/mthca_memfree.h b/drivers/infiniband/hw/mthca/mthca_memfree.h
index a1ab06847b7..da9b8f9b884 100644
--- a/drivers/infiniband/hw/mthca/mthca_memfree.h
+++ b/drivers/infiniband/hw/mthca/mthca_memfree.h
@@ -30,8 +30,6 @@
  * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
  * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
  * SOFTWARE.
- *
- * $Id$
  */
 
 #ifndef MTHCA_MEMFREE_H
diff --git a/drivers/infiniband/hw/mthca/mthca_mr.c b/drivers/infiniband/hw/mthca/mthca_mr.c
index 820205dec56..8489b1e81c0 100644
--- a/drivers/infiniband/hw/mthca/mthca_mr.c
+++ b/drivers/infiniband/hw/mthca/mthca_mr.c
@@ -29,8 +29,6 @@
  * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
  * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
  * SOFTWARE.
- *
- * $Id: mthca_mr.c 1349 2004-12-16 21:09:43Z roland $
  */
 
 #include <linux/slab.h>
diff --git a/drivers/infiniband/hw/mthca/mthca_pd.c b/drivers/infiniband/hw/mthca/mthca_pd.c
index c1e950764bd..266f14e4740 100644
--- a/drivers/infiniband/hw/mthca/mthca_pd.c
+++ b/drivers/infiniband/hw/mthca/mthca_pd.c
@@ -30,8 +30,6 @@
  * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
  * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
  * SOFTWARE.
- *
- * $Id: mthca_pd.c 1349 2004-12-16 21:09:43Z roland $
  */
 
 #include <linux/errno.h>
diff --git a/drivers/infiniband/hw/mthca/mthca_profile.c b/drivers/infiniband/hw/mthca/mthca_profile.c
index 605a8d57fac..d168c254061 100644
--- a/drivers/infiniband/hw/mthca/mthca_profile.c
+++ b/drivers/infiniband/hw/mthca/mthca_profile.c
@@ -29,8 +29,6 @@
  * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
  * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
  * SOFTWARE.
- *
- * $Id: mthca_profile.c 1349 2004-12-16 21:09:43Z roland $
  */
 
 #include <linux/module.h>
diff --git a/drivers/infiniband/hw/mthca/mthca_profile.h b/drivers/infiniband/hw/mthca/mthca_profile.h
index e76cb62d8e3..62b009cc873 100644
--- a/drivers/infiniband/hw/mthca/mthca_profile.h
+++ b/drivers/infiniband/hw/mthca/mthca_profile.h
@@ -29,8 +29,6 @@
  * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
  * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
  * SOFTWARE.
- *
- * $Id: mthca_profile.h 1349 2004-12-16 21:09:43Z roland $
  */
 
 #ifndef MTHCA_PROFILE_H
diff --git a/drivers/infiniband/hw/mthca/mthca_provider.c b/drivers/infiniband/hw/mthca/mthca_provider.c
index be34f99ca62..87ad889e367 100644
--- a/drivers/infiniband/hw/mthca/mthca_provider.c
+++ b/drivers/infiniband/hw/mthca/mthca_provider.c
@@ -32,8 +32,6 @@
  * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
  * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
  * SOFTWARE.
- *
- * $Id: mthca_provider.c 4859 2006-01-09 21:55:10Z roland $
  */
 
 #include <rdma/ib_smi.h>
diff --git a/drivers/infiniband/hw/mthca/mthca_provider.h b/drivers/infiniband/hw/mthca/mthca_provider.h
index 934bf954403..c621f8794b8 100644
--- a/drivers/infiniband/hw/mthca/mthca_provider.h
+++ b/drivers/infiniband/hw/mthca/mthca_provider.h
@@ -30,8 +30,6 @@
  * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
  * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
  * SOFTWARE.
- *
- * $Id: mthca_provider.h 1349 2004-12-16 21:09:43Z roland $
  */
 
 #ifndef MTHCA_PROVIDER_H
diff --git a/drivers/infiniband/hw/mthca/mthca_qp.c b/drivers/infiniband/hw/mthca/mthca_qp.c
index 09dc3614cf2..3b1c5baf13b 100644
--- a/drivers/infiniband/hw/mthca/mthca_qp.c
+++ b/drivers/infiniband/hw/mthca/mthca_qp.c
@@ -31,8 +31,6 @@
  * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
  * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
  * SOFTWARE.
- *
- * $Id: mthca_qp.c 1355 2004-12-17 15:23:43Z roland $
  */
 
 #include <linux/string.h>
diff --git a/drivers/infiniband/hw/mthca/mthca_reset.c b/drivers/infiniband/hw/mthca/mthca_reset.c
index 91934f2d9db..acb6817f606 100644
--- a/drivers/infiniband/hw/mthca/mthca_reset.c
+++ b/drivers/infiniband/hw/mthca/mthca_reset.c
@@ -28,8 +28,6 @@
  * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
  * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
  * SOFTWARE.
- *
- * $Id: mthca_reset.c 1349 2004-12-16 21:09:43Z roland $
  */
 
 #include <linux/init.h>
diff --git a/drivers/infiniband/hw/mthca/mthca_srq.c b/drivers/infiniband/hw/mthca/mthca_srq.c
index a5ffff6e102..4fabe62aab8 100644
--- a/drivers/infiniband/hw/mthca/mthca_srq.c
+++ b/drivers/infiniband/hw/mthca/mthca_srq.c
@@ -28,8 +28,6 @@
  * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
  * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
  * SOFTWARE.
- *
- * $Id: mthca_srq.c 3047 2005-08-10 03:59:35Z roland $
  */
 
 #include <linux/slab.h>
diff --git a/drivers/infiniband/hw/mthca/mthca_uar.c b/drivers/infiniband/hw/mthca/mthca_uar.c
index 8b728486410..ca5900c96fc 100644
--- a/drivers/infiniband/hw/mthca/mthca_uar.c
+++ b/drivers/infiniband/hw/mthca/mthca_uar.c
@@ -28,8 +28,6 @@
  * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
  * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
  * SOFTWARE.
- *
- * $Id$
  */
 
 #include <asm/page.h>		/* PAGE_SHIFT */
diff --git a/drivers/infiniband/hw/mthca/mthca_user.h b/drivers/infiniband/hw/mthca/mthca_user.h
index e1262c942db..5fe56e81073 100644
--- a/drivers/infiniband/hw/mthca/mthca_user.h
+++ b/drivers/infiniband/hw/mthca/mthca_user.h
@@ -29,7 +29,6 @@
  * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
  * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
  * SOFTWARE.
- *
  */
 
 #ifndef MTHCA_USER_H
diff --git a/drivers/infiniband/hw/mthca/mthca_wqe.h b/drivers/infiniband/hw/mthca/mthca_wqe.h
index b3551a8dea1..341a5ae881c 100644
--- a/drivers/infiniband/hw/mthca/mthca_wqe.h
+++ b/drivers/infiniband/hw/mthca/mthca_wqe.h
@@ -28,8 +28,6 @@
  * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
  * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
  * SOFTWARE.
- *
- * $Id: mthca_wqe.h 3047 2005-08-10 03:59:35Z roland $
  */
 
 #ifndef MTHCA_WQE_H
-- 
cgit v1.2.3


From 00f7ec36c9324928e4cd23f02e6d8550f30c32ca Mon Sep 17 00:00:00 2001
From: Steve Wise <swise@opengridcomputing.com>
Date: Mon, 14 Jul 2008 23:48:45 -0700
Subject: RDMA/core: Add memory management extensions support

This patch adds support for the IB "base memory management extension"
(BMME) and the equivalent iWARP operations (which the iWARP verbs
mandates all devices must implement).  The new operations are:

 - Allocate an ib_mr for use in fast register work requests.

 - Allocate/free a physical buffer lists for use in fast register work
   requests.  This allows device drivers to allocate this memory as
   needed for use in posting send requests (eg via dma_alloc_coherent).

 - New send queue work requests:
   * send with remote invalidate
   * fast register memory region
   * local invalidate memory region
   * RDMA read with invalidate local memory region (iWARP only)

Consumer interface details:

 - A new device capability flag IB_DEVICE_MEM_MGT_EXTENSIONS is added
   to indicate device support for these features.

 - New send work request opcodes IB_WR_FAST_REG_MR, IB_WR_LOCAL_INV,
   IB_WR_RDMA_READ_WITH_INV are added.

 - A new consumer API function, ib_alloc_mr() is added to allocate
   fast register memory regions.

 - New consumer API functions, ib_alloc_fast_reg_page_list() and
   ib_free_fast_reg_page_list() are added to allocate and free
   device-specific memory for fast registration page lists.

 - A new consumer API function, ib_update_fast_reg_key(), is added to
   allow the key portion of the R_Key and L_Key of a fast registration
   MR to be updated.  Consumers call this if desired before posting
   a IB_WR_FAST_REG_MR work request.

Consumers can use this as follows:

 - MR is allocated with ib_alloc_mr().

 - Page list memory is allocated with ib_alloc_fast_reg_page_list().

 - MR R_Key/L_Key "key" field is updated with ib_update_fast_reg_key().

 - MR made VALID and bound to a specific page list via
   ib_post_send(IB_WR_FAST_REG_MR)

 - MR made INVALID via ib_post_send(IB_WR_LOCAL_INV),
   ib_post_send(IB_WR_RDMA_READ_WITH_INV) or an incoming send with
   invalidate operation.

 - MR is deallocated with ib_dereg_mr()

 - page lists dealloced via ib_free_fast_reg_page_list().

Applications can allocate a fast register MR once, and then can
repeatedly bind the MR to different physical block lists (PBLs) via
posting work requests to a send queue (SQ).  For each outstanding
MR-to-PBL binding in the SQ pipe, a fast_reg_page_list needs to be
allocated (the fast_reg_page_list is owned by the low-level driver
from the consumer posting a work request until the request completes).
Thus pipelining can be achieved while still allowing device-specific
page_list processing.

The 32-bit fast register memory key/STag is composed of a 24-bit index
and an 8-bit key.  The application can change the key each time it
fast registers thus allowing more control over the peer's use of the
key/STag (ie it can effectively be changed each time the rkey is
rebound to a page list).

Signed-off-by: Steve Wise <swise@opengridcomputing.com>
Signed-off-by: Roland Dreier <rolandd@cisco.com>
---
 drivers/infiniband/hw/ehca/ehca_reqs.c  |  2 +-
 drivers/infiniband/hw/ipath/ipath_cq.c  |  2 +-
 drivers/infiniband/hw/ipath/ipath_rc.c  |  4 ++--
 drivers/infiniband/hw/ipath/ipath_ruc.c |  4 ++--
 drivers/infiniband/hw/ipath/ipath_uc.c  |  8 ++++----
 drivers/infiniband/hw/ipath/ipath_ud.c  |  8 ++++----
 drivers/infiniband/hw/mlx4/cq.c         | 12 ++++++------
 drivers/infiniband/hw/mthca/mthca_cq.c  |  4 ++--
 8 files changed, 22 insertions(+), 22 deletions(-)

(limited to 'drivers/infiniband/hw')

diff --git a/drivers/infiniband/hw/ehca/ehca_reqs.c b/drivers/infiniband/hw/ehca/ehca_reqs.c
index f093b0033da..b799b271021 100644
--- a/drivers/infiniband/hw/ehca/ehca_reqs.c
+++ b/drivers/infiniband/hw/ehca/ehca_reqs.c
@@ -681,7 +681,7 @@ poll_cq_one_read_cqe:
 	wc->dlid_path_bits = cqe->dlid;
 	wc->src_qp = cqe->remote_qp_number;
 	wc->wc_flags = cqe->w_completion_flags;
-	wc->imm_data = cpu_to_be32(cqe->immediate_data);
+	wc->ex.imm_data = cpu_to_be32(cqe->immediate_data);
 	wc->sl = cqe->service_level;
 
 poll_cq_one_exit0:
diff --git a/drivers/infiniband/hw/ipath/ipath_cq.c b/drivers/infiniband/hw/ipath/ipath_cq.c
index a03bd28d9b4..d385e4168c9 100644
--- a/drivers/infiniband/hw/ipath/ipath_cq.c
+++ b/drivers/infiniband/hw/ipath/ipath_cq.c
@@ -82,7 +82,7 @@ void ipath_cq_enter(struct ipath_cq *cq, struct ib_wc *entry, int solicited)
 		wc->uqueue[head].opcode = entry->opcode;
 		wc->uqueue[head].vendor_err = entry->vendor_err;
 		wc->uqueue[head].byte_len = entry->byte_len;
-		wc->uqueue[head].imm_data = (__u32 __force)entry->imm_data;
+		wc->uqueue[head].ex.imm_data = (__u32 __force) entry->ex.imm_data;
 		wc->uqueue[head].qp_num = entry->qp->qp_num;
 		wc->uqueue[head].src_qp = entry->src_qp;
 		wc->uqueue[head].wc_flags = entry->wc_flags;
diff --git a/drivers/infiniband/hw/ipath/ipath_rc.c b/drivers/infiniband/hw/ipath/ipath_rc.c
index 108df667d2e..97710522624 100644
--- a/drivers/infiniband/hw/ipath/ipath_rc.c
+++ b/drivers/infiniband/hw/ipath/ipath_rc.c
@@ -1703,11 +1703,11 @@ void ipath_rc_rcv(struct ipath_ibdev *dev, struct ipath_ib_header *hdr,
 	case OP(SEND_LAST_WITH_IMMEDIATE):
 	send_last_imm:
 		if (header_in_data) {
-			wc.imm_data = *(__be32 *) data;
+			wc.ex.imm_data = *(__be32 *) data;
 			data += sizeof(__be32);
 		} else {
 			/* Immediate data comes after BTH */
-			wc.imm_data = ohdr->u.imm_data;
+			wc.ex.imm_data = ohdr->u.imm_data;
 		}
 		hdrsize += 4;
 		wc.wc_flags = IB_WC_WITH_IMM;
diff --git a/drivers/infiniband/hw/ipath/ipath_ruc.c b/drivers/infiniband/hw/ipath/ipath_ruc.c
index a4b5521567f..af051f75766 100644
--- a/drivers/infiniband/hw/ipath/ipath_ruc.c
+++ b/drivers/infiniband/hw/ipath/ipath_ruc.c
@@ -331,7 +331,7 @@ again:
 	switch (wqe->wr.opcode) {
 	case IB_WR_SEND_WITH_IMM:
 		wc.wc_flags = IB_WC_WITH_IMM;
-		wc.imm_data = wqe->wr.ex.imm_data;
+		wc.ex.imm_data = wqe->wr.ex.imm_data;
 		/* FALLTHROUGH */
 	case IB_WR_SEND:
 		if (!ipath_get_rwqe(qp, 0))
@@ -342,7 +342,7 @@ again:
 		if (unlikely(!(qp->qp_access_flags & IB_ACCESS_REMOTE_WRITE)))
 			goto inv_err;
 		wc.wc_flags = IB_WC_WITH_IMM;
-		wc.imm_data = wqe->wr.ex.imm_data;
+		wc.ex.imm_data = wqe->wr.ex.imm_data;
 		if (!ipath_get_rwqe(qp, 1))
 			goto rnr_nak;
 		/* FALLTHROUGH */
diff --git a/drivers/infiniband/hw/ipath/ipath_uc.c b/drivers/infiniband/hw/ipath/ipath_uc.c
index 0596ec16fcb..82cc588b8bf 100644
--- a/drivers/infiniband/hw/ipath/ipath_uc.c
+++ b/drivers/infiniband/hw/ipath/ipath_uc.c
@@ -379,11 +379,11 @@ void ipath_uc_rcv(struct ipath_ibdev *dev, struct ipath_ib_header *hdr,
 	case OP(SEND_LAST_WITH_IMMEDIATE):
 	send_last_imm:
 		if (header_in_data) {
-			wc.imm_data = *(__be32 *) data;
+			wc.ex.imm_data = *(__be32 *) data;
 			data += sizeof(__be32);
 		} else {
 			/* Immediate data comes after BTH */
-			wc.imm_data = ohdr->u.imm_data;
+			wc.ex.imm_data = ohdr->u.imm_data;
 		}
 		hdrsize += 4;
 		wc.wc_flags = IB_WC_WITH_IMM;
@@ -483,11 +483,11 @@ void ipath_uc_rcv(struct ipath_ibdev *dev, struct ipath_ib_header *hdr,
 	case OP(RDMA_WRITE_LAST_WITH_IMMEDIATE):
 	rdma_last_imm:
 		if (header_in_data) {
-			wc.imm_data = *(__be32 *) data;
+			wc.ex.imm_data = *(__be32 *) data;
 			data += sizeof(__be32);
 		} else {
 			/* Immediate data comes after BTH */
-			wc.imm_data = ohdr->u.imm_data;
+			wc.ex.imm_data = ohdr->u.imm_data;
 		}
 		hdrsize += 4;
 		wc.wc_flags = IB_WC_WITH_IMM;
diff --git a/drivers/infiniband/hw/ipath/ipath_ud.c b/drivers/infiniband/hw/ipath/ipath_ud.c
index 77ca8ca74e7..36aa242c487 100644
--- a/drivers/infiniband/hw/ipath/ipath_ud.c
+++ b/drivers/infiniband/hw/ipath/ipath_ud.c
@@ -96,7 +96,7 @@ static void ipath_ud_loopback(struct ipath_qp *sqp, struct ipath_swqe *swqe)
 
 	if (swqe->wr.opcode == IB_WR_SEND_WITH_IMM) {
 		wc.wc_flags = IB_WC_WITH_IMM;
-		wc.imm_data = swqe->wr.ex.imm_data;
+		wc.ex.imm_data = swqe->wr.ex.imm_data;
 	}
 
 	/*
@@ -492,14 +492,14 @@ void ipath_ud_rcv(struct ipath_ibdev *dev, struct ipath_ib_header *hdr,
 	if (qp->ibqp.qp_num > 1 &&
 	    opcode == IB_OPCODE_UD_SEND_ONLY_WITH_IMMEDIATE) {
 		if (header_in_data) {
-			wc.imm_data = *(__be32 *) data;
+			wc.ex.imm_data = *(__be32 *) data;
 			data += sizeof(__be32);
 		} else
-			wc.imm_data = ohdr->u.ud.imm_data;
+			wc.ex.imm_data = ohdr->u.ud.imm_data;
 		wc.wc_flags = IB_WC_WITH_IMM;
 		hdrsize += sizeof(u32);
 	} else if (opcode == IB_OPCODE_UD_SEND_ONLY) {
-		wc.imm_data = 0;
+		wc.ex.imm_data = 0;
 		wc.wc_flags = 0;
 	} else {
 		dev->n_pkt_drops++;
diff --git a/drivers/infiniband/hw/mlx4/cq.c b/drivers/infiniband/hw/mlx4/cq.c
index 4521319b140..299f20832ab 100644
--- a/drivers/infiniband/hw/mlx4/cq.c
+++ b/drivers/infiniband/hw/mlx4/cq.c
@@ -663,18 +663,18 @@ repoll:
 
 		switch (cqe->owner_sr_opcode & MLX4_CQE_OPCODE_MASK) {
 		case MLX4_RECV_OPCODE_RDMA_WRITE_IMM:
-			wc->opcode   = IB_WC_RECV_RDMA_WITH_IMM;
-			wc->wc_flags = IB_WC_WITH_IMM;
-			wc->imm_data = cqe->immed_rss_invalid;
+			wc->opcode	= IB_WC_RECV_RDMA_WITH_IMM;
+			wc->wc_flags	= IB_WC_WITH_IMM;
+			wc->ex.imm_data = cqe->immed_rss_invalid;
 			break;
 		case MLX4_RECV_OPCODE_SEND:
 			wc->opcode   = IB_WC_RECV;
 			wc->wc_flags = 0;
 			break;
 		case MLX4_RECV_OPCODE_SEND_IMM:
-			wc->opcode   = IB_WC_RECV;
-			wc->wc_flags = IB_WC_WITH_IMM;
-			wc->imm_data = cqe->immed_rss_invalid;
+			wc->opcode	= IB_WC_RECV;
+			wc->wc_flags	= IB_WC_WITH_IMM;
+			wc->ex.imm_data = cqe->immed_rss_invalid;
 			break;
 		}
 
diff --git a/drivers/infiniband/hw/mthca/mthca_cq.c b/drivers/infiniband/hw/mthca/mthca_cq.c
index f788fce71ac..d9f4735c2b3 100644
--- a/drivers/infiniband/hw/mthca/mthca_cq.c
+++ b/drivers/infiniband/hw/mthca/mthca_cq.c
@@ -620,13 +620,13 @@ static inline int mthca_poll_one(struct mthca_dev *dev,
 		case IB_OPCODE_SEND_LAST_WITH_IMMEDIATE:
 		case IB_OPCODE_SEND_ONLY_WITH_IMMEDIATE:
 			entry->wc_flags = IB_WC_WITH_IMM;
-			entry->imm_data = cqe->imm_etype_pkey_eec;
+			entry->ex.imm_data = cqe->imm_etype_pkey_eec;
 			entry->opcode = IB_WC_RECV;
 			break;
 		case IB_OPCODE_RDMA_WRITE_LAST_WITH_IMMEDIATE:
 		case IB_OPCODE_RDMA_WRITE_ONLY_WITH_IMMEDIATE:
 			entry->wc_flags = IB_WC_WITH_IMM;
-			entry->imm_data = cqe->imm_etype_pkey_eec;
+			entry->ex.imm_data = cqe->imm_etype_pkey_eec;
 			entry->opcode = IB_WC_RECV_RDMA_WITH_IMM;
 			break;
 		default:
-- 
cgit v1.2.3


From e7e55829999deaab3f43e201a087731c02c54cf9 Mon Sep 17 00:00:00 2001
From: Steve Wise <swise@opengridcomputing.com>
Date: Mon, 14 Jul 2008 23:48:45 -0700
Subject: RDMA/cxgb3: MEM_MGT_EXTENSIONS support

- set IB_DEVICE_MEM_MGT_EXTENSIONS capability bit if fw supports it.
- set max_fast_reg_page_list_len device attribute.
- add iwch_alloc_fast_reg_mr function.
- add iwch_alloc_fastreg_pbl
- add iwch_free_fastreg_pbl
- adjust the WQ depth for kernel mode work queues to account for
  fastreg possibly taking 2 WR slots.
- add fastreg_mr work request support.
- add local_inv work request support.
- add send_with_inv and send_with_se_inv work request support.
- removed useless duplicate enums/defines for TPT/MW/MR stuff.

Signed-off-by: Steve Wise <swise@opengridcomputing.com>
Signed-off-by: Roland Dreier <rolandd@cisco.com>
---
 drivers/infiniband/hw/cxgb3/cxio_hal.c      |  15 ++-
 drivers/infiniband/hw/cxgb3/cxio_hal.h      |   1 +
 drivers/infiniband/hw/cxgb3/cxio_wr.h       |  90 ++++++++++++++----
 drivers/infiniband/hw/cxgb3/iwch_cq.c       |  15 ++-
 drivers/infiniband/hw/cxgb3/iwch_provider.c | 104 +++++++++++++++++++-
 drivers/infiniband/hw/cxgb3/iwch_provider.h |   8 --
 drivers/infiniband/hw/cxgb3/iwch_qp.c       | 142 +++++++++++++++++++---------
 7 files changed, 293 insertions(+), 82 deletions(-)

(limited to 'drivers/infiniband/hw')

diff --git a/drivers/infiniband/hw/cxgb3/cxio_hal.c b/drivers/infiniband/hw/cxgb3/cxio_hal.c
index 3f441fc57c1..340e4181c76 100644
--- a/drivers/infiniband/hw/cxgb3/cxio_hal.c
+++ b/drivers/infiniband/hw/cxgb3/cxio_hal.c
@@ -145,7 +145,9 @@ static int cxio_hal_clear_qp_ctx(struct cxio_rdev *rdev_p, u32 qpid)
 	}
 	wqe = (struct t3_modify_qp_wr *) skb_put(skb, sizeof(*wqe));
 	memset(wqe, 0, sizeof(*wqe));
-	build_fw_riwrh((struct fw_riwrh *) wqe, T3_WR_QP_MOD, 3, 0, qpid, 7);
+	build_fw_riwrh((struct fw_riwrh *) wqe, T3_WR_QP_MOD,
+		       T3_COMPLETION_FLAG | T3_NOTIFY_FLAG, 0, qpid, 7,
+		       T3_SOPEOP);
 	wqe->flags = cpu_to_be32(MODQP_WRITE_EC);
 	sge_cmd = qpid << 8 | 3;
 	wqe->sge_cmd = cpu_to_be64(sge_cmd);
@@ -558,7 +560,7 @@ static int cxio_hal_init_ctrl_qp(struct cxio_rdev *rdev_p)
 	wqe = (struct t3_modify_qp_wr *) skb_put(skb, sizeof(*wqe));
 	memset(wqe, 0, sizeof(*wqe));
 	build_fw_riwrh((struct fw_riwrh *) wqe, T3_WR_QP_MOD, 0, 0,
-		       T3_CTL_QP_TID, 7);
+		       T3_CTL_QP_TID, 7, T3_SOPEOP);
 	wqe->flags = cpu_to_be32(MODQP_WRITE_EC);
 	sge_cmd = (3ULL << 56) | FW_RI_SGEEC_START << 8 | 3;
 	wqe->sge_cmd = cpu_to_be64(sge_cmd);
@@ -674,7 +676,7 @@ static int cxio_hal_ctrl_qp_write_mem(struct cxio_rdev *rdev_p, u32 addr,
 		build_fw_riwrh((struct fw_riwrh *) wqe, T3_WR_BP, flag,
 			       Q_GENBIT(rdev_p->ctrl_qp.wptr,
 					T3_CTRL_QP_SIZE_LOG2), T3_CTRL_QP_ID,
-			       wr_len);
+			       wr_len, T3_SOPEOP);
 		if (flag == T3_COMPLETION_FLAG)
 			ring_doorbell(rdev_p->ctrl_qp.doorbell, T3_CTRL_QP_ID);
 		len -= 96;
@@ -816,6 +818,13 @@ int cxio_deallocate_window(struct cxio_rdev *rdev_p, u32 stag)
 			     0, 0);
 }
 
+int cxio_allocate_stag(struct cxio_rdev *rdev_p, u32 *stag, u32 pdid, u32 pbl_size, u32 pbl_addr)
+{
+	*stag = T3_STAG_UNSET;
+	return __cxio_tpt_op(rdev_p, 0, stag, 0, pdid, TPT_NON_SHARED_MR,
+			     0, 0, 0ULL, 0, 0, pbl_size, pbl_addr);
+}
+
 int cxio_rdma_init(struct cxio_rdev *rdev_p, struct t3_rdma_init_attr *attr)
 {
 	struct t3_rdma_init_wr *wqe;
diff --git a/drivers/infiniband/hw/cxgb3/cxio_hal.h b/drivers/infiniband/hw/cxgb3/cxio_hal.h
index 6e128f6bab0..25a880664e6 100644
--- a/drivers/infiniband/hw/cxgb3/cxio_hal.h
+++ b/drivers/infiniband/hw/cxgb3/cxio_hal.h
@@ -165,6 +165,7 @@ int cxio_reregister_phys_mem(struct cxio_rdev *rdev, u32 * stag, u32 pdid,
 int cxio_dereg_mem(struct cxio_rdev *rdev, u32 stag, u32 pbl_size,
 		   u32 pbl_addr);
 int cxio_allocate_window(struct cxio_rdev *rdev, u32 * stag, u32 pdid);
+int cxio_allocate_stag(struct cxio_rdev *rdev, u32 *stag, u32 pdid, u32 pbl_size, u32 pbl_addr);
 int cxio_deallocate_window(struct cxio_rdev *rdev, u32 stag);
 int cxio_rdma_init(struct cxio_rdev *rdev, struct t3_rdma_init_attr *attr);
 void cxio_register_ev_cb(cxio_hal_ev_callback_func_t ev_cb);
diff --git a/drivers/infiniband/hw/cxgb3/cxio_wr.h b/drivers/infiniband/hw/cxgb3/cxio_wr.h
index f1a25a821a4..de760e9f1cc 100644
--- a/drivers/infiniband/hw/cxgb3/cxio_wr.h
+++ b/drivers/infiniband/hw/cxgb3/cxio_wr.h
@@ -72,7 +72,8 @@ enum t3_wr_opcode {
 	T3_WR_BIND = FW_WROPCODE_RI_BIND_MW,
 	T3_WR_RCV = FW_WROPCODE_RI_RECEIVE,
 	T3_WR_INIT = FW_WROPCODE_RI_RDMA_INIT,
-	T3_WR_QP_MOD = FW_WROPCODE_RI_MODIFY_QP
+	T3_WR_QP_MOD = FW_WROPCODE_RI_MODIFY_QP,
+	T3_WR_FASTREG = FW_WROPCODE_RI_FASTREGISTER_MR
 } __attribute__ ((packed));
 
 enum t3_rdma_opcode {
@@ -89,7 +90,8 @@ enum t3_rdma_opcode {
 	T3_FAST_REGISTER,
 	T3_LOCAL_INV,
 	T3_QP_MOD,
-	T3_BYPASS
+	T3_BYPASS,
+	T3_RDMA_READ_REQ_WITH_INV,
 } __attribute__ ((packed));
 
 static inline enum t3_rdma_opcode wr2opcode(enum t3_wr_opcode wrop)
@@ -103,6 +105,7 @@ static inline enum t3_rdma_opcode wr2opcode(enum t3_wr_opcode wrop)
 		case T3_WR_BIND: return T3_BIND_MW;
 		case T3_WR_INIT: return T3_RDMA_INIT;
 		case T3_WR_QP_MOD: return T3_QP_MOD;
+		case T3_WR_FASTREG: return T3_FAST_REGISTER;
 		default: break;
 	}
 	return -1;
@@ -170,11 +173,54 @@ struct t3_send_wr {
 	struct t3_sge sgl[T3_MAX_SGE];	/* 4+ */
 };
 
+#define T3_MAX_FASTREG_DEPTH 24
+#define T3_MAX_FASTREG_FRAG 10
+
+struct t3_fastreg_wr {
+	struct fw_riwrh wrh;	/* 0 */
+	union t3_wrid wrid;	/* 1 */
+	__be32 stag;		/* 2 */
+	__be32 len;
+	__be32 va_base_hi;	/* 3 */
+	__be32 va_base_lo_fbo;
+	__be32 page_type_perms; /* 4 */
+	__be32 reserved1;
+	__be64 pbl_addrs[0];	/* 5+ */
+};
+
+/*
+ * If a fastreg wr spans multiple wqes, then the 2nd fragment look like this.
+ */
+struct t3_pbl_frag {
+	struct fw_riwrh wrh;	/* 0 */
+	__be64 pbl_addrs[14];	/* 1..14 */
+};
+
+#define S_FR_PAGE_COUNT		24
+#define M_FR_PAGE_COUNT		0xff
+#define V_FR_PAGE_COUNT(x)	((x) << S_FR_PAGE_COUNT)
+#define G_FR_PAGE_COUNT(x)	((((x) >> S_FR_PAGE_COUNT)) & M_FR_PAGE_COUNT)
+
+#define S_FR_PAGE_SIZE		16
+#define M_FR_PAGE_SIZE		0x1f
+#define V_FR_PAGE_SIZE(x)	((x) << S_FR_PAGE_SIZE)
+#define G_FR_PAGE_SIZE(x)	((((x) >> S_FR_PAGE_SIZE)) & M_FR_PAGE_SIZE)
+
+#define S_FR_TYPE		8
+#define M_FR_TYPE		0x1
+#define V_FR_TYPE(x)		((x) << S_FR_TYPE)
+#define G_FR_TYPE(x)		((((x) >> S_FR_TYPE)) & M_FR_TYPE)
+
+#define S_FR_PERMS		0
+#define M_FR_PERMS		0xff
+#define V_FR_PERMS(x)		((x) << S_FR_PERMS)
+#define G_FR_PERMS(x)		((((x) >> S_FR_PERMS)) & M_FR_PERMS)
+
 struct t3_local_inv_wr {
 	struct fw_riwrh wrh;	/* 0 */
 	union t3_wrid wrid;	/* 1 */
 	__be32 stag;		/* 2 */
-	__be32 reserved3;
+	__be32 reserved;
 };
 
 struct t3_rdma_write_wr {
@@ -193,7 +239,8 @@ struct t3_rdma_read_wr {
 	struct fw_riwrh wrh;	/* 0 */
 	union t3_wrid wrid;	/* 1 */
 	u8 rdmaop;		/* 2 */
-	u8 reserved[3];
+	u8 local_inv;
+	u8 reserved[2];
 	__be32 rem_stag;
 	__be64 rem_to;		/* 3 */
 	__be32 local_stag;	/* 4 */
@@ -201,18 +248,6 @@ struct t3_rdma_read_wr {
 	__be64 local_to;	/* 5 */
 };
 
-enum t3_addr_type {
-	T3_VA_BASED_TO = 0x0,
-	T3_ZERO_BASED_TO = 0x1
-} __attribute__ ((packed));
-
-enum t3_mem_perms {
-	T3_MEM_ACCESS_LOCAL_READ = 0x1,
-	T3_MEM_ACCESS_LOCAL_WRITE = 0x2,
-	T3_MEM_ACCESS_REM_READ = 0x4,
-	T3_MEM_ACCESS_REM_WRITE = 0x8
-} __attribute__ ((packed));
-
 struct t3_bind_mw_wr {
 	struct fw_riwrh wrh;	/* 0 */
 	union t3_wrid wrid;	/* 1 */
@@ -336,6 +371,11 @@ struct t3_genbit {
 	__be64 genbit;
 };
 
+struct t3_wq_in_err {
+	u64 flit[13];
+	u64 err;
+};
+
 enum rdma_init_wr_flags {
 	MPA_INITIATOR = (1<<0),
 	PRIV_QP = (1<<1),
@@ -346,13 +386,16 @@ union t3_wr {
 	struct t3_rdma_write_wr write;
 	struct t3_rdma_read_wr read;
 	struct t3_receive_wr recv;
+	struct t3_fastreg_wr fastreg;
+	struct t3_pbl_frag pbl_frag;
 	struct t3_local_inv_wr local_inv;
 	struct t3_bind_mw_wr bind;
 	struct t3_bypass_wr bypass;
 	struct t3_rdma_init_wr init;
 	struct t3_modify_qp_wr qp_mod;
 	struct t3_genbit genbit;
-	u64 flit[16];
+	struct t3_wq_in_err wq_in_err;
+	__be64 flit[16];
 };
 
 #define T3_SQ_CQE_FLIT	  13
@@ -366,12 +409,18 @@ static inline enum t3_wr_opcode fw_riwrh_opcode(struct fw_riwrh *wqe)
 	return G_FW_RIWR_OP(be32_to_cpu(wqe->op_seop_flags));
 }
 
+enum t3_wr_hdr_bits {
+	T3_EOP = 1,
+	T3_SOP = 2,
+	T3_SOPEOP = T3_EOP|T3_SOP,
+};
+
 static inline void build_fw_riwrh(struct fw_riwrh *wqe, enum t3_wr_opcode op,
 				  enum t3_wr_flags flags, u8 genbit, u32 tid,
-				  u8 len)
+				  u8 len, u8 sopeop)
 {
 	wqe->op_seop_flags = cpu_to_be32(V_FW_RIWR_OP(op) |
-					 V_FW_RIWR_SOPEOP(M_FW_RIWR_SOPEOP) |
+					 V_FW_RIWR_SOPEOP(sopeop) |
 					 V_FW_RIWR_FLAGS(flags));
 	wmb();
 	wqe->gen_tid_len = cpu_to_be32(V_FW_RIWR_GEN(genbit) |
@@ -404,6 +453,7 @@ enum tpt_addr_type {
 };
 
 enum tpt_mem_perm {
+	TPT_MW_BIND = 0x10,
 	TPT_LOCAL_READ = 0x8,
 	TPT_LOCAL_WRITE = 0x4,
 	TPT_REMOTE_READ = 0x2,
@@ -659,7 +709,7 @@ struct t3_cq {
 
 static inline void cxio_set_wq_in_error(struct t3_wq *wq)
 {
-	wq->queue->flit[13] = 1;
+	wq->queue->wq_in_err.err = 1;
 }
 
 static inline struct t3_cqe *cxio_next_hw_cqe(struct t3_cq *cq)
diff --git a/drivers/infiniband/hw/cxgb3/iwch_cq.c b/drivers/infiniband/hw/cxgb3/iwch_cq.c
index 4ee8ccd0a9e..cf5474ae68f 100644
--- a/drivers/infiniband/hw/cxgb3/iwch_cq.c
+++ b/drivers/infiniband/hw/cxgb3/iwch_cq.c
@@ -81,6 +81,7 @@ static int iwch_poll_cq_one(struct iwch_dev *rhp, struct iwch_cq *chp,
 	wc->wr_id = cookie;
 	wc->qp = &qhp->ibqp;
 	wc->vendor_err = CQE_STATUS(cqe);
+	wc->wc_flags = 0;
 
 	PDBG("%s qpid 0x%x type %d opcode %d status 0x%x wrid hi 0x%x "
 	     "lo 0x%x cookie 0x%llx\n", __func__,
@@ -94,6 +95,11 @@ static int iwch_poll_cq_one(struct iwch_dev *rhp, struct iwch_cq *chp,
 		else
 			wc->byte_len = 0;
 		wc->opcode = IB_WC_RECV;
+		if (CQE_OPCODE(cqe) == T3_SEND_WITH_INV ||
+		    CQE_OPCODE(cqe) == T3_SEND_WITH_SE_INV) {
+			wc->ex.invalidate_rkey = CQE_WRID_STAG(cqe);
+			wc->wc_flags |= IB_WC_WITH_INVALIDATE;
+		}
 	} else {
 		switch (CQE_OPCODE(cqe)) {
 		case T3_RDMA_WRITE:
@@ -105,17 +111,20 @@ static int iwch_poll_cq_one(struct iwch_dev *rhp, struct iwch_cq *chp,
 			break;
 		case T3_SEND:
 		case T3_SEND_WITH_SE:
+		case T3_SEND_WITH_INV:
+		case T3_SEND_WITH_SE_INV:
 			wc->opcode = IB_WC_SEND;
 			break;
 		case T3_BIND_MW:
 			wc->opcode = IB_WC_BIND_MW;
 			break;
 
-		/* these aren't supported yet */
-		case T3_SEND_WITH_INV:
-		case T3_SEND_WITH_SE_INV:
 		case T3_LOCAL_INV:
+			wc->opcode = IB_WC_LOCAL_INV;
+			break;
 		case T3_FAST_REGISTER:
+			wc->opcode = IB_WC_FAST_REG_MR;
+			break;
 		default:
 			printk(KERN_ERR MOD "Unexpected opcode %d "
 			       "in the CQE received for QPID=0x%0x\n",
diff --git a/drivers/infiniband/hw/cxgb3/iwch_provider.c b/drivers/infiniband/hw/cxgb3/iwch_provider.c
index 95f82cfb6c5..5d504f3ed68 100644
--- a/drivers/infiniband/hw/cxgb3/iwch_provider.c
+++ b/drivers/infiniband/hw/cxgb3/iwch_provider.c
@@ -768,6 +768,68 @@ static int iwch_dealloc_mw(struct ib_mw *mw)
 	return 0;
 }
 
+static struct ib_mr *iwch_alloc_fast_reg_mr(struct ib_pd *pd, int pbl_depth)
+{
+	struct iwch_dev *rhp;
+	struct iwch_pd *php;
+	struct iwch_mr *mhp;
+	u32 mmid;
+	u32 stag = 0;
+	int ret;
+
+	php = to_iwch_pd(pd);
+	rhp = php->rhp;
+	mhp = kzalloc(sizeof(*mhp), GFP_KERNEL);
+	if (!mhp)
+		return ERR_PTR(-ENOMEM);
+
+	mhp->rhp = rhp;
+	ret = iwch_alloc_pbl(mhp, pbl_depth);
+	if (ret) {
+		kfree(mhp);
+		return ERR_PTR(ret);
+	}
+	mhp->attr.pbl_size = pbl_depth;
+	ret = cxio_allocate_stag(&rhp->rdev, &stag, php->pdid,
+				 mhp->attr.pbl_size, mhp->attr.pbl_addr);
+	if (ret) {
+		iwch_free_pbl(mhp);
+		kfree(mhp);
+		return ERR_PTR(ret);
+	}
+	mhp->attr.pdid = php->pdid;
+	mhp->attr.type = TPT_NON_SHARED_MR;
+	mhp->attr.stag = stag;
+	mhp->attr.state = 1;
+	mmid = (stag) >> 8;
+	mhp->ibmr.rkey = mhp->ibmr.lkey = stag;
+	insert_handle(rhp, &rhp->mmidr, mhp, mmid);
+	PDBG("%s mmid 0x%x mhp %p stag 0x%x\n", __func__, mmid, mhp, stag);
+	return &(mhp->ibmr);
+}
+
+static struct ib_fast_reg_page_list *iwch_alloc_fastreg_pbl(
+					struct ib_device *device,
+					int page_list_len)
+{
+	struct ib_fast_reg_page_list *page_list;
+
+	page_list = kmalloc(sizeof *page_list + page_list_len * sizeof(u64),
+			    GFP_KERNEL);
+	if (!page_list)
+		return ERR_PTR(-ENOMEM);
+
+	page_list->page_list = (u64 *)(page_list + 1);
+	page_list->max_page_list_len = page_list_len;
+
+	return page_list;
+}
+
+static void iwch_free_fastreg_pbl(struct ib_fast_reg_page_list *page_list)
+{
+	kfree(page_list);
+}
+
 static int iwch_destroy_qp(struct ib_qp *ib_qp)
 {
 	struct iwch_dev *rhp;
@@ -843,6 +905,15 @@ static struct ib_qp *iwch_create_qp(struct ib_pd *pd,
 	 */
 	sqsize = roundup_pow_of_two(attrs->cap.max_send_wr);
 	wqsize = roundup_pow_of_two(rqsize + sqsize);
+
+	/*
+	 * Kernel users need more wq space for fastreg WRs which can take
+	 * 2 WR fragments.
+	 */
+	ucontext = pd->uobject ? to_iwch_ucontext(pd->uobject->context) : NULL;
+	if (!ucontext && wqsize < (rqsize + (2 * sqsize)))
+		wqsize = roundup_pow_of_two(rqsize +
+				roundup_pow_of_two(attrs->cap.max_send_wr * 2));
 	PDBG("%s wqsize %d sqsize %d rqsize %d\n", __func__,
 	     wqsize, sqsize, rqsize);
 	qhp = kzalloc(sizeof(*qhp), GFP_KERNEL);
@@ -851,7 +922,6 @@ static struct ib_qp *iwch_create_qp(struct ib_pd *pd,
 	qhp->wq.size_log2 = ilog2(wqsize);
 	qhp->wq.rq_size_log2 = ilog2(rqsize);
 	qhp->wq.sq_size_log2 = ilog2(sqsize);
-	ucontext = pd->uobject ? to_iwch_ucontext(pd->uobject->context) : NULL;
 	if (cxio_create_qp(&rhp->rdev, !udata, &qhp->wq,
 			   ucontext ? &ucontext->uctx : &rhp->rdev.uctx)) {
 		kfree(qhp);
@@ -1048,6 +1118,7 @@ static int iwch_query_device(struct ib_device *ibdev,
 	props->max_mr = dev->attr.max_mem_regs;
 	props->max_pd = dev->attr.max_pds;
 	props->local_ca_ack_delay = 0;
+	props->max_fast_reg_page_list_len = T3_MAX_FASTREG_DEPTH;
 
 	return 0;
 }
@@ -1088,6 +1159,28 @@ static ssize_t show_rev(struct device *dev, struct device_attribute *attr,
 	return sprintf(buf, "%d\n", iwch_dev->rdev.t3cdev_p->type);
 }
 
+static int fw_supports_fastreg(struct iwch_dev *iwch_dev)
+{
+	struct ethtool_drvinfo info;
+	struct net_device *lldev = iwch_dev->rdev.t3cdev_p->lldev;
+	char *cp, *next;
+	unsigned fw_maj, fw_min;
+
+	rtnl_lock();
+	lldev->ethtool_ops->get_drvinfo(lldev, &info);
+	rtnl_unlock();
+
+	next = info.fw_version+1;
+	cp = strsep(&next, ".");
+	sscanf(cp, "%i", &fw_maj);
+	cp = strsep(&next, ".");
+	sscanf(cp, "%i", &fw_min);
+
+	PDBG("%s maj %u min %u\n", __func__, fw_maj, fw_min);
+
+	return fw_maj > 6 || (fw_maj == 6 && fw_min > 0);
+}
+
 static ssize_t show_fw_ver(struct device *dev, struct device_attribute *attr, char *buf)
 {
 	struct iwch_dev *iwch_dev = container_of(dev, struct iwch_dev,
@@ -1149,8 +1242,10 @@ int iwch_register_device(struct iwch_dev *dev)
 	memset(&dev->ibdev.node_guid, 0, sizeof(dev->ibdev.node_guid));
 	memcpy(&dev->ibdev.node_guid, dev->rdev.t3cdev_p->lldev->dev_addr, 6);
 	dev->ibdev.owner = THIS_MODULE;
-	dev->device_cap_flags =
-	    (IB_DEVICE_ZERO_STAG | IB_DEVICE_MEM_WINDOW);
+	dev->device_cap_flags = IB_DEVICE_ZERO_STAG |
+				IB_DEVICE_MEM_WINDOW;
+	if (fw_supports_fastreg(dev))
+		dev->device_cap_flags |= IB_DEVICE_MEM_MGT_EXTENSIONS;
 
 	dev->ibdev.uverbs_cmd_mask =
 	    (1ull << IB_USER_VERBS_CMD_GET_CONTEXT) |
@@ -1202,6 +1297,9 @@ int iwch_register_device(struct iwch_dev *dev)
 	dev->ibdev.alloc_mw = iwch_alloc_mw;
 	dev->ibdev.bind_mw = iwch_bind_mw;
 	dev->ibdev.dealloc_mw = iwch_dealloc_mw;
+	dev->ibdev.alloc_fast_reg_mr = iwch_alloc_fast_reg_mr;
+	dev->ibdev.alloc_fast_reg_page_list = iwch_alloc_fastreg_pbl;
+	dev->ibdev.free_fast_reg_page_list = iwch_free_fastreg_pbl;
 
 	dev->ibdev.attach_mcast = iwch_multicast_attach;
 	dev->ibdev.detach_mcast = iwch_multicast_detach;
diff --git a/drivers/infiniband/hw/cxgb3/iwch_provider.h b/drivers/infiniband/hw/cxgb3/iwch_provider.h
index 836163fc542..f5ceca05c43 100644
--- a/drivers/infiniband/hw/cxgb3/iwch_provider.h
+++ b/drivers/infiniband/hw/cxgb3/iwch_provider.h
@@ -296,14 +296,6 @@ static inline u32 iwch_ib_to_tpt_access(int acc)
 	       TPT_LOCAL_READ;
 }
 
-static inline u32 iwch_ib_to_mwbind_access(int acc)
-{
-	return (acc & IB_ACCESS_REMOTE_WRITE ? T3_MEM_ACCESS_REM_WRITE : 0) |
-	       (acc & IB_ACCESS_REMOTE_READ ? T3_MEM_ACCESS_REM_READ : 0) |
-	       (acc & IB_ACCESS_LOCAL_WRITE ? T3_MEM_ACCESS_LOCAL_WRITE : 0) |
-	       T3_MEM_ACCESS_LOCAL_READ;
-}
-
 enum iwch_mmid_state {
 	IWCH_STAG_STATE_VALID,
 	IWCH_STAG_STATE_INVALID
diff --git a/drivers/infiniband/hw/cxgb3/iwch_qp.c b/drivers/infiniband/hw/cxgb3/iwch_qp.c
index 99261379922..3b44300a303 100644
--- a/drivers/infiniband/hw/cxgb3/iwch_qp.c
+++ b/drivers/infiniband/hw/cxgb3/iwch_qp.c
@@ -44,54 +44,39 @@ static int iwch_build_rdma_send(union t3_wr *wqe, struct ib_send_wr *wr,
 
 	switch (wr->opcode) {
 	case IB_WR_SEND:
-	case IB_WR_SEND_WITH_IMM:
 		if (wr->send_flags & IB_SEND_SOLICITED)
 			wqe->send.rdmaop = T3_SEND_WITH_SE;
 		else
 			wqe->send.rdmaop = T3_SEND;
 		wqe->send.rem_stag = 0;
 		break;
-#if 0				/* Not currently supported */
-	case TYPE_SEND_INVALIDATE:
-	case TYPE_SEND_INVALIDATE_IMMEDIATE:
-		wqe->send.rdmaop = T3_SEND_WITH_INV;
-		wqe->send.rem_stag = cpu_to_be32(wr->wr.rdma.rkey);
-		break;
-	case TYPE_SEND_SE_INVALIDATE:
-		wqe->send.rdmaop = T3_SEND_WITH_SE_INV;
-		wqe->send.rem_stag = cpu_to_be32(wr->wr.rdma.rkey);
+	case IB_WR_SEND_WITH_INV:
+		if (wr->send_flags & IB_SEND_SOLICITED)
+			wqe->send.rdmaop = T3_SEND_WITH_SE_INV;
+		else
+			wqe->send.rdmaop = T3_SEND_WITH_INV;
+		wqe->send.rem_stag = cpu_to_be32(wr->ex.invalidate_rkey);
 		break;
-#endif
 	default:
-		break;
+		return -EINVAL;
 	}
 	if (wr->num_sge > T3_MAX_SGE)
 		return -EINVAL;
 	wqe->send.reserved[0] = 0;
 	wqe->send.reserved[1] = 0;
 	wqe->send.reserved[2] = 0;
-	if (wr->opcode == IB_WR_SEND_WITH_IMM) {
-		plen = 4;
-		wqe->send.sgl[0].stag = wr->ex.imm_data;
-		wqe->send.sgl[0].len = __constant_cpu_to_be32(0);
-		wqe->send.num_sgle = __constant_cpu_to_be32(0);
-		*flit_cnt = 5;
-	} else {
-		plen = 0;
-		for (i = 0; i < wr->num_sge; i++) {
-			if ((plen + wr->sg_list[i].length) < plen) {
-				return -EMSGSIZE;
-			}
-			plen += wr->sg_list[i].length;
-			wqe->send.sgl[i].stag =
-			    cpu_to_be32(wr->sg_list[i].lkey);
-			wqe->send.sgl[i].len =
-			    cpu_to_be32(wr->sg_list[i].length);
-			wqe->send.sgl[i].to = cpu_to_be64(wr->sg_list[i].addr);
-		}
-		wqe->send.num_sgle = cpu_to_be32(wr->num_sge);
-		*flit_cnt = 4 + ((wr->num_sge) << 1);
+	plen = 0;
+	for (i = 0; i < wr->num_sge; i++) {
+		if ((plen + wr->sg_list[i].length) < plen)
+			return -EMSGSIZE;
+
+		plen += wr->sg_list[i].length;
+		wqe->send.sgl[i].stag = cpu_to_be32(wr->sg_list[i].lkey);
+		wqe->send.sgl[i].len = cpu_to_be32(wr->sg_list[i].length);
+		wqe->send.sgl[i].to = cpu_to_be64(wr->sg_list[i].addr);
 	}
+	wqe->send.num_sgle = cpu_to_be32(wr->num_sge);
+	*flit_cnt = 4 + ((wr->num_sge) << 1);
 	wqe->send.plen = cpu_to_be32(plen);
 	return 0;
 }
@@ -143,9 +128,12 @@ static int iwch_build_rdma_read(union t3_wr *wqe, struct ib_send_wr *wr,
 	if (wr->num_sge > 1)
 		return -EINVAL;
 	wqe->read.rdmaop = T3_READ_REQ;
+	if (wr->opcode == IB_WR_RDMA_READ_WITH_INV)
+		wqe->read.local_inv = 1;
+	else
+		wqe->read.local_inv = 0;
 	wqe->read.reserved[0] = 0;
 	wqe->read.reserved[1] = 0;
-	wqe->read.reserved[2] = 0;
 	wqe->read.rem_stag = cpu_to_be32(wr->wr.rdma.rkey);
 	wqe->read.rem_to = cpu_to_be64(wr->wr.rdma.remote_addr);
 	wqe->read.local_stag = cpu_to_be32(wr->sg_list[0].lkey);
@@ -155,6 +143,57 @@ static int iwch_build_rdma_read(union t3_wr *wqe, struct ib_send_wr *wr,
 	return 0;
 }
 
+static int iwch_build_fastreg(union t3_wr *wqe, struct ib_send_wr *wr,
+				u8 *flit_cnt, int *wr_cnt, struct t3_wq *wq)
+{
+	int i;
+	__be64 *p;
+
+	if (wr->wr.fast_reg.page_list_len > T3_MAX_FASTREG_DEPTH)
+		return -EINVAL;
+	*wr_cnt = 1;
+	wqe->fastreg.stag = cpu_to_be32(wr->wr.fast_reg.rkey);
+	wqe->fastreg.len = cpu_to_be32(wr->wr.fast_reg.length);
+	wqe->fastreg.va_base_hi = cpu_to_be32(wr->wr.fast_reg.iova_start >> 32);
+	wqe->fastreg.va_base_lo_fbo =
+				cpu_to_be32(wr->wr.fast_reg.iova_start & 0xffffffff);
+	wqe->fastreg.page_type_perms = cpu_to_be32(
+		V_FR_PAGE_COUNT(wr->wr.fast_reg.page_list_len) |
+		V_FR_PAGE_SIZE(wr->wr.fast_reg.page_shift-12) |
+		V_FR_TYPE(TPT_VATO) |
+		V_FR_PERMS(iwch_ib_to_tpt_access(wr->wr.fast_reg.access_flags)));
+	p = &wqe->fastreg.pbl_addrs[0];
+	for (i = 0; i < wr->wr.fast_reg.page_list_len; i++, p++) {
+
+		/* If we need a 2nd WR, then set it up */
+		if (i == T3_MAX_FASTREG_FRAG) {
+			*wr_cnt = 2;
+			wqe = (union t3_wr *)(wq->queue +
+				Q_PTR2IDX((wq->wptr+1), wq->size_log2));
+			build_fw_riwrh((void *)wqe, T3_WR_FASTREG, 0,
+			       Q_GENBIT(wq->wptr + 1, wq->size_log2),
+			       0, 1 + wr->wr.fast_reg.page_list_len - T3_MAX_FASTREG_FRAG,
+			       T3_EOP);
+
+			p = &wqe->pbl_frag.pbl_addrs[0];
+		}
+		*p = cpu_to_be64((u64)wr->wr.fast_reg.page_list->page_list[i]);
+	}
+	*flit_cnt = 5 + wr->wr.fast_reg.page_list_len;
+	if (*flit_cnt > 15)
+		*flit_cnt = 15;
+	return 0;
+}
+
+static int iwch_build_inv_stag(union t3_wr *wqe, struct ib_send_wr *wr,
+				u8 *flit_cnt)
+{
+	wqe->local_inv.stag = cpu_to_be32(wr->ex.invalidate_rkey);
+	wqe->local_inv.reserved = 0;
+	*flit_cnt = sizeof(struct t3_local_inv_wr) >> 3;
+	return 0;
+}
+
 /*
  * TBD: this is going to be moved to firmware. Missing pdid/qpid check for now.
  */
@@ -238,6 +277,7 @@ int iwch_post_send(struct ib_qp *ibqp, struct ib_send_wr *wr,
 	u32 num_wrs;
 	unsigned long flag;
 	struct t3_swsq *sqp;
+	int wr_cnt = 1;
 
 	qhp = to_iwch_qp(ibqp);
 	spin_lock_irqsave(&qhp->lock, flag);
@@ -262,15 +302,15 @@ int iwch_post_send(struct ib_qp *ibqp, struct ib_send_wr *wr,
 		t3_wr_flags = 0;
 		if (wr->send_flags & IB_SEND_SOLICITED)
 			t3_wr_flags |= T3_SOLICITED_EVENT_FLAG;
-		if (wr->send_flags & IB_SEND_FENCE)
-			t3_wr_flags |= T3_READ_FENCE_FLAG;
 		if (wr->send_flags & IB_SEND_SIGNALED)
 			t3_wr_flags |= T3_COMPLETION_FLAG;
 		sqp = qhp->wq.sq +
 		      Q_PTR2IDX(qhp->wq.sq_wptr, qhp->wq.sq_size_log2);
 		switch (wr->opcode) {
 		case IB_WR_SEND:
-		case IB_WR_SEND_WITH_IMM:
+		case IB_WR_SEND_WITH_INV:
+			if (wr->send_flags & IB_SEND_FENCE)
+				t3_wr_flags |= T3_READ_FENCE_FLAG;
 			t3_wr_opcode = T3_WR_SEND;
 			err = iwch_build_rdma_send(wqe, wr, &t3_wr_flit_cnt);
 			break;
@@ -280,6 +320,7 @@ int iwch_post_send(struct ib_qp *ibqp, struct ib_send_wr *wr,
 			err = iwch_build_rdma_write(wqe, wr, &t3_wr_flit_cnt);
 			break;
 		case IB_WR_RDMA_READ:
+		case IB_WR_RDMA_READ_WITH_INV:
 			t3_wr_opcode = T3_WR_READ;
 			t3_wr_flags = 0; /* T3 reads are always signaled */
 			err = iwch_build_rdma_read(wqe, wr, &t3_wr_flit_cnt);
@@ -289,6 +330,17 @@ int iwch_post_send(struct ib_qp *ibqp, struct ib_send_wr *wr,
 			if (!qhp->wq.oldest_read)
 				qhp->wq.oldest_read = sqp;
 			break;
+		case IB_WR_FAST_REG_MR:
+			t3_wr_opcode = T3_WR_FASTREG;
+			err = iwch_build_fastreg(wqe, wr, &t3_wr_flit_cnt,
+						 &wr_cnt, &qhp->wq);
+			break;
+		case IB_WR_LOCAL_INV:
+			if (wr->send_flags & IB_SEND_FENCE)
+				t3_wr_flags |= T3_LOCAL_FENCE_FLAG;
+			t3_wr_opcode = T3_WR_INV_STAG;
+			err = iwch_build_inv_stag(wqe, wr, &t3_wr_flit_cnt);
+			break;
 		default:
 			PDBG("%s post of type=%d TBD!\n", __func__,
 			     wr->opcode);
@@ -307,14 +359,15 @@ int iwch_post_send(struct ib_qp *ibqp, struct ib_send_wr *wr,
 
 		build_fw_riwrh((void *) wqe, t3_wr_opcode, t3_wr_flags,
 			       Q_GENBIT(qhp->wq.wptr, qhp->wq.size_log2),
-			       0, t3_wr_flit_cnt);
+			       0, t3_wr_flit_cnt,
+			       (wr_cnt == 1) ? T3_SOPEOP : T3_SOP);
 		PDBG("%s cookie 0x%llx wq idx 0x%x swsq idx %ld opcode %d\n",
 		     __func__, (unsigned long long) wr->wr_id, idx,
 		     Q_PTR2IDX(qhp->wq.sq_wptr, qhp->wq.sq_size_log2),
 		     sqp->opcode);
 		wr = wr->next;
 		num_wrs--;
-		++(qhp->wq.wptr);
+		qhp->wq.wptr += wr_cnt;
 		++(qhp->wq.sq_wptr);
 	}
 	spin_unlock_irqrestore(&qhp->lock, flag);
@@ -359,7 +412,7 @@ int iwch_post_receive(struct ib_qp *ibqp, struct ib_recv_wr *wr,
 			wr->wr_id;
 		build_fw_riwrh((void *) wqe, T3_WR_RCV, T3_COMPLETION_FLAG,
 			       Q_GENBIT(qhp->wq.wptr, qhp->wq.size_log2),
-			       0, sizeof(struct t3_receive_wr) >> 3);
+			       0, sizeof(struct t3_receive_wr) >> 3, T3_SOPEOP);
 		PDBG("%s cookie 0x%llx idx 0x%x rq_wptr 0x%x rw_rptr 0x%x "
 		     "wqe %p \n", __func__, (unsigned long long) wr->wr_id,
 		     idx, qhp->wq.rq_wptr, qhp->wq.rq_rptr, wqe);
@@ -419,10 +472,10 @@ int iwch_bind_mw(struct ib_qp *qp,
 	sgl.lkey = mw_bind->mr->lkey;
 	sgl.length = mw_bind->length;
 	wqe->bind.reserved = 0;
-	wqe->bind.type = T3_VA_BASED_TO;
+	wqe->bind.type = TPT_VATO;
 
 	/* TBD: check perms */
-	wqe->bind.perms = iwch_ib_to_mwbind_access(mw_bind->mw_access_flags);
+	wqe->bind.perms = iwch_ib_to_tpt_access(mw_bind->mw_access_flags);
 	wqe->bind.mr_stag = cpu_to_be32(mw_bind->mr->lkey);
 	wqe->bind.mw_stag = cpu_to_be32(mw->rkey);
 	wqe->bind.mw_len = cpu_to_be32(mw_bind->length);
@@ -430,7 +483,7 @@ int iwch_bind_mw(struct ib_qp *qp,
 	err = iwch_sgl2pbl_map(rhp, &sgl, 1, &pbl_addr, &page_size);
 	if (err) {
 		spin_unlock_irqrestore(&qhp->lock, flag);
-	        return err;
+		return err;
 	}
 	wqe->send.wrid.id0.hi = qhp->wq.sq_wptr;
 	sqp = qhp->wq.sq + Q_PTR2IDX(qhp->wq.sq_wptr, qhp->wq.sq_size_log2);
@@ -441,10 +494,9 @@ int iwch_bind_mw(struct ib_qp *qp,
 	sqp->signaled = (mw_bind->send_flags & IB_SEND_SIGNALED);
 	wqe->bind.mr_pbl_addr = cpu_to_be32(pbl_addr);
 	wqe->bind.mr_pagesz = page_size;
-	wqe->flit[T3_SQ_COOKIE_FLIT] = mw_bind->wr_id;
 	build_fw_riwrh((void *)wqe, T3_WR_BIND, t3_wr_flags,
 		       Q_GENBIT(qhp->wq.wptr, qhp->wq.size_log2), 0,
-			        sizeof(struct t3_bind_mw_wr) >> 3);
+		       sizeof(struct t3_bind_mw_wr) >> 3, T3_SOPEOP);
 	++(qhp->wq.wptr);
 	++(qhp->wq.sq_wptr);
 	spin_unlock_irqrestore(&qhp->lock, flag);
-- 
cgit v1.2.3


From d1f2cd895f8733faa9d79d09d825a2ed80002ac7 Mon Sep 17 00:00:00 2001
From: Eli Cohen <eli@dev.mellanox.co.il>
Date: Mon, 14 Jul 2008 23:48:45 -0700
Subject: IB/mlx4: Configure QPs' max message size based on real device
 capability

ConnectX returns the max message size it supports through the
QUERY_DEV_CAP firmware command.  When modifying a QP to RTR, the max
message size for the QP must be specified.  This value must not exceed
the value declared through QUERY_DEV_CAP.  The current code ignores
the max allowed size and unconditionally sets the value to 2^31.  This
patch sets all QPs to the max value allowed as returned from firmware.

Signed-off-by: Eli Cohen <eli@mellanox.co.il>
Signed-off-by: Roland Dreier <rolandd@cisco.com>
---
 drivers/infiniband/hw/mlx4/qp.c | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)

(limited to 'drivers/infiniband/hw')

diff --git a/drivers/infiniband/hw/mlx4/qp.c b/drivers/infiniband/hw/mlx4/qp.c
index 4b0ac5d68c4..76054a72f71 100644
--- a/drivers/infiniband/hw/mlx4/qp.c
+++ b/drivers/infiniband/hw/mlx4/qp.c
@@ -908,7 +908,8 @@ static int __mlx4_ib_modify_qp(struct ib_qp *ibqp,
 			       attr->path_mtu);
 			goto out;
 		}
-		context->mtu_msgmax = (attr->path_mtu << 5) | 31;
+		context->mtu_msgmax = (attr->path_mtu << 5) |
+			ilog2(dev->dev->caps.max_msg_sz);
 	}
 
 	if (qp->rq.wqe_cnt)
-- 
cgit v1.2.3


From 6578cf33989a594bab37af988d45d87812b946b8 Mon Sep 17 00:00:00 2001
From: Eli Cohen <eli@dev.mellanox.co.il>
Date: Mon, 14 Jul 2008 23:48:45 -0700
Subject: IB/mlx4: Pass congestion management class MADs to the HCA

ConnectX HCAs support the IB_MGMT_CLASS_CONG_MGMT management class, so
process MADs of this class through the MAD_IFC firmware command.

Signed-off-by: Eli Cohen <eli@mellanox.co.il>
Signed-off-by: Roland Dreier <rolandd@cisco.com>
---
 drivers/infiniband/hw/mlx4/mad.c | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)

(limited to 'drivers/infiniband/hw')

diff --git a/drivers/infiniband/hw/mlx4/mad.c b/drivers/infiniband/hw/mlx4/mad.c
index 4c1e72fc8f5..cdca3a511e1 100644
--- a/drivers/infiniband/hw/mlx4/mad.c
+++ b/drivers/infiniband/hw/mlx4/mad.c
@@ -255,7 +255,8 @@ int mlx4_ib_process_mad(struct ib_device *ibdev, int mad_flags,	u8 port_num,
 			return IB_MAD_RESULT_SUCCESS;
 	} else if (in_mad->mad_hdr.mgmt_class == IB_MGMT_CLASS_PERF_MGMT ||
 		   in_mad->mad_hdr.mgmt_class == MLX4_IB_VENDOR_CLASS1   ||
-		   in_mad->mad_hdr.mgmt_class == MLX4_IB_VENDOR_CLASS2) {
+		   in_mad->mad_hdr.mgmt_class == MLX4_IB_VENDOR_CLASS2   ||
+		   in_mad->mad_hdr.mgmt_class == IB_MGMT_CLASS_CONG_MGMT) {
 		if (in_mad->mad_hdr.method  != IB_MGMT_METHOD_GET &&
 		    in_mad->mad_hdr.method  != IB_MGMT_METHOD_SET)
 			return IB_MAD_RESULT_SUCCESS;
-- 
cgit v1.2.3


From d3809ad0972297fbc7ef0585049ef465d9d8d79d Mon Sep 17 00:00:00 2001
From: Roland Dreier <rolandd@cisco.com>
Date: Mon, 14 Jul 2008 23:48:46 -0700
Subject: IB/mthca: Remove extra code for RESET->ERR QP state transition

Commit b18aad71 ("IB/mthca: Fix RESET to ERROR transition") added some
extra code to handle a QP state transition from RESET to ERROR.
However, the latest 1.2.1 version of the IB spec has clarified that
this transition is actually not allowed, so we can remove this extra
code again.

Signed-off-by: Roland Dreier <rolandd@cisco.com>
---
 drivers/infiniband/hw/mthca/mthca_qp.c | 26 --------------------------
 1 file changed, 26 deletions(-)

(limited to 'drivers/infiniband/hw')

diff --git a/drivers/infiniband/hw/mthca/mthca_qp.c b/drivers/infiniband/hw/mthca/mthca_qp.c
index 3b1c5baf13b..77b52b3adcc 100644
--- a/drivers/infiniband/hw/mthca/mthca_qp.c
+++ b/drivers/infiniband/hw/mthca/mthca_qp.c
@@ -848,23 +848,6 @@ out:
 	return err;
 }
 
-static const struct ib_qp_attr dummy_init_attr = { .port_num = 1 };
-static const int dummy_init_attr_mask[] = {
-	[IB_QPT_UD]  = (IB_QP_PKEY_INDEX		|
-			IB_QP_PORT			|
-			IB_QP_QKEY),
-	[IB_QPT_UC]  = (IB_QP_PKEY_INDEX		|
-			IB_QP_PORT			|
-			IB_QP_ACCESS_FLAGS),
-	[IB_QPT_RC]  = (IB_QP_PKEY_INDEX		|
-			IB_QP_PORT			|
-			IB_QP_ACCESS_FLAGS),
-	[IB_QPT_SMI] = (IB_QP_PKEY_INDEX		|
-			IB_QP_QKEY),
-	[IB_QPT_GSI] = (IB_QP_PKEY_INDEX		|
-			IB_QP_QKEY),
-};
-
 int mthca_modify_qp(struct ib_qp *ibqp, struct ib_qp_attr *attr, int attr_mask,
 		    struct ib_udata *udata)
 {
@@ -926,15 +909,6 @@ int mthca_modify_qp(struct ib_qp *ibqp, struct ib_qp_attr *attr, int attr_mask,
 		goto out;
 	}
 
-	if (cur_state == IB_QPS_RESET && new_state == IB_QPS_ERR) {
-		err = __mthca_modify_qp(ibqp, &dummy_init_attr,
-					dummy_init_attr_mask[ibqp->qp_type],
-					IB_QPS_RESET, IB_QPS_INIT);
-		if (err)
-			goto out;
-		cur_state = IB_QPS_INIT;
-	}
-
 	err = __mthca_modify_qp(ibqp, attr, attr_mask, cur_state, new_state);
 
 out:
-- 
cgit v1.2.3


From 7c27f358209a8ce7c57b584346d7b611e823f1b1 Mon Sep 17 00:00:00 2001
From: Roland Dreier <rolandd@cisco.com>
Date: Mon, 14 Jul 2008 23:48:46 -0700
Subject: IB/mlx4: Remove extra code for RESET->ERR QP state transition

Commit 65adfa91 ("IB/mlx4: Fix RESET to RESET and RESET to ERROR
transitions") added some extra code to handle a QP state transition
from RESET to ERROR.  However, the latest 1.2.1 version of the IB spec
has clarified that this transition is actually not allowed, so we can
remove this extra code again.

Signed-off-by: Roland Dreier <rolandd@cisco.com>
---
 drivers/infiniband/hw/mlx4/qp.c | 26 --------------------------
 1 file changed, 26 deletions(-)

(limited to 'drivers/infiniband/hw')

diff --git a/drivers/infiniband/hw/mlx4/qp.c b/drivers/infiniband/hw/mlx4/qp.c
index 76054a72f71..44bbd6c2e31 100644
--- a/drivers/infiniband/hw/mlx4/qp.c
+++ b/drivers/infiniband/hw/mlx4/qp.c
@@ -1132,23 +1132,6 @@ out:
 	return err;
 }
 
-static const struct ib_qp_attr mlx4_ib_qp_attr = { .port_num = 1 };
-static const int mlx4_ib_qp_attr_mask_table[IB_QPT_UD + 1] = {
-		[IB_QPT_UD]  = (IB_QP_PKEY_INDEX		|
-				IB_QP_PORT			|
-				IB_QP_QKEY),
-		[IB_QPT_UC]  = (IB_QP_PKEY_INDEX		|
-				IB_QP_PORT			|
-				IB_QP_ACCESS_FLAGS),
-		[IB_QPT_RC]  = (IB_QP_PKEY_INDEX		|
-				IB_QP_PORT			|
-				IB_QP_ACCESS_FLAGS),
-		[IB_QPT_SMI] = (IB_QP_PKEY_INDEX		|
-				IB_QP_QKEY),
-		[IB_QPT_GSI] = (IB_QP_PKEY_INDEX		|
-				IB_QP_QKEY),
-};
-
 int mlx4_ib_modify_qp(struct ib_qp *ibqp, struct ib_qp_attr *attr,
 		      int attr_mask, struct ib_udata *udata)
 {
@@ -1191,15 +1174,6 @@ int mlx4_ib_modify_qp(struct ib_qp *ibqp, struct ib_qp_attr *attr,
 		goto out;
 	}
 
-	if (cur_state == IB_QPS_RESET && new_state == IB_QPS_ERR) {
-		err = __mlx4_ib_modify_qp(ibqp, &mlx4_ib_qp_attr,
-					  mlx4_ib_qp_attr_mask_table[ibqp->qp_type],
-					  IB_QPS_RESET, IB_QPS_INIT);
-		if (err)
-			goto out;
-		cur_state = IB_QPS_INIT;
-	}
-
 	err = __mlx4_ib_modify_qp(ibqp, attr, attr_mask, cur_state, new_state);
 
 out:
-- 
cgit v1.2.3


From 3e255eac561672cbc92844b9f16cae9304c2a783 Mon Sep 17 00:00:00 2001
From: Joachim Fenkes <fenkes@de.ibm.com>
Date: Mon, 14 Jul 2008 23:48:47 -0700
Subject: IB/ehca: Reject receive work requests if QP is in RESET state

Signed-off-by: Joachim Fenkes <fenkes@de.ibm.com>
Signed-off-by: Roland Dreier <rolandd@cisco.com>
---
 drivers/infiniband/hw/ehca/ehca_reqs.c | 12 ++++++++++--
 1 file changed, 10 insertions(+), 2 deletions(-)

(limited to 'drivers/infiniband/hw')

diff --git a/drivers/infiniband/hw/ehca/ehca_reqs.c b/drivers/infiniband/hw/ehca/ehca_reqs.c
index b799b271021..dd9bc68f1c7 100644
--- a/drivers/infiniband/hw/ehca/ehca_reqs.c
+++ b/drivers/infiniband/hw/ehca/ehca_reqs.c
@@ -544,8 +544,16 @@ int ehca_post_recv(struct ib_qp *qp,
 		   struct ib_recv_wr *recv_wr,
 		   struct ib_recv_wr **bad_recv_wr)
 {
-	return internal_post_recv(container_of(qp, struct ehca_qp, ib_qp),
-				  qp->device, recv_wr, bad_recv_wr);
+	struct ehca_qp *my_qp = container_of(qp, struct ehca_qp, ib_qp);
+
+	/* Reject WR if QP is in RESET state */
+	if (unlikely(my_qp->state == IB_QPS_RESET)) {
+		ehca_err(qp->device, "Invalid QP state  qp_state=%d qpn=%x",
+			 my_qp->state, qp->qp_num);
+		return -EINVAL;
+	}
+
+	return internal_post_recv(my_qp, qp->device, recv_wr, bad_recv_wr);
 }
 
 int ehca_post_srq_recv(struct ib_srq *srq,
-- 
cgit v1.2.3


From 6f7bc01a7382641c61ec036d68ff3a9140b48a1c Mon Sep 17 00:00:00 2001
From: Stefan Roscher <ossrosch@linux.vnet.ibm.com>
Date: Mon, 14 Jul 2008 23:48:47 -0700
Subject: IB/ehca: In case of lost interrupts, trigger EOI to reenable
 interrupts

During corner case testing, we noticed that some versions of ehca do
not properly transition to interrupt done in special load situations.
This can be resolved by periodically triggering EOI through H_EOI, if
EQEs are pending.

Signed-off-by: Stefan Roscher <stefan.roscher@de.ibm.com>
Acked-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Signed-off-by: Roland Dreier <rolandd@cisco.com>
---
 drivers/infiniband/hw/ehca/ehca_irq.c |  9 +++++++--
 drivers/infiniband/hw/ehca/hcp_if.c   | 10 ++++++++++
 drivers/infiniband/hw/ehca/hcp_if.h   |  1 +
 3 files changed, 18 insertions(+), 2 deletions(-)

(limited to 'drivers/infiniband/hw')

diff --git a/drivers/infiniband/hw/ehca/ehca_irq.c b/drivers/infiniband/hw/ehca/ehca_irq.c
index ce1ab0571be..0792d930c48 100644
--- a/drivers/infiniband/hw/ehca/ehca_irq.c
+++ b/drivers/infiniband/hw/ehca/ehca_irq.c
@@ -531,7 +531,7 @@ void ehca_process_eq(struct ehca_shca *shca, int is_irq)
 {
 	struct ehca_eq *eq = &shca->eq;
 	struct ehca_eqe_cache_entry *eqe_cache = eq->eqe_cache;
-	u64 eqe_value;
+	u64 eqe_value, ret;
 	unsigned long flags;
 	int eqe_cnt, i;
 	int eq_empty = 0;
@@ -583,8 +583,13 @@ void ehca_process_eq(struct ehca_shca *shca, int is_irq)
 			ehca_dbg(&shca->ib_device,
 				 "No eqe found for irq event");
 		goto unlock_irq_spinlock;
-	} else if (!is_irq)
+	} else if (!is_irq) {
+		ret = hipz_h_eoi(eq->ist);
+		if (ret != H_SUCCESS)
+			ehca_err(&shca->ib_device,
+				 "bad return code EOI -rc = %ld\n", ret);
 		ehca_dbg(&shca->ib_device, "deadman found %x eqe", eqe_cnt);
+	}
 	if (unlikely(eqe_cnt == EHCA_EQE_CACHE_SIZE))
 		ehca_dbg(&shca->ib_device, "too many eqes for one irq event");
 	/* enable irq for new packets */
diff --git a/drivers/infiniband/hw/ehca/hcp_if.c b/drivers/infiniband/hw/ehca/hcp_if.c
index 5245e13c3a3..415d3a465de 100644
--- a/drivers/infiniband/hw/ehca/hcp_if.c
+++ b/drivers/infiniband/hw/ehca/hcp_if.c
@@ -933,3 +933,13 @@ u64 hipz_h_error_data(const struct ipz_adapter_handle adapter_handle,
 				       r_cb,
 				       0, 0, 0, 0);
 }
+
+u64 hipz_h_eoi(int irq)
+{
+	unsigned long xirr;
+
+	iosync();
+	xirr = (0xffULL << 24) | irq;
+
+	return plpar_hcall_norets(H_EOI, xirr);
+}
diff --git a/drivers/infiniband/hw/ehca/hcp_if.h b/drivers/infiniband/hw/ehca/hcp_if.h
index 60ce02b7066..2c3c6e0ea5c 100644
--- a/drivers/infiniband/hw/ehca/hcp_if.h
+++ b/drivers/infiniband/hw/ehca/hcp_if.h
@@ -260,5 +260,6 @@ u64 hipz_h_error_data(const struct ipz_adapter_handle adapter_handle,
 		      const u64 ressource_handle,
 		      void *rblock,
 		      unsigned long *byte_count);
+u64 hipz_h_eoi(int irq);
 
 #endif /* __HCP_IF_H__ */
-- 
cgit v1.2.3


From 97d1cc8055c7b3fbd35bf693775d61102e65d174 Mon Sep 17 00:00:00 2001
From: Steve Wise <swise@opengridcomputing.com>
Date: Mon, 14 Jul 2008 23:48:47 -0700
Subject: RDMA/cxgb3: Fix up some ib_device_attr fields

- set fw_ver
- set hw_ver
- set max_qp_wr to something reasonable
- set max_cqe to something reasonable

Signed-off-by: Steve Wise <swise@opengridcomputing.com>
Signed-off-by: Roland Dreier <rolandd@cisco.com>
---
 drivers/infiniband/hw/cxgb3/cxio_hal.h      |  3 ++-
 drivers/infiniband/hw/cxgb3/iwch.c          |  4 ++--
 drivers/infiniband/hw/cxgb3/iwch_provider.c | 25 +++++++++++++++++++++++++
 3 files changed, 29 insertions(+), 3 deletions(-)

(limited to 'drivers/infiniband/hw')

diff --git a/drivers/infiniband/hw/cxgb3/cxio_hal.h b/drivers/infiniband/hw/cxgb3/cxio_hal.h
index 25a880664e6..a9ff32c3621 100644
--- a/drivers/infiniband/hw/cxgb3/cxio_hal.h
+++ b/drivers/infiniband/hw/cxgb3/cxio_hal.h
@@ -45,13 +45,14 @@
 #define T3_CTRL_QP_SIZE_LOG2  8
 #define T3_CTRL_CQ_ID    0
 
-/* TBD */
 #define T3_MAX_NUM_RI (1<<15)
 #define T3_MAX_NUM_QP (1<<15)
 #define T3_MAX_NUM_CQ (1<<15)
 #define T3_MAX_NUM_PD (1<<15)
 #define T3_MAX_PBL_SIZE 256
 #define T3_MAX_RQ_SIZE 1024
+#define T3_MAX_QP_DEPTH (T3_MAX_RQ_SIZE-1)
+#define T3_MAX_CQ_DEPTH 8192
 #define T3_MAX_NUM_STAG (1<<15)
 #define T3_MAX_MR_SIZE 0x100000000ULL
 
diff --git a/drivers/infiniband/hw/cxgb3/iwch.c b/drivers/infiniband/hw/cxgb3/iwch.c
index 71554eacb13..e09cc1a8199 100644
--- a/drivers/infiniband/hw/cxgb3/iwch.c
+++ b/drivers/infiniband/hw/cxgb3/iwch.c
@@ -74,11 +74,11 @@ static void rnic_init(struct iwch_dev *rnicp)
 	rnicp->attr.vendor_id = 0x168;
 	rnicp->attr.vendor_part_id = 7;
 	rnicp->attr.max_qps = T3_MAX_NUM_QP - 32;
-	rnicp->attr.max_wrs = (1UL << 24) - 1;
+	rnicp->attr.max_wrs = T3_MAX_QP_DEPTH;
 	rnicp->attr.max_sge_per_wr = T3_MAX_SGE;
 	rnicp->attr.max_sge_per_rdma_write_wr = T3_MAX_SGE;
 	rnicp->attr.max_cqs = T3_MAX_NUM_CQ - 1;
-	rnicp->attr.max_cqes_per_cq = (1UL << 24) - 1;
+	rnicp->attr.max_cqes_per_cq = T3_MAX_CQ_DEPTH;
 	rnicp->attr.max_mem_regs = cxio_num_stags(&rnicp->rdev);
 	rnicp->attr.max_phys_buf_entries = T3_MAX_PBL_SIZE;
 	rnicp->attr.max_pds = T3_MAX_NUM_PD - 1;
diff --git a/drivers/infiniband/hw/cxgb3/iwch_provider.c b/drivers/infiniband/hw/cxgb3/iwch_provider.c
index 5d504f3ed68..5f438065739 100644
--- a/drivers/infiniband/hw/cxgb3/iwch_provider.c
+++ b/drivers/infiniband/hw/cxgb3/iwch_provider.c
@@ -1093,6 +1093,29 @@ static int iwch_query_gid(struct ib_device *ibdev, u8 port,
 	return 0;
 }
 
+static u64 fw_vers_string_to_u64(struct iwch_dev *iwch_dev)
+{
+	struct ethtool_drvinfo info;
+	struct net_device *lldev = iwch_dev->rdev.t3cdev_p->lldev;
+	char *cp, *next;
+	unsigned fw_maj, fw_min, fw_mic;
+
+	rtnl_lock();
+	lldev->ethtool_ops->get_drvinfo(lldev, &info);
+	rtnl_unlock();
+
+	next = info.fw_version + 1;
+	cp = strsep(&next, ".");
+	sscanf(cp, "%i", &fw_maj);
+	cp = strsep(&next, ".");
+	sscanf(cp, "%i", &fw_min);
+	cp = strsep(&next, ".");
+	sscanf(cp, "%i", &fw_mic);
+
+	return (((u64)fw_maj & 0xffff) << 32) | ((fw_min & 0xffff) << 16) |
+	       (fw_mic & 0xffff);
+}
+
 static int iwch_query_device(struct ib_device *ibdev,
 			     struct ib_device_attr *props)
 {
@@ -1103,6 +1126,8 @@ static int iwch_query_device(struct ib_device *ibdev,
 	dev = to_iwch_dev(ibdev);
 	memset(props, 0, sizeof *props);
 	memcpy(&props->sys_image_guid, dev->rdev.t3cdev_p->lldev->dev_addr, 6);
+	props->hw_ver = dev->rdev.t3cdev_p->type;
+	props->fw_ver = fw_vers_string_to_u64(dev);
 	props->device_cap_flags = dev->device_cap_flags;
 	props->vendor_id = (u32)dev->rdev.rnic_info.pdev->vendor;
 	props->vendor_part_id = (u32)dev->rdev.rnic_info.pdev->device;
-- 
cgit v1.2.3


From eec8845d29504a12fbd434e192d61aed3d9d74fa Mon Sep 17 00:00:00 2001
From: Roland Dreier <rolandd@cisco.com>
Date: Mon, 14 Jul 2008 23:48:47 -0700
Subject: RDMA/cxgb3: Remove write-only iwch_rnic_attributes fields

The members struct iwch_rnic_attributes.vendor_id and .vendor_part_id
are write-only, so we might as well get rid of them.

Signed-off-by: Roland Dreier <rolandd@cisco.com>
Acked-by: Steve Wise <swise@opengridcomputing.com>
---
 drivers/infiniband/hw/cxgb3/iwch.c | 2 --
 drivers/infiniband/hw/cxgb3/iwch.h | 2 --
 2 files changed, 4 deletions(-)

(limited to 'drivers/infiniband/hw')

diff --git a/drivers/infiniband/hw/cxgb3/iwch.c b/drivers/infiniband/hw/cxgb3/iwch.c
index e09cc1a8199..06086256ca1 100644
--- a/drivers/infiniband/hw/cxgb3/iwch.c
+++ b/drivers/infiniband/hw/cxgb3/iwch.c
@@ -71,8 +71,6 @@ static void rnic_init(struct iwch_dev *rnicp)
 	idr_init(&rnicp->mmidr);
 	spin_lock_init(&rnicp->lock);
 
-	rnicp->attr.vendor_id = 0x168;
-	rnicp->attr.vendor_part_id = 7;
 	rnicp->attr.max_qps = T3_MAX_NUM_QP - 32;
 	rnicp->attr.max_wrs = T3_MAX_QP_DEPTH;
 	rnicp->attr.max_sge_per_wr = T3_MAX_SGE;
diff --git a/drivers/infiniband/hw/cxgb3/iwch.h b/drivers/infiniband/hw/cxgb3/iwch.h
index d2409a505e8..3773453b2cf 100644
--- a/drivers/infiniband/hw/cxgb3/iwch.h
+++ b/drivers/infiniband/hw/cxgb3/iwch.h
@@ -48,8 +48,6 @@ struct iwch_qp;
 struct iwch_mr;
 
 struct iwch_rnic_attributes {
-	u32 vendor_id;
-	u32 vendor_part_id;
 	u32 max_qps;
 	u32 max_wrs;				/* Max for any SQ/RQ */
 	u32 max_sge_per_wr;
-- 
cgit v1.2.3


From 14cc180f7b032f8484c1a3d0533b1129ffe307fd Mon Sep 17 00:00:00 2001
From: Steve Wise <swise@opengridcomputing.com>
Date: Mon, 14 Jul 2008 23:48:48 -0700
Subject: RDMA/cxgb3: Add support for protocol statistics

- Add a new rdma ctl command called RDMA_GET_MIB to the cxgb3 low
  level driver to obtain the protocol mib from the rnic hardware.

- Add new iw_cxgb3 provider method to get the MIB from the low level
  driver.

Signed-off-by: Steve Wise <swise@opengridcomputing.com>
Signed-off-by: Roland Dreier <rolandd@cisco.com>
---
 drivers/infiniband/hw/cxgb3/iwch_provider.c | 62 +++++++++++++++++++++++++++--
 1 file changed, 58 insertions(+), 4 deletions(-)

(limited to 'drivers/infiniband/hw')

diff --git a/drivers/infiniband/hw/cxgb3/iwch_provider.c b/drivers/infiniband/hw/cxgb3/iwch_provider.c
index 5f438065739..18a6609f5e0 100644
--- a/drivers/infiniband/hw/cxgb3/iwch_provider.c
+++ b/drivers/infiniband/hw/cxgb3/iwch_provider.c
@@ -56,6 +56,7 @@
 #include "iwch_provider.h"
 #include "iwch_cm.h"
 #include "iwch_user.h"
+#include "common.h"
 
 static int iwch_modify_port(struct ib_device *ibdev,
 			    u8 port, int port_modify_mask,
@@ -1245,6 +1246,61 @@ static ssize_t show_board(struct device *dev, struct device_attribute *attr,
 		       iwch_dev->rdev.rnic_info.pdev->device);
 }
 
+static int iwch_get_mib(struct ib_device *ibdev,
+			union rdma_protocol_stats *stats)
+{
+	struct iwch_dev *dev;
+	struct tp_mib_stats m;
+	int ret;
+
+	PDBG("%s ibdev %p\n", __func__, ibdev);
+	dev = to_iwch_dev(ibdev);
+	ret = dev->rdev.t3cdev_p->ctl(dev->rdev.t3cdev_p, RDMA_GET_MIB, &m);
+	if (ret)
+		return -ENOSYS;
+
+	memset(stats, 0, sizeof *stats);
+	stats->iw.ipInReceives = ((u64) m.ipInReceive_hi << 32) +
+				m.ipInReceive_lo;
+	stats->iw.ipInHdrErrors = ((u64) m.ipInHdrErrors_hi << 32) +
+				  m.ipInHdrErrors_lo;
+	stats->iw.ipInAddrErrors = ((u64) m.ipInAddrErrors_hi << 32) +
+				   m.ipInAddrErrors_lo;
+	stats->iw.ipInUnknownProtos = ((u64) m.ipInUnknownProtos_hi << 32) +
+				      m.ipInUnknownProtos_lo;
+	stats->iw.ipInDiscards = ((u64) m.ipInDiscards_hi << 32) +
+				 m.ipInDiscards_lo;
+	stats->iw.ipInDelivers = ((u64) m.ipInDelivers_hi << 32) +
+				 m.ipInDelivers_lo;
+	stats->iw.ipOutRequests = ((u64) m.ipOutRequests_hi << 32) +
+				  m.ipOutRequests_lo;
+	stats->iw.ipOutDiscards = ((u64) m.ipOutDiscards_hi << 32) +
+				  m.ipOutDiscards_lo;
+	stats->iw.ipOutNoRoutes = ((u64) m.ipOutNoRoutes_hi << 32) +
+				  m.ipOutNoRoutes_lo;
+	stats->iw.ipReasmTimeout = (u64) m.ipReasmTimeout;
+	stats->iw.ipReasmReqds = (u64) m.ipReasmReqds;
+	stats->iw.ipReasmOKs = (u64) m.ipReasmOKs;
+	stats->iw.ipReasmFails = (u64) m.ipReasmFails;
+	stats->iw.tcpActiveOpens = (u64) m.tcpActiveOpens;
+	stats->iw.tcpPassiveOpens = (u64) m.tcpPassiveOpens;
+	stats->iw.tcpAttemptFails = (u64) m.tcpAttemptFails;
+	stats->iw.tcpEstabResets = (u64) m.tcpEstabResets;
+	stats->iw.tcpOutRsts = (u64) m.tcpOutRsts;
+	stats->iw.tcpCurrEstab = (u64) m.tcpCurrEstab;
+	stats->iw.tcpInSegs = ((u64) m.tcpInSegs_hi << 32) +
+			      m.tcpInSegs_lo;
+	stats->iw.tcpOutSegs = ((u64) m.tcpOutSegs_hi << 32) +
+			       m.tcpOutSegs_lo;
+	stats->iw.tcpRetransSegs = ((u64) m.tcpRetransSeg_hi << 32) +
+				  m.tcpRetransSeg_lo;
+	stats->iw.tcpInErrs = ((u64) m.tcpInErrs_hi << 32) +
+			      m.tcpInErrs_lo;
+	stats->iw.tcpRtoMin = (u64) m.tcpRtoMin;
+	stats->iw.tcpRtoMax = (u64) m.tcpRtoMax;
+	return 0;
+}
+
 static DEVICE_ATTR(hw_rev, S_IRUGO, show_rev, NULL);
 static DEVICE_ATTR(fw_ver, S_IRUGO, show_fw_ver, NULL);
 static DEVICE_ATTR(hca_type, S_IRUGO, show_hca, NULL);
@@ -1254,7 +1310,7 @@ static struct device_attribute *iwch_class_attributes[] = {
 	&dev_attr_hw_rev,
 	&dev_attr_fw_ver,
 	&dev_attr_hca_type,
-	&dev_attr_board_id
+	&dev_attr_board_id,
 };
 
 int iwch_register_device(struct iwch_dev *dev)
@@ -1325,15 +1381,13 @@ int iwch_register_device(struct iwch_dev *dev)
 	dev->ibdev.alloc_fast_reg_mr = iwch_alloc_fast_reg_mr;
 	dev->ibdev.alloc_fast_reg_page_list = iwch_alloc_fastreg_pbl;
 	dev->ibdev.free_fast_reg_page_list = iwch_free_fastreg_pbl;
-
 	dev->ibdev.attach_mcast = iwch_multicast_attach;
 	dev->ibdev.detach_mcast = iwch_multicast_detach;
 	dev->ibdev.process_mad = iwch_process_mad;
-
 	dev->ibdev.req_notify_cq = iwch_arm_cq;
 	dev->ibdev.post_send = iwch_post_send;
 	dev->ibdev.post_recv = iwch_post_receive;
-
+	dev->ibdev.get_protocol_stats = iwch_get_mib;
 
 	dev->ibdev.iwcm = kmalloc(sizeof(struct iw_cm_verbs), GFP_KERNEL);
 	if (!dev->ibdev.iwcm)
-- 
cgit v1.2.3


From 521e575b9a7324a0bca762622139f69582a042bf Mon Sep 17 00:00:00 2001
From: Ron Livne <ronli@voltaire.com>
Date: Mon, 14 Jul 2008 23:48:48 -0700
Subject: IB/mlx4: Add support for blocking multicast loopback packets

Add support for handling the IB_QP_CREATE_MULTICAST_BLOCK_LOOPBACK
flag by using the per-multicast group loopback blocking feature of
mlx4 hardware.

Signed-off-by: Ron Livne <ronli@voltaire.com>
Signed-off-by: Roland Dreier <rolandd@cisco.com>
---
 drivers/infiniband/hw/mlx4/main.c    |  7 +++++--
 drivers/infiniband/hw/mlx4/mlx4_ib.h |  3 ++-
 drivers/infiniband/hw/mlx4/qp.c      | 21 ++++++++++++++++++---
 3 files changed, 25 insertions(+), 6 deletions(-)

(limited to 'drivers/infiniband/hw')

diff --git a/drivers/infiniband/hw/mlx4/main.c b/drivers/infiniband/hw/mlx4/main.c
index 4d61e32866c..bcf50648fa1 100644
--- a/drivers/infiniband/hw/mlx4/main.c
+++ b/drivers/infiniband/hw/mlx4/main.c
@@ -90,7 +90,8 @@ static int mlx4_ib_query_device(struct ib_device *ibdev,
 	props->device_cap_flags    = IB_DEVICE_CHANGE_PHY_PORT |
 		IB_DEVICE_PORT_ACTIVE_EVENT		|
 		IB_DEVICE_SYS_IMAGE_GUID		|
-		IB_DEVICE_RC_RNR_NAK_GEN;
+		IB_DEVICE_RC_RNR_NAK_GEN		|
+		IB_DEVICE_BLOCK_MULTICAST_LOOPBACK;
 	if (dev->dev->caps.flags & MLX4_DEV_CAP_FLAG_BAD_PKEY_CNTR)
 		props->device_cap_flags |= IB_DEVICE_BAD_PKEY_CNTR;
 	if (dev->dev->caps.flags & MLX4_DEV_CAP_FLAG_BAD_QKEY_CNTR)
@@ -437,7 +438,9 @@ static int mlx4_ib_dealloc_pd(struct ib_pd *pd)
 static int mlx4_ib_mcg_attach(struct ib_qp *ibqp, union ib_gid *gid, u16 lid)
 {
 	return mlx4_multicast_attach(to_mdev(ibqp->device)->dev,
-				     &to_mqp(ibqp)->mqp, gid->raw);
+				     &to_mqp(ibqp)->mqp, gid->raw,
+				     !!(to_mqp(ibqp)->flags &
+					MLX4_IB_QP_BLOCK_MULTICAST_LOOPBACK));
 }
 
 static int mlx4_ib_mcg_detach(struct ib_qp *ibqp, union ib_gid *gid, u16 lid)
diff --git a/drivers/infiniband/hw/mlx4/mlx4_ib.h b/drivers/infiniband/hw/mlx4/mlx4_ib.h
index 5cf994794d2..c4cf5b69eef 100644
--- a/drivers/infiniband/hw/mlx4/mlx4_ib.h
+++ b/drivers/infiniband/hw/mlx4/mlx4_ib.h
@@ -101,7 +101,8 @@ struct mlx4_ib_wq {
 };
 
 enum mlx4_ib_qp_flags {
-	MLX4_IB_QP_LSO		= 1 << 0
+	MLX4_IB_QP_LSO				= 1 << 0,
+	MLX4_IB_QP_BLOCK_MULTICAST_LOOPBACK	= 1 << 1,
 };
 
 struct mlx4_ib_qp {
diff --git a/drivers/infiniband/hw/mlx4/qp.c b/drivers/infiniband/hw/mlx4/qp.c
index 44bbd6c2e31..91590e7fba0 100644
--- a/drivers/infiniband/hw/mlx4/qp.c
+++ b/drivers/infiniband/hw/mlx4/qp.c
@@ -511,6 +511,9 @@ 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;
 
@@ -684,10 +687,15 @@ struct ib_qp *mlx4_ib_create_qp(struct ib_pd *pd,
 	struct mlx4_ib_qp *qp;
 	int err;
 
-	/* We only support LSO, and only for kernel UD QPs. */
-	if (init_attr->create_flags & ~IB_QP_CREATE_IPOIB_UD_LSO)
+	/*
+	 * We only support LSO and multicast loopback blocking, and
+	 * only for kernel UD QPs.
+	 */
+	if (init_attr->create_flags & ~(IB_QP_CREATE_IPOIB_UD_LSO |
+					IB_QP_CREATE_BLOCK_MULTICAST_LOOPBACK))
 		return ERR_PTR(-EINVAL);
-	if (init_attr->create_flags & IB_QP_CREATE_IPOIB_UD_LSO &&
+
+	if (init_attr->create_flags &&
 	    (pd->uobject || init_attr->qp_type != IB_QPT_UD))
 		return ERR_PTR(-EINVAL);
 
@@ -1844,6 +1852,13 @@ done:
 
 	qp_init_attr->cap	     = qp_attr->cap;
 
+	qp_init_attr->create_flags = 0;
+	if (qp->flags & MLX4_IB_QP_BLOCK_MULTICAST_LOOPBACK)
+		qp_init_attr->create_flags |= IB_QP_CREATE_BLOCK_MULTICAST_LOOPBACK;
+
+	if (qp->flags & MLX4_IB_QP_LSO)
+		qp_init_attr->create_flags |= IB_QP_CREATE_IPOIB_UD_LSO;
+
 out:
 	mutex_unlock(&qp->mutex);
 	return err;
-- 
cgit v1.2.3


From 038919f29682b00ea95506e959210fc72d1aaf64 Mon Sep 17 00:00:00 2001
From: Joachim Fenkes <fenkes@de.ibm.com>
Date: Mon, 14 Jul 2008 23:48:49 -0700
Subject: IB/ehca: Make device table externally visible

This gives ehca an autogenerated modalias and therefore enables automatic loading.

Signed-off-by: Joachim Fenkes <fenkes@de.ibm.com>
Signed-off-by: Roland Dreier <rolandd@cisco.com>
---
 drivers/infiniband/hw/ehca/ehca_main.c | 1 +
 1 file changed, 1 insertion(+)

(limited to 'drivers/infiniband/hw')

diff --git a/drivers/infiniband/hw/ehca/ehca_main.c b/drivers/infiniband/hw/ehca/ehca_main.c
index 482103eb6ea..598844d2edc 100644
--- a/drivers/infiniband/hw/ehca/ehca_main.c
+++ b/drivers/infiniband/hw/ehca/ehca_main.c
@@ -923,6 +923,7 @@ static struct of_device_id ehca_device_table[] =
 	},
 	{},
 };
+MODULE_DEVICE_TABLE(of, ehca_device_table);
 
 static struct of_platform_driver ehca_driver = {
 	.name        = "ehca",
-- 
cgit v1.2.3


From 1ff66e8c1faee7c2711b84b9c89e1c5fcd767839 Mon Sep 17 00:00:00 2001
From: Roland Dreier <rolandd@cisco.com>
Date: Mon, 14 Jul 2008 23:48:49 -0700
Subject: RDMA/nes: Encapsulate logic nes_put_cqp_request()

The iw_nes driver repeats the logic

	if (atomic_dec_and_test(&cqp_request->refcount)) {
		if (cqp_request->dynamic) {
			kfree(cqp_request);
		} else {
			spin_lock_irqsave(&nesdev->cqp.lock, flags);
			list_add_tail(&cqp_request->list, &nesdev->cqp_avail_reqs);
			spin_unlock_irqrestore(&nesdev->cqp.lock, flags);
		}
	}

over and over.  Wrap this up in functions nes_free_cqp_request() and
nes_put_cqp_request() to simplify such code.

In addition to making the source smaller and more readable, this shrinks
the compiled code quite a bit:

add/remove: 2/0 grow/shrink: 0/13 up/down: 164/-1692 (-1528)
function                                     old     new   delta
nes_free_cqp_request                           -     147    +147
nes_put_cqp_request                            -      17     +17
nes_modify_qp                               2316    2293     -23
nes_hw_modify_qp                             737     657     -80
nes_dereg_mr                                 945     860     -85
flush_wqes                                   501     416     -85
nes_manage_apbvt                             648     560     -88
nes_reg_mr                                  1117    1026     -91
nes_cqp_ce_handler                           927     769    -158
nes_alloc_mw                                1052     884    -168
nes_create_qp                               5314    5141    -173
nes_alloc_fmr                               2212    2035    -177
nes_destroy_cq                              1097     918    -179
nes_create_cq                               2787    2598    -189
nes_dealloc_mw                               762     566    -196

Signed-off-by: Roland Dreier <rolandd@cisco.com>
Acked-by: Faisal Latif <flatif@neteffect.com>
---
 drivers/infiniband/hw/nes/nes.h       |   4 +
 drivers/infiniband/hw/nes/nes_hw.c    |  60 ++---------
 drivers/infiniband/hw/nes/nes_utils.c |  24 +++++
 drivers/infiniband/hw/nes/nes_verbs.c | 189 ++++++----------------------------
 4 files changed, 66 insertions(+), 211 deletions(-)

(limited to 'drivers/infiniband/hw')

diff --git a/drivers/infiniband/hw/nes/nes.h b/drivers/infiniband/hw/nes/nes.h
index 61b46e9c7d2..fe88bec4894 100644
--- a/drivers/infiniband/hw/nes/nes.h
+++ b/drivers/infiniband/hw/nes/nes.h
@@ -538,6 +538,10 @@ void nes_read_1G_phy_reg(struct nes_device *, u8, u8, u16 *);
 void nes_write_10G_phy_reg(struct nes_device *, u16, u8, u16, u16);
 void nes_read_10G_phy_reg(struct nes_device *, u8, u8, u16);
 struct nes_cqp_request *nes_get_cqp_request(struct nes_device *);
+void nes_free_cqp_request(struct nes_device *nesdev,
+			  struct nes_cqp_request *cqp_request);
+void nes_put_cqp_request(struct nes_device *nesdev,
+			 struct nes_cqp_request *cqp_request);
 void nes_post_cqp_request(struct nes_device *, struct nes_cqp_request *, int);
 int nes_arp_table(struct nes_device *, u32, u8 *, u32);
 void nes_mh_fix(unsigned long);
diff --git a/drivers/infiniband/hw/nes/nes_hw.c b/drivers/infiniband/hw/nes/nes_hw.c
index d3278f111ca..80e486653ec 100644
--- a/drivers/infiniband/hw/nes/nes_hw.c
+++ b/drivers/infiniband/hw/nes/nes_hw.c
@@ -2710,39 +2710,11 @@ static void nes_cqp_ce_handler(struct nes_device *nesdev, struct nes_hw_cq *cq)
 					barrier();
 					cqp_request->request_done = 1;
 					wake_up(&cqp_request->waitq);
-					if (atomic_dec_and_test(&cqp_request->refcount)) {
-						nes_debug(NES_DBG_CQP, "CQP request %p (opcode 0x%02X) freed.\n",
-								cqp_request,
-								le32_to_cpu(cqp_request->cqp_wqe.wqe_words[NES_CQP_WQE_OPCODE_IDX])&0x3f);
-						if (cqp_request->dynamic) {
-							kfree(cqp_request);
-						} else {
-							spin_lock_irqsave(&nesdev->cqp.lock, flags);
-							list_add_tail(&cqp_request->list, &nesdev->cqp_avail_reqs);
-							spin_unlock_irqrestore(&nesdev->cqp.lock, flags);
-						}
-					}
-				} else if (cqp_request->callback) {
-					/* Envoke the callback routine */
-					cqp_request->cqp_callback(nesdev, cqp_request);
-					if (cqp_request->dynamic) {
-						kfree(cqp_request);
-					} else {
-						spin_lock_irqsave(&nesdev->cqp.lock, flags);
-						list_add_tail(&cqp_request->list, &nesdev->cqp_avail_reqs);
-						spin_unlock_irqrestore(&nesdev->cqp.lock, flags);
-					}
+					nes_put_cqp_request(nesdev, cqp_request);
 				} else {
-					nes_debug(NES_DBG_CQP, "CQP request %p (opcode 0x%02X) freed.\n",
-							cqp_request,
-							le32_to_cpu(cqp_request->cqp_wqe.wqe_words[NES_CQP_WQE_OPCODE_IDX]) & 0x3f);
-					if (cqp_request->dynamic) {
-						kfree(cqp_request);
-					} else {
-						spin_lock_irqsave(&nesdev->cqp.lock, flags);
-						list_add_tail(&cqp_request->list, &nesdev->cqp_avail_reqs);
-						spin_unlock_irqrestore(&nesdev->cqp.lock, flags);
-					}
+					if (cqp_request->callback)
+						cqp_request->cqp_callback(nesdev, cqp_request);
+					nes_free_cqp_request(nesdev, cqp_request);
 				}
 			} else {
 				wake_up(&nesdev->cqp.waitq);
@@ -3149,7 +3121,6 @@ int nes_manage_apbvt(struct nes_vnic *nesvnic, u32 accel_local_port,
 {
 	struct nes_device *nesdev = nesvnic->nesdev;
 	struct nes_hw_cqp_wqe *cqp_wqe;
-	unsigned long flags;
 	struct nes_cqp_request *cqp_request;
 	int ret = 0;
 	u16 major_code;
@@ -3184,15 +3155,9 @@ int nes_manage_apbvt(struct nes_vnic *nesvnic, u32 accel_local_port,
 	nes_debug(NES_DBG_QP, "Completed, ret=%u,  CQP Major:Minor codes = 0x%04X:0x%04X\n",
 			ret, cqp_request->major_code, cqp_request->minor_code);
 	major_code = cqp_request->major_code;
-	if (atomic_dec_and_test(&cqp_request->refcount)) {
-		if (cqp_request->dynamic) {
-			kfree(cqp_request);
-		} else {
-			spin_lock_irqsave(&nesdev->cqp.lock, flags);
-			list_add_tail(&cqp_request->list, &nesdev->cqp_avail_reqs);
-			spin_unlock_irqrestore(&nesdev->cqp.lock, flags);
-		}
-	}
+
+	nes_put_cqp_request(nesdev, cqp_request);
+
 	if (!ret)
 		return -ETIME;
 	else if (major_code)
@@ -3262,7 +3227,6 @@ void nes_manage_arp_cache(struct net_device *netdev, unsigned char *mac_addr,
 void flush_wqes(struct nes_device *nesdev, struct nes_qp *nesqp,
 		u32 which_wq, u32 wait_completion)
 {
-	unsigned long flags;
 	struct nes_cqp_request *cqp_request;
 	struct nes_hw_cqp_wqe *cqp_wqe;
 	int ret;
@@ -3294,14 +3258,6 @@ void flush_wqes(struct nes_device *nesdev, struct nes_qp *nesqp,
 		nes_debug(NES_DBG_QP, "Flush SQ QP WQEs completed, ret=%u,"
 				" CQP Major:Minor codes = 0x%04X:0x%04X\n",
 				ret, cqp_request->major_code, cqp_request->minor_code);
-		if (atomic_dec_and_test(&cqp_request->refcount)) {
-			if (cqp_request->dynamic) {
-				kfree(cqp_request);
-			} else {
-				spin_lock_irqsave(&nesdev->cqp.lock, flags);
-				list_add_tail(&cqp_request->list, &nesdev->cqp_avail_reqs);
-				spin_unlock_irqrestore(&nesdev->cqp.lock, flags);
-			}
-		}
+		nes_put_cqp_request(nesdev, cqp_request);
 	}
 }
diff --git a/drivers/infiniband/hw/nes/nes_utils.c b/drivers/infiniband/hw/nes/nes_utils.c
index fe83d1b2b17..8f519c86087 100644
--- a/drivers/infiniband/hw/nes/nes_utils.c
+++ b/drivers/infiniband/hw/nes/nes_utils.c
@@ -567,6 +567,30 @@ struct nes_cqp_request *nes_get_cqp_request(struct nes_device *nesdev)
 	return cqp_request;
 }
 
+void nes_free_cqp_request(struct nes_device *nesdev,
+			  struct nes_cqp_request *cqp_request)
+{
+	unsigned long flags;
+
+	nes_debug(NES_DBG_CQP, "CQP request %p (opcode 0x%02X) freed.\n",
+		  cqp_request,
+		  le32_to_cpu(cqp_request->cqp_wqe.wqe_words[NES_CQP_WQE_OPCODE_IDX]) & 0x3f);
+
+	if (cqp_request->dynamic) {
+		kfree(cqp_request);
+	} else {
+		spin_lock_irqsave(&nesdev->cqp.lock, flags);
+		list_add_tail(&cqp_request->list, &nesdev->cqp_avail_reqs);
+		spin_unlock_irqrestore(&nesdev->cqp.lock, flags);
+	}
+}
+
+void nes_put_cqp_request(struct nes_device *nesdev,
+			 struct nes_cqp_request *cqp_request)
+{
+	if (atomic_dec_and_test(&cqp_request->refcount))
+		nes_free_cqp_request(nesdev, cqp_request);
+}
 
 /**
  * nes_post_cqp_request
diff --git a/drivers/infiniband/hw/nes/nes_verbs.c b/drivers/infiniband/hw/nes/nes_verbs.c
index d617da9bd35..464a98a6e11 100644
--- a/drivers/infiniband/hw/nes/nes_verbs.c
+++ b/drivers/infiniband/hw/nes/nes_verbs.c
@@ -55,7 +55,6 @@ static void nes_unregister_ofa_device(struct nes_ib_device *nesibdev);
  * nes_alloc_mw
  */
 static struct ib_mw *nes_alloc_mw(struct ib_pd *ibpd) {
-	unsigned long flags;
 	struct nes_pd *nespd = to_nespd(ibpd);
 	struct nes_vnic *nesvnic = to_nesvnic(ibpd->device);
 	struct nes_device *nesdev = nesvnic->nesdev;
@@ -128,15 +127,7 @@ static struct ib_mw *nes_alloc_mw(struct ib_pd *ibpd) {
 			" CQP Major:Minor codes = 0x%04X:0x%04X.\n",
 			stag, ret, cqp_request->major_code, cqp_request->minor_code);
 	if ((!ret) || (cqp_request->major_code)) {
-		if (atomic_dec_and_test(&cqp_request->refcount)) {
-			if (cqp_request->dynamic) {
-				kfree(cqp_request);
-			} else {
-				spin_lock_irqsave(&nesdev->cqp.lock, flags);
-				list_add_tail(&cqp_request->list, &nesdev->cqp_avail_reqs);
-				spin_unlock_irqrestore(&nesdev->cqp.lock, flags);
-			}
-		}
+		nes_put_cqp_request(nesdev, cqp_request);
 		kfree(nesmr);
 		nes_free_resource(nesadapter, nesadapter->allocated_mrs, stag_index);
 		if (!ret) {
@@ -144,17 +135,8 @@ static struct ib_mw *nes_alloc_mw(struct ib_pd *ibpd) {
 		} else {
 			return ERR_PTR(-ENOMEM);
 		}
-	} else {
-		if (atomic_dec_and_test(&cqp_request->refcount)) {
-			if (cqp_request->dynamic) {
-				kfree(cqp_request);
-			} else {
-				spin_lock_irqsave(&nesdev->cqp.lock, flags);
-				list_add_tail(&cqp_request->list, &nesdev->cqp_avail_reqs);
-				spin_unlock_irqrestore(&nesdev->cqp.lock, flags);
-			}
-		}
 	}
+	nes_put_cqp_request(nesdev, cqp_request);
 
 	nesmr->ibmw.rkey = stag;
 	nesmr->mode = IWNES_MEMREG_TYPE_MW;
@@ -178,7 +160,6 @@ static int nes_dealloc_mw(struct ib_mw *ibmw)
 	struct nes_hw_cqp_wqe *cqp_wqe;
 	struct nes_cqp_request *cqp_request;
 	int err = 0;
-	unsigned long flags;
 	int ret;
 
 	/* Deallocate the window with the adapter */
@@ -204,32 +185,12 @@ static int nes_dealloc_mw(struct ib_mw *ibmw)
 	nes_debug(NES_DBG_MR, "Deallocate STag completed, wait_event_timeout ret = %u,"
 			" CQP Major:Minor codes = 0x%04X:0x%04X.\n",
 			ret, cqp_request->major_code, cqp_request->minor_code);
-	if ((!ret) || (cqp_request->major_code)) {
-		if (atomic_dec_and_test(&cqp_request->refcount)) {
-			if (cqp_request->dynamic) {
-				kfree(cqp_request);
-			} else {
-				spin_lock_irqsave(&nesdev->cqp.lock, flags);
-				list_add_tail(&cqp_request->list, &nesdev->cqp_avail_reqs);
-				spin_unlock_irqrestore(&nesdev->cqp.lock, flags);
-			}
-		}
-		if (!ret) {
-			err = -ETIME;
-		} else {
-			err = -EIO;
-		}
-	} else {
-		if (atomic_dec_and_test(&cqp_request->refcount)) {
-			if (cqp_request->dynamic) {
-				kfree(cqp_request);
-			} else {
-				spin_lock_irqsave(&nesdev->cqp.lock, flags);
-				list_add_tail(&cqp_request->list, &nesdev->cqp_avail_reqs);
-				spin_unlock_irqrestore(&nesdev->cqp.lock, flags);
-			}
-		}
-	}
+	if (!ret)
+		err = -ETIME;
+	else if (cqp_request->major_code)
+		err = -EIO;
+
+	nes_put_cqp_request(nesdev, cqp_request);
 
 	nes_free_resource(nesadapter, nesadapter->allocated_mrs,
 			(ibmw->rkey & 0x0fffff00) >> 8);
@@ -526,29 +487,11 @@ static struct ib_fmr *nes_alloc_fmr(struct ib_pd *ibpd,
 			stag, ret, cqp_request->major_code, cqp_request->minor_code);
 
 	if ((!ret) || (cqp_request->major_code)) {
-		if (atomic_dec_and_test(&cqp_request->refcount)) {
-			if (cqp_request->dynamic) {
-				kfree(cqp_request);
-			} else {
-				spin_lock_irqsave(&nesdev->cqp.lock, flags);
-				list_add_tail(&cqp_request->list, &nesdev->cqp_avail_reqs);
-				spin_unlock_irqrestore(&nesdev->cqp.lock, flags);
-			}
-		}
+		nes_put_cqp_request(nesdev, cqp_request);
 		ret = (!ret) ? -ETIME : -EIO;
 		goto failed_leaf_vpbl_pages_alloc;
-	} else {
-		if (atomic_dec_and_test(&cqp_request->refcount)) {
-			if (cqp_request->dynamic) {
-				kfree(cqp_request);
-			} else {
-				spin_lock_irqsave(&nesdev->cqp.lock, flags);
-				list_add_tail(&cqp_request->list, &nesdev->cqp_avail_reqs);
-				spin_unlock_irqrestore(&nesdev->cqp.lock, flags);
-			}
-		}
 	}
-
+	nes_put_cqp_request(nesdev, cqp_request);
 	nesfmr->nesmr.ibfmr.lkey = stag;
 	nesfmr->nesmr.ibfmr.rkey = stag;
 	nesfmr->attr = *ibfmr_attr;
@@ -1487,15 +1430,7 @@ static struct ib_qp *nes_create_qp(struct ib_pd *ibpd,
 					nesqp->hwqp.qp_id, ret, nesdev->cqp.sq_head, nesdev->cqp.sq_tail,
 					cqp_request->major_code, cqp_request->minor_code);
 			if ((!ret) || (cqp_request->major_code)) {
-				if (atomic_dec_and_test(&cqp_request->refcount)) {
-					if (cqp_request->dynamic) {
-						kfree(cqp_request);
-					} else {
-						spin_lock_irqsave(&nesdev->cqp.lock, flags);
-						list_add_tail(&cqp_request->list, &nesdev->cqp_avail_reqs);
-						spin_unlock_irqrestore(&nesdev->cqp.lock, flags);
-					}
-				}
+				nes_put_cqp_request(nesdev, cqp_request);
 				nes_free_resource(nesadapter, nesadapter->allocated_qps, qp_num);
 				nes_free_qp_mem(nesdev, nesqp,virt_wqs);
 				kfree(nesqp->allocated_buffer);
@@ -1504,18 +1439,10 @@ static struct ib_qp *nes_create_qp(struct ib_pd *ibpd,
 				} else {
 					return ERR_PTR(-EIO);
 				}
-			} else {
-				if (atomic_dec_and_test(&cqp_request->refcount)) {
-					if (cqp_request->dynamic) {
-						kfree(cqp_request);
-					} else {
-						spin_lock_irqsave(&nesdev->cqp.lock, flags);
-						list_add_tail(&cqp_request->list, &nesdev->cqp_avail_reqs);
-						spin_unlock_irqrestore(&nesdev->cqp.lock, flags);
-					}
-				}
 			}
 
+			nes_put_cqp_request(nesdev, cqp_request);
+
 			if (ibpd->uobject) {
 				uresp.mmap_sq_db_index = nesqp->mmap_sq_db_index;
 				uresp.actual_sq_size = sq_size;
@@ -1827,32 +1754,15 @@ static struct ib_cq *nes_create_cq(struct ib_device *ibdev, int entries,
 	nes_debug(NES_DBG_CQ, "Create iWARP CQ%u completed, wait_event_timeout ret = %d.\n",
 			nescq->hw_cq.cq_number, ret);
 	if ((!ret) || (cqp_request->major_code)) {
-		if (atomic_dec_and_test(&cqp_request->refcount)) {
-			if (cqp_request->dynamic) {
-				kfree(cqp_request);
-			} else {
-				spin_lock_irqsave(&nesdev->cqp.lock, flags);
-				list_add_tail(&cqp_request->list, &nesdev->cqp_avail_reqs);
-				spin_unlock_irqrestore(&nesdev->cqp.lock, flags);
-			}
-		}
+		nes_put_cqp_request(nesdev, cqp_request);
 		if (!context)
 			pci_free_consistent(nesdev->pcidev, nescq->cq_mem_size, mem,
 					nescq->hw_cq.cq_pbase);
 		nes_free_resource(nesadapter, nesadapter->allocated_cqs, cq_num);
 		kfree(nescq);
 		return ERR_PTR(-EIO);
-	} else {
-		if (atomic_dec_and_test(&cqp_request->refcount)) {
-			if (cqp_request->dynamic) {
-				kfree(cqp_request);
-			} else {
-				spin_lock_irqsave(&nesdev->cqp.lock, flags);
-				list_add_tail(&cqp_request->list, &nesdev->cqp_avail_reqs);
-				spin_unlock_irqrestore(&nesdev->cqp.lock, flags);
-			}
-		}
 	}
+	nes_put_cqp_request(nesdev, cqp_request);
 
 	if (context) {
 		/* free the nespbl */
@@ -1942,37 +1852,18 @@ static int nes_destroy_cq(struct ib_cq *ib_cq)
 			" CQP Major:Minor codes = 0x%04X:0x%04X.\n",
 			nescq->hw_cq.cq_number, ret, cqp_request->major_code,
 			cqp_request->minor_code);
-	if ((!ret) || (cqp_request->major_code)) {
-		if (atomic_dec_and_test(&cqp_request->refcount)) {
-			if (cqp_request->dynamic) {
-				kfree(cqp_request);
-			} else {
-				spin_lock_irqsave(&nesdev->cqp.lock, flags);
-				list_add_tail(&cqp_request->list, &nesdev->cqp_avail_reqs);
-				spin_unlock_irqrestore(&nesdev->cqp.lock, flags);
-			}
-		}
-		if (!ret) {
-			nes_debug(NES_DBG_CQ, "iWARP CQ%u destroy timeout expired\n",
+	if (!ret) {
+		nes_debug(NES_DBG_CQ, "iWARP CQ%u destroy timeout expired\n",
 					nescq->hw_cq.cq_number);
-			ret = -ETIME;
-		} else {
-			nes_debug(NES_DBG_CQ, "iWARP CQ%u destroy failed\n",
+		ret = -ETIME;
+	} else if (cqp_request->major_code) {
+		nes_debug(NES_DBG_CQ, "iWARP CQ%u destroy failed\n",
 					nescq->hw_cq.cq_number);
-			ret = -EIO;
-		}
+		ret = -EIO;
 	} else {
 		ret = 0;
-		if (atomic_dec_and_test(&cqp_request->refcount)) {
-			if (cqp_request->dynamic) {
-				kfree(cqp_request);
-			} else {
-				spin_lock_irqsave(&nesdev->cqp.lock, flags);
-				list_add_tail(&cqp_request->list, &nesdev->cqp_avail_reqs);
-				spin_unlock_irqrestore(&nesdev->cqp.lock, flags);
-			}
-		}
 	}
+	nes_put_cqp_request(nesdev, cqp_request);
 
 	if (nescq->cq_mem_size)
 		pci_free_consistent(nesdev->pcidev, nescq->cq_mem_size,
@@ -2105,15 +1996,8 @@ static int nes_reg_mr(struct nes_device *nesdev, struct nes_pd *nespd,
 			" CQP Major:Minor codes = 0x%04X:0x%04X.\n",
 			stag, ret, cqp_request->major_code, cqp_request->minor_code);
 	major_code = cqp_request->major_code;
-	if (atomic_dec_and_test(&cqp_request->refcount)) {
-		if (cqp_request->dynamic) {
-			kfree(cqp_request);
-		} else {
-			spin_lock_irqsave(&nesdev->cqp.lock, flags);
-			list_add_tail(&cqp_request->list, &nesdev->cqp_avail_reqs);
-			spin_unlock_irqrestore(&nesdev->cqp.lock, flags);
-		}
-	}
+	nes_put_cqp_request(nesdev, cqp_request);
+
 	if (!ret)
 		return -ETIME;
 	else if (major_code)
@@ -2771,15 +2655,9 @@ static int nes_dereg_mr(struct ib_mr *ib_mr)
 
 	major_code = cqp_request->major_code;
 	minor_code = cqp_request->minor_code;
-	if (atomic_dec_and_test(&cqp_request->refcount)) {
-		if (cqp_request->dynamic) {
-			kfree(cqp_request);
-		} else {
-			spin_lock_irqsave(&nesdev->cqp.lock, flags);
-			list_add_tail(&cqp_request->list, &nesdev->cqp_avail_reqs);
-			spin_unlock_irqrestore(&nesdev->cqp.lock, flags);
-		}
-	}
+
+	nes_put_cqp_request(nesdev, cqp_request);
+
 	if (!ret) {
 		nes_debug(NES_DBG_MR, "Timeout waiting to destroy STag,"
 				" ib_mr=%p, rkey = 0x%08X\n",
@@ -2904,7 +2782,6 @@ int nes_hw_modify_qp(struct nes_device *nesdev, struct nes_qp *nesqp,
 	/* struct iw_cm_id *cm_id = nesqp->cm_id; */
 	/* struct iw_cm_event cm_event; */
 	struct nes_cqp_request *cqp_request;
-	unsigned long flags;
 	int ret;
 	u16 major_code;
 
@@ -2950,15 +2827,9 @@ int nes_hw_modify_qp(struct nes_device *nesdev, struct nes_qp *nesqp,
 					nesqp->hwqp.qp_id, cqp_request->major_code,
 					cqp_request->minor_code, next_iwarp_state);
 		}
-		if (atomic_dec_and_test(&cqp_request->refcount)) {
-			if (cqp_request->dynamic) {
-				kfree(cqp_request);
-			} else {
-				spin_lock_irqsave(&nesdev->cqp.lock, flags);
-				list_add_tail(&cqp_request->list, &nesdev->cqp_avail_reqs);
-				spin_unlock_irqrestore(&nesdev->cqp.lock, flags);
-			}
-		}
+
+		nes_put_cqp_request(nesdev, cqp_request);
+
 		if (!ret)
 			return -ETIME;
 		else if (major_code)
-- 
cgit v1.2.3


From 52c8084b740c42af27d5bfa62cec7079d12fbc2b Mon Sep 17 00:00:00 2001
From: Jon Mason <jon@opengridcomputing.com>
Date: Mon, 14 Jul 2008 23:48:49 -0700
Subject: RDMA/cxgb3: Propagate HW page size capabilities

cxgb3 does not currently report the page size capabilities, and
incorrectly reports them internally.

This version changes the bit-shifting to a static value (per Steve's
request).

Signed-off-by: Jon Mason <jon@opengridcomputing.com>
Acked-by: Steve Wise <swise@opengridcomputing.com>
Signed-off-by: Roland Dreier <rolandd@cisco.com>
---
 drivers/infiniband/hw/cxgb3/cxio_hal.h      | 1 +
 drivers/infiniband/hw/cxgb3/iwch.c          | 2 +-
 drivers/infiniband/hw/cxgb3/iwch_provider.c | 1 +
 3 files changed, 3 insertions(+), 1 deletion(-)

(limited to 'drivers/infiniband/hw')

diff --git a/drivers/infiniband/hw/cxgb3/cxio_hal.h b/drivers/infiniband/hw/cxgb3/cxio_hal.h
index a9ff32c3621..656fe47bc84 100644
--- a/drivers/infiniband/hw/cxgb3/cxio_hal.h
+++ b/drivers/infiniband/hw/cxgb3/cxio_hal.h
@@ -55,6 +55,7 @@
 #define T3_MAX_CQ_DEPTH 8192
 #define T3_MAX_NUM_STAG (1<<15)
 #define T3_MAX_MR_SIZE 0x100000000ULL
+#define T3_PAGESIZE_MASK 0xffff000  /* 4KB-128MB */
 
 #define T3_STAG_UNSET 0xffffffff
 
diff --git a/drivers/infiniband/hw/cxgb3/iwch.c b/drivers/infiniband/hw/cxgb3/iwch.c
index 06086256ca1..4489c89d671 100644
--- a/drivers/infiniband/hw/cxgb3/iwch.c
+++ b/drivers/infiniband/hw/cxgb3/iwch.c
@@ -80,7 +80,7 @@ static void rnic_init(struct iwch_dev *rnicp)
 	rnicp->attr.max_mem_regs = cxio_num_stags(&rnicp->rdev);
 	rnicp->attr.max_phys_buf_entries = T3_MAX_PBL_SIZE;
 	rnicp->attr.max_pds = T3_MAX_NUM_PD - 1;
-	rnicp->attr.mem_pgsizes_bitmask = 0x7FFF;	/* 4KB-128MB */
+	rnicp->attr.mem_pgsizes_bitmask = T3_PAGESIZE_MASK;
 	rnicp->attr.max_mr_size = T3_MAX_MR_SIZE;
 	rnicp->attr.can_resize_wq = 0;
 	rnicp->attr.max_rdma_reads_per_qp = 8;
diff --git a/drivers/infiniband/hw/cxgb3/iwch_provider.c b/drivers/infiniband/hw/cxgb3/iwch_provider.c
index 18a6609f5e0..249d99f4a3c 100644
--- a/drivers/infiniband/hw/cxgb3/iwch_provider.c
+++ b/drivers/infiniband/hw/cxgb3/iwch_provider.c
@@ -1130,6 +1130,7 @@ static int iwch_query_device(struct ib_device *ibdev,
 	props->hw_ver = dev->rdev.t3cdev_p->type;
 	props->fw_ver = fw_vers_string_to_u64(dev);
 	props->device_cap_flags = dev->device_cap_flags;
+	props->page_size_cap = dev->attr.mem_pgsizes_bitmask;
 	props->vendor_id = (u32)dev->rdev.rnic_info.pdev->vendor;
 	props->vendor_part_id = (u32)dev->rdev.rnic_info.pdev->device;
 	props->max_mr_size = dev->attr.max_mr_size;
-- 
cgit v1.2.3


From 8294f29767c53e97664a27db9974adea8e2ea95b Mon Sep 17 00:00:00 2001
From: Roland Dreier <rolandd@cisco.com>
Date: Mon, 14 Jul 2008 23:48:49 -0700
Subject: RDMA/nes: Get rid of ring_doorbell parameter of
 nes_post_cqp_request()

Every caller of nes_post_cqp_request() passed it NES_CQP_REQUEST_RING_DOORBELL,
so just remove that parameter and always ring the doorbell.

Signed-off-by: Roland Dreier <rolandd@cisco.com>
Acked-by: Faisal Latif <flatif@neteffect.com>
---
 drivers/infiniband/hw/nes/nes.c       |  2 +-
 drivers/infiniband/hw/nes/nes.h       |  5 +----
 drivers/infiniband/hw/nes/nes_hw.c    |  6 +++---
 drivers/infiniband/hw/nes/nes_hw.h    |  2 +-
 drivers/infiniband/hw/nes/nes_utils.c |  9 ++++-----
 drivers/infiniband/hw/nes/nes_verbs.c | 18 +++++++++---------
 6 files changed, 19 insertions(+), 23 deletions(-)

(limited to 'drivers/infiniband/hw')

diff --git a/drivers/infiniband/hw/nes/nes.c b/drivers/infiniband/hw/nes/nes.c
index a4e9269a29b..d2884e77809 100644
--- a/drivers/infiniband/hw/nes/nes.c
+++ b/drivers/infiniband/hw/nes/nes.c
@@ -328,7 +328,7 @@ void nes_rem_ref(struct ib_qp *ibqp)
 		set_wqe_32bit_value(cqp_wqe->wqe_words, NES_CQP_WQE_ID_IDX, nesqp->hwqp.qp_id);
 		u64temp = (u64)nesqp->nesqp_context_pbase;
 		set_wqe_64bit_value(cqp_wqe->wqe_words, NES_CQP_QP_WQE_CONTEXT_LOW_IDX, u64temp);
-		nes_post_cqp_request(nesdev, cqp_request, NES_CQP_REQUEST_RING_DOORBELL);
+		nes_post_cqp_request(nesdev, cqp_request);
 	}
 }
 
diff --git a/drivers/infiniband/hw/nes/nes.h b/drivers/infiniband/hw/nes/nes.h
index fe88bec4894..39bd897b40c 100644
--- a/drivers/infiniband/hw/nes/nes.h
+++ b/drivers/infiniband/hw/nes/nes.h
@@ -94,9 +94,6 @@
 
 #define MAX_DPC_ITERATIONS               128
 
-#define NES_CQP_REQUEST_NO_DOORBELL_RING 0
-#define NES_CQP_REQUEST_RING_DOORBELL    1
-
 #define NES_DRV_OPT_ENABLE_MPA_VER_0     0x00000001
 #define NES_DRV_OPT_DISABLE_MPA_CRC      0x00000002
 #define NES_DRV_OPT_DISABLE_FIRST_WRITE  0x00000004
@@ -542,7 +539,7 @@ void nes_free_cqp_request(struct nes_device *nesdev,
 			  struct nes_cqp_request *cqp_request);
 void nes_put_cqp_request(struct nes_device *nesdev,
 			 struct nes_cqp_request *cqp_request);
-void nes_post_cqp_request(struct nes_device *, struct nes_cqp_request *, int);
+void nes_post_cqp_request(struct nes_device *, struct nes_cqp_request *);
 int nes_arp_table(struct nes_device *, u32, u8 *, u32);
 void nes_mh_fix(unsigned long);
 void nes_clc(unsigned long);
diff --git a/drivers/infiniband/hw/nes/nes_hw.c b/drivers/infiniband/hw/nes/nes_hw.c
index 80e486653ec..902b1375a5d 100644
--- a/drivers/infiniband/hw/nes/nes_hw.c
+++ b/drivers/infiniband/hw/nes/nes_hw.c
@@ -3147,7 +3147,7 @@ int nes_manage_apbvt(struct nes_vnic *nesvnic, u32 accel_local_port,
 	nes_debug(NES_DBG_QP, "Waiting for CQP completion for APBVT.\n");
 
 	atomic_set(&cqp_request->refcount, 2);
-	nes_post_cqp_request(nesdev, cqp_request, NES_CQP_REQUEST_RING_DOORBELL);
+	nes_post_cqp_request(nesdev, cqp_request);
 
 	if (add_port == NES_MANAGE_APBVT_ADD)
 		ret = wait_event_timeout(cqp_request->waitq, (cqp_request->request_done != 0),
@@ -3217,7 +3217,7 @@ void nes_manage_arp_cache(struct net_device *netdev, unsigned char *mac_addr,
 			nesdev->cqp.sq_head, nesdev->cqp.sq_tail);
 
 	atomic_set(&cqp_request->refcount, 1);
-	nes_post_cqp_request(nesdev, cqp_request, NES_CQP_REQUEST_RING_DOORBELL);
+	nes_post_cqp_request(nesdev, cqp_request);
 }
 
 
@@ -3249,7 +3249,7 @@ void flush_wqes(struct nes_device *nesdev, struct nes_qp *nesqp,
 			cpu_to_le32(NES_CQP_FLUSH_WQES | which_wq);
 	cqp_wqe->wqe_words[NES_CQP_WQE_ID_IDX] = cpu_to_le32(nesqp->hwqp.qp_id);
 
-	nes_post_cqp_request(nesdev, cqp_request, NES_CQP_REQUEST_RING_DOORBELL);
+	nes_post_cqp_request(nesdev, cqp_request);
 
 	if (wait_completion) {
 		/* Wait for CQP */
diff --git a/drivers/infiniband/hw/nes/nes_hw.h b/drivers/infiniband/hw/nes/nes_hw.h
index 745bf94f3f0..7b81e0ae007 100644
--- a/drivers/infiniband/hw/nes/nes_hw.h
+++ b/drivers/infiniband/hw/nes/nes_hw.h
@@ -1172,7 +1172,7 @@ struct nes_vnic {
 	u32    mcrq_qp_id;
 	struct nes_ucontext *mcrq_ucontext;
 	struct nes_cqp_request* (*get_cqp_request)(struct nes_device *nesdev);
-	void (*post_cqp_request)(struct nes_device*, struct nes_cqp_request *, int);
+	void (*post_cqp_request)(struct nes_device*, struct nes_cqp_request *);
 	int (*mcrq_mcast_filter)( struct nes_vnic* nesvnic, __u8* dmi_addr );
 	struct net_device_stats netstats;
 	/* used to put the netdev on the adapters logical port list */
diff --git a/drivers/infiniband/hw/nes/nes_utils.c b/drivers/infiniband/hw/nes/nes_utils.c
index 8f519c86087..fb8cbd71a2e 100644
--- a/drivers/infiniband/hw/nes/nes_utils.c
+++ b/drivers/infiniband/hw/nes/nes_utils.c
@@ -596,7 +596,7 @@ void nes_put_cqp_request(struct nes_device *nesdev,
  * nes_post_cqp_request
  */
 void nes_post_cqp_request(struct nes_device *nesdev,
-		struct nes_cqp_request *cqp_request, int ring_doorbell)
+			  struct nes_cqp_request *cqp_request)
 {
 	struct nes_hw_cqp_wqe *cqp_wqe;
 	unsigned long flags;
@@ -624,10 +624,9 @@ void nes_post_cqp_request(struct nes_device *nesdev,
 				nesdev->cqp.sq_head, nesdev->cqp.sq_tail, nesdev->cqp.sq_size,
 				cqp_request->waiting, atomic_read(&cqp_request->refcount));
 		barrier();
-		if (ring_doorbell) {
-			/* Ring doorbell (1 WQEs) */
-			nes_write32(nesdev->regs+NES_WQE_ALLOC, 0x01800000 | nesdev->cqp.qp_id);
-		}
+
+		/* Ring doorbell (1 WQEs) */
+		nes_write32(nesdev->regs+NES_WQE_ALLOC, 0x01800000 | nesdev->cqp.qp_id);
 
 		barrier();
 	} else {
diff --git a/drivers/infiniband/hw/nes/nes_verbs.c b/drivers/infiniband/hw/nes/nes_verbs.c
index 464a98a6e11..e3939d13484 100644
--- a/drivers/infiniband/hw/nes/nes_verbs.c
+++ b/drivers/infiniband/hw/nes/nes_verbs.c
@@ -118,7 +118,7 @@ static struct ib_mw *nes_alloc_mw(struct ib_pd *ibpd) {
 	set_wqe_32bit_value(cqp_wqe->wqe_words, NES_CQP_STAG_WQE_STAG_IDX, stag);
 
 	atomic_set(&cqp_request->refcount, 2);
-	nes_post_cqp_request(nesdev, cqp_request, NES_CQP_REQUEST_RING_DOORBELL);
+	nes_post_cqp_request(nesdev, cqp_request);
 
 	/* Wait for CQP */
 	ret = wait_event_timeout(cqp_request->waitq, (cqp_request->request_done != 0),
@@ -175,7 +175,7 @@ static int nes_dealloc_mw(struct ib_mw *ibmw)
 	set_wqe_32bit_value(cqp_wqe->wqe_words, NES_CQP_STAG_WQE_STAG_IDX, ibmw->rkey);
 
 	atomic_set(&cqp_request->refcount, 2);
-	nes_post_cqp_request(nesdev, cqp_request, NES_CQP_REQUEST_RING_DOORBELL);
+	nes_post_cqp_request(nesdev, cqp_request);
 
 	/* Wait for CQP */
 	nes_debug(NES_DBG_MR, "Waiting for deallocate STag 0x%08X to complete.\n",
@@ -477,7 +477,7 @@ static struct ib_fmr *nes_alloc_fmr(struct ib_pd *ibpd,
 			(nesfmr->nesmr.pbls_used-1) : nesfmr->nesmr.pbls_used);
 
 	atomic_set(&cqp_request->refcount, 2);
-	nes_post_cqp_request(nesdev, cqp_request, NES_CQP_REQUEST_RING_DOORBELL);
+	nes_post_cqp_request(nesdev, cqp_request);
 
 	/* Wait for CQP */
 	ret = wait_event_timeout(cqp_request->waitq, (cqp_request->request_done != 0),
@@ -1417,7 +1417,7 @@ static struct ib_qp *nes_create_qp(struct ib_pd *ibpd,
 			set_wqe_64bit_value(cqp_wqe->wqe_words, NES_CQP_QP_WQE_CONTEXT_LOW_IDX, u64temp);
 
 			atomic_set(&cqp_request->refcount, 2);
-			nes_post_cqp_request(nesdev, cqp_request, NES_CQP_REQUEST_RING_DOORBELL);
+			nes_post_cqp_request(nesdev, cqp_request);
 
 			/* Wait for CQP */
 			nes_debug(NES_DBG_QP, "Waiting for create iWARP QP%u to complete.\n",
@@ -1744,7 +1744,7 @@ static struct ib_cq *nes_create_cq(struct ib_device *ibdev, int entries,
 			cpu_to_le32(((u32)((u64temp) >> 33)) & 0x7FFFFFFF);
 
 	atomic_set(&cqp_request->refcount, 2);
-	nes_post_cqp_request(nesdev, cqp_request, NES_CQP_REQUEST_RING_DOORBELL);
+	nes_post_cqp_request(nesdev, cqp_request);
 
 	/* Wait for CQP */
 	nes_debug(NES_DBG_CQ, "Waiting for create iWARP CQ%u to complete.\n",
@@ -1841,7 +1841,7 @@ static int nes_destroy_cq(struct ib_cq *ib_cq)
 		(nescq->hw_cq.cq_number | ((u32)PCI_FUNC(nesdev->pcidev->devfn) << 16)));
 	nes_free_resource(nesadapter, nesadapter->allocated_cqs, nescq->hw_cq.cq_number);
 	atomic_set(&cqp_request->refcount, 2);
-	nes_post_cqp_request(nesdev, cqp_request, NES_CQP_REQUEST_RING_DOORBELL);
+	nes_post_cqp_request(nesdev, cqp_request);
 
 	/* Wait for CQP */
 	nes_debug(NES_DBG_CQ, "Waiting for destroy iWARP CQ%u to complete.\n",
@@ -1987,7 +1987,7 @@ static int nes_reg_mr(struct nes_device *nesdev, struct nes_pd *nespd,
 	barrier();
 
 	atomic_set(&cqp_request->refcount, 2);
-	nes_post_cqp_request(nesdev, cqp_request, NES_CQP_REQUEST_RING_DOORBELL);
+	nes_post_cqp_request(nesdev, cqp_request);
 
 	/* Wait for CQP */
 	ret = wait_event_timeout(cqp_request->waitq, (0 != cqp_request->request_done),
@@ -2638,7 +2638,7 @@ static int nes_dereg_mr(struct ib_mr *ib_mr)
 	set_wqe_32bit_value(cqp_wqe->wqe_words, NES_CQP_STAG_WQE_STAG_IDX, ib_mr->rkey);
 
 	atomic_set(&cqp_request->refcount, 2);
-	nes_post_cqp_request(nesdev, cqp_request, NES_CQP_REQUEST_RING_DOORBELL);
+	nes_post_cqp_request(nesdev, cqp_request);
 
 	/* Wait for CQP */
 	nes_debug(NES_DBG_MR, "Waiting for deallocate STag 0x%08X completed\n", ib_mr->rkey);
@@ -2809,7 +2809,7 @@ int nes_hw_modify_qp(struct nes_device *nesdev, struct nes_qp *nesqp,
 	set_wqe_64bit_value(cqp_wqe->wqe_words, NES_CQP_QP_WQE_CONTEXT_LOW_IDX, (u64)nesqp->nesqp_context_pbase);
 
 	atomic_set(&cqp_request->refcount, 2);
-	nes_post_cqp_request(nesdev, cqp_request, NES_CQP_REQUEST_RING_DOORBELL);
+	nes_post_cqp_request(nesdev, cqp_request);
 
 	/* Wait for CQP */
 	if (wait_completion) {
-- 
cgit v1.2.3


From 70fe1796a5ebc5f955be39bba5c42eee9eb89e1f Mon Sep 17 00:00:00 2001
From: Steve Wise <swise@opengridcomputing.com>
Date: Mon, 14 Jul 2008 23:48:49 -0700
Subject: RDMA/cxgb3: Set rkey field for new memory windows in iwch_alloc_mw()

Signed-off-by: Steve Wise <swise@opengridcomputing.com>
Signed-off-by: Roland Dreier <rolandd@cisco.com>
---
 drivers/infiniband/hw/cxgb3/iwch_provider.c | 1 +
 1 file changed, 1 insertion(+)

(limited to 'drivers/infiniband/hw')

diff --git a/drivers/infiniband/hw/cxgb3/iwch_provider.c b/drivers/infiniband/hw/cxgb3/iwch_provider.c
index 249d99f4a3c..c9a3893b38e 100644
--- a/drivers/infiniband/hw/cxgb3/iwch_provider.c
+++ b/drivers/infiniband/hw/cxgb3/iwch_provider.c
@@ -748,6 +748,7 @@ static struct ib_mw *iwch_alloc_mw(struct ib_pd *pd)
 	mhp->attr.type = TPT_MW;
 	mhp->attr.stag = stag;
 	mmid = (stag) >> 8;
+	mhp->ibmw.rkey = stag;
 	insert_handle(rhp, &rhp->mmidr, mhp, mmid);
 	PDBG("%s mmid 0x%x mhp %p stag 0x%x\n", __func__, mmid, mhp, stag);
 	return &(mhp->ibmw);
-- 
cgit v1.2.3


From df8666198dd058b9498ebdbc52c61957206d30a5 Mon Sep 17 00:00:00 2001
From: Ralph Campbell <ralph.campbell@qlogic.com>
Date: Mon, 14 Jul 2008 23:48:52 -0700
Subject: IB/ipath: Use IEEE OUI for vendor_id reported by ibv_query_device()

The IB spe. for SubnGet(NodeInfo) and query HCA says that the vendor
ID field should be the IEEE OUI assigned to the vendor.  The ipath
driver was returning the PCI vendor ID instead.  This will affect
applications which call ibv_query_device().  The old value was
0x001fc1 or 0x001077, the new value is 0x001175.

The vendor ID doesn't appear to be exported via /sys so that should
reduce possible compatibility issues.  I'm only aware of Open MPI as a
major application which depends on this change, and they have made
necessary adjustments.

Signed-off-by: Ralph Campbell <ralph.campbell@qlogic.com>
Signed-off-by: Roland Dreier <rolandd@cisco.com>
---
 drivers/infiniband/hw/ipath/ipath_mad.c   | 6 +++---
 drivers/infiniband/hw/ipath/ipath_verbs.c | 3 ++-
 2 files changed, 5 insertions(+), 4 deletions(-)

(limited to 'drivers/infiniband/hw')

diff --git a/drivers/infiniband/hw/ipath/ipath_mad.c b/drivers/infiniband/hw/ipath/ipath_mad.c
index 5f9315d77a4..be4fc9ada8e 100644
--- a/drivers/infiniband/hw/ipath/ipath_mad.c
+++ b/drivers/infiniband/hw/ipath/ipath_mad.c
@@ -111,9 +111,9 @@ static int recv_subn_get_nodeinfo(struct ib_smp *smp,
 	nip->revision = cpu_to_be32((majrev << 16) | minrev);
 	nip->local_port_num = port;
 	vendor = dd->ipath_vendorid;
-	nip->vendor_id[0] = 0;
-	nip->vendor_id[1] = vendor >> 8;
-	nip->vendor_id[2] = vendor;
+	nip->vendor_id[0] = IPATH_SRC_OUI_1;
+	nip->vendor_id[1] = IPATH_SRC_OUI_2;
+	nip->vendor_id[2] = IPATH_SRC_OUI_3;
 
 	return reply(smp);
 }
diff --git a/drivers/infiniband/hw/ipath/ipath_verbs.c b/drivers/infiniband/hw/ipath/ipath_verbs.c
index 7779165b2c2..9e23ab0b51a 100644
--- a/drivers/infiniband/hw/ipath/ipath_verbs.c
+++ b/drivers/infiniband/hw/ipath/ipath_verbs.c
@@ -1497,7 +1497,8 @@ static int ipath_query_device(struct ib_device *ibdev,
 		IB_DEVICE_SYS_IMAGE_GUID | IB_DEVICE_RC_RNR_NAK_GEN |
 		IB_DEVICE_PORT_ACTIVE_EVENT | IB_DEVICE_SRQ_RESIZE;
 	props->page_size_cap = PAGE_SIZE;
-	props->vendor_id = dev->dd->ipath_vendorid;
+	props->vendor_id =
+		IPATH_SRC_OUI_1 << 16 | IPATH_SRC_OUI_2 << 8 | IPATH_SRC_OUI_3;
 	props->vendor_part_id = dev->dd->ipath_deviceid;
 	props->hw_ver = dev->dd->ipath_pcirev;
 
-- 
cgit v1.2.3


From 4522e08ced48baaf28990e2674e940aae9940310 Mon Sep 17 00:00:00 2001
From: Roland Dreier <rolandd@cisco.com>
Date: Mon, 14 Jul 2008 23:48:52 -0700
Subject: IB/mthca: Remove "stop" flag for catastrophic error polling timer

Since we use del_timer_sync() anyway, there's no need for an
additional flag to tell the timer not to rearm.

Signed-off-by: Roland Dreier <rolandd@cisco.com>
---
 drivers/infiniband/hw/mthca/mthca_catas.c | 15 ++-------------
 drivers/infiniband/hw/mthca/mthca_dev.h   |  1 -
 2 files changed, 2 insertions(+), 14 deletions(-)

(limited to 'drivers/infiniband/hw')

diff --git a/drivers/infiniband/hw/mthca/mthca_catas.c b/drivers/infiniband/hw/mthca/mthca_catas.c
index 40573e48439..50e2792e60c 100644
--- a/drivers/infiniband/hw/mthca/mthca_catas.c
+++ b/drivers/infiniband/hw/mthca/mthca_catas.c
@@ -126,7 +126,6 @@ static void handle_catas(struct mthca_dev *dev)
 static void poll_catas(unsigned long dev_ptr)
 {
 	struct mthca_dev *dev = (struct mthca_dev *) dev_ptr;
-	unsigned long flags;
 	int i;
 
 	for (i = 0; i < dev->catas_err.size; ++i)
@@ -135,13 +134,8 @@ static void poll_catas(unsigned long dev_ptr)
 			return;
 		}
 
-	spin_lock_irqsave(&catas_lock, flags);
-	if (!dev->catas_err.stop)
-		mod_timer(&dev->catas_err.timer,
-			  jiffies + MTHCA_CATAS_POLL_INTERVAL);
-	spin_unlock_irqrestore(&catas_lock, flags);
-
-	return;
+	mod_timer(&dev->catas_err.timer,
+		  jiffies + MTHCA_CATAS_POLL_INTERVAL);
 }
 
 void mthca_start_catas_poll(struct mthca_dev *dev)
@@ -149,7 +143,6 @@ void mthca_start_catas_poll(struct mthca_dev *dev)
 	unsigned long addr;
 
 	init_timer(&dev->catas_err.timer);
-	dev->catas_err.stop = 0;
 	dev->catas_err.map  = NULL;
 
 	addr = pci_resource_start(dev->pdev, 0) +
@@ -180,10 +173,6 @@ void mthca_start_catas_poll(struct mthca_dev *dev)
 
 void mthca_stop_catas_poll(struct mthca_dev *dev)
 {
-	spin_lock_irq(&catas_lock);
-	dev->catas_err.stop = 1;
-	spin_unlock_irq(&catas_lock);
-
 	del_timer_sync(&dev->catas_err.timer);
 
 	if (dev->catas_err.map) {
diff --git a/drivers/infiniband/hw/mthca/mthca_dev.h b/drivers/infiniband/hw/mthca/mthca_dev.h
index 2997d8d564e..ee4d073c889 100644
--- a/drivers/infiniband/hw/mthca/mthca_dev.h
+++ b/drivers/infiniband/hw/mthca/mthca_dev.h
@@ -277,7 +277,6 @@ struct mthca_mcg_table {
 struct mthca_catas_err {
 	u64			addr;
 	u32 __iomem	       *map;
-	unsigned long		stop;
 	u32			size;
 	struct timer_list	timer;
 	struct list_head	list;
-- 
cgit v1.2.3


From c036925ac0e940a4e8525b08e89d2c64fe282c51 Mon Sep 17 00:00:00 2001
From: Roland Dreier <rolandd@cisco.com>
Date: Mon, 14 Jul 2008 23:48:52 -0700
Subject: IB/mthca: Use round_jiffies() for catastrophic error polling timer

Exactly when the catastrophic error polling timer function runs is not
important, so use round_jiffies() to save unnecessary wakeups.

Signed-off-by: Roland Dreier <rolandd@cisco.com>
---
 drivers/infiniband/hw/mthca/mthca_catas.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

(limited to 'drivers/infiniband/hw')

diff --git a/drivers/infiniband/hw/mthca/mthca_catas.c b/drivers/infiniband/hw/mthca/mthca_catas.c
index 50e2792e60c..cc440f90000 100644
--- a/drivers/infiniband/hw/mthca/mthca_catas.c
+++ b/drivers/infiniband/hw/mthca/mthca_catas.c
@@ -135,7 +135,7 @@ static void poll_catas(unsigned long dev_ptr)
 		}
 
 	mod_timer(&dev->catas_err.timer,
-		  jiffies + MTHCA_CATAS_POLL_INTERVAL);
+		  round_jiffies(jiffies + MTHCA_CATAS_POLL_INTERVAL));
 }
 
 void mthca_start_catas_poll(struct mthca_dev *dev)
-- 
cgit v1.2.3


From aed012279d35e88e29fd55737d8821604433f50a Mon Sep 17 00:00:00 2001
From: Roland Dreier <rolandd@cisco.com>
Date: Mon, 14 Jul 2008 23:48:52 -0700
Subject: IB/mthca: Fix check of max_send_sge for special QPs

The MLX transport requires two extra gather entries for sends (one for
the header and one for the checksum at the end, as the comment says).
However the code checked that max_recv_sge was not too big, instead of
checking max_send_sge as it should have.  Fix the code to check the
correct condition.

Signed-off-by: Roland Dreier <rolandd@cisco.com>
---
 drivers/infiniband/hw/mthca/mthca_qp.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

(limited to 'drivers/infiniband/hw')

diff --git a/drivers/infiniband/hw/mthca/mthca_qp.c b/drivers/infiniband/hw/mthca/mthca_qp.c
index 77b52b3adcc..f5081bfde6d 100644
--- a/drivers/infiniband/hw/mthca/mthca_qp.c
+++ b/drivers/infiniband/hw/mthca/mthca_qp.c
@@ -1249,10 +1249,10 @@ static int mthca_set_qp_size(struct mthca_dev *dev, struct ib_qp_cap *cap,
 		return -EINVAL;
 
 	/*
-	 * For MLX transport we need 2 extra S/G entries:
+	 * For MLX transport we need 2 extra send gather entries:
 	 * one for the header and one for the checksum at the end
 	 */
-	if (qp->transport == MLX && cap->max_recv_sge + 2 > dev->limits.max_sg)
+	if (qp->transport == MLX && cap->max_send_sge + 2 > dev->limits.max_sg)
 		return -EINVAL;
 
 	if (mthca_is_memfree(dev)) {
-- 
cgit v1.2.3


From 96f15c03532282366364ecfd20f04e49b5d96f3a Mon Sep 17 00:00:00 2001
From: Steve Wise <swise@opengridcomputing.com>
Date: Mon, 14 Jul 2008 23:48:53 -0700
Subject: RDMA/core: Add local DMA L_Key support

- Change the IB_DEVICE_ZERO_STAG flag to the transport-neutral name
  IB_DEVICE_LOCAL_DMA_LKEY, which is used by iWARP RNICs to indicate 0
  STag support and IB HCAs to indicate reserved L_Key support.

- Add a u32 local_dma_lkey member to struct ib_device.  Drivers fill
  this in with the appropriate local DMA L_Key (if they support it).

- Fix up the drivers using this flag.

Signed-off-by: Steve Wise <swise@opengridcomputing.com>
Signed-off-by: Roland Dreier <rolandd@cisco.com>
---
 drivers/infiniband/hw/amso1100/c2_rnic.c    | 2 +-
 drivers/infiniband/hw/cxgb3/iwch_provider.c | 6 ++++--
 drivers/infiniband/hw/nes/nes_hw.c          | 2 +-
 3 files changed, 6 insertions(+), 4 deletions(-)

(limited to 'drivers/infiniband/hw')

diff --git a/drivers/infiniband/hw/amso1100/c2_rnic.c b/drivers/infiniband/hw/amso1100/c2_rnic.c
index b1441aeb60c..dd05c483564 100644
--- a/drivers/infiniband/hw/amso1100/c2_rnic.c
+++ b/drivers/infiniband/hw/amso1100/c2_rnic.c
@@ -454,7 +454,7 @@ int __devinit c2_rnic_init(struct c2_dev *c2dev)
 	    (IB_DEVICE_RESIZE_MAX_WR |
 	     IB_DEVICE_CURR_QP_STATE_MOD |
 	     IB_DEVICE_SYS_IMAGE_GUID |
-	     IB_DEVICE_ZERO_STAG |
+	     IB_DEVICE_LOCAL_DMA_LKEY |
 	     IB_DEVICE_MEM_WINDOW);
 
 	/* Allocate the qptr_array */
diff --git a/drivers/infiniband/hw/cxgb3/iwch_provider.c b/drivers/infiniband/hw/cxgb3/iwch_provider.c
index c9a3893b38e..7ecfd4d638c 100644
--- a/drivers/infiniband/hw/cxgb3/iwch_provider.c
+++ b/drivers/infiniband/hw/cxgb3/iwch_provider.c
@@ -1325,8 +1325,10 @@ int iwch_register_device(struct iwch_dev *dev)
 	memset(&dev->ibdev.node_guid, 0, sizeof(dev->ibdev.node_guid));
 	memcpy(&dev->ibdev.node_guid, dev->rdev.t3cdev_p->lldev->dev_addr, 6);
 	dev->ibdev.owner = THIS_MODULE;
-	dev->device_cap_flags = IB_DEVICE_ZERO_STAG |
-				IB_DEVICE_MEM_WINDOW;
+	dev->device_cap_flags = IB_DEVICE_LOCAL_DMA_LKEY | IB_DEVICE_MEM_WINDOW;
+
+	/* cxgb3 supports STag 0. */
+	dev->ibdev.local_dma_lkey = 0;
 	if (fw_supports_fastreg(dev))
 		dev->device_cap_flags |= IB_DEVICE_MEM_MGT_EXTENSIONS;
 
diff --git a/drivers/infiniband/hw/nes/nes_hw.c b/drivers/infiniband/hw/nes/nes_hw.c
index 902b1375a5d..85f26d19a32 100644
--- a/drivers/infiniband/hw/nes/nes_hw.c
+++ b/drivers/infiniband/hw/nes/nes_hw.c
@@ -398,7 +398,7 @@ struct nes_adapter *nes_init_adapter(struct nes_device *nesdev, u8 hw_rev) {
 	nesadapter->base_pd = 1;
 
 	nesadapter->device_cap_flags =
-		IB_DEVICE_ZERO_STAG | IB_DEVICE_MEM_WINDOW;
+		IB_DEVICE_LOCAL_DMA_LKEY | IB_DEVICE_MEM_WINDOW;
 
 	nesadapter->allocated_qps = (unsigned long *)&(((unsigned char *)nesadapter)
 			[(sizeof(struct nes_adapter)+(sizeof(unsigned long)-1))&(~(sizeof(unsigned long)-1))]);
-- 
cgit v1.2.3


From 4ab928f69208d240d3681336f34589e4b151824f Mon Sep 17 00:00:00 2001
From: Steve Wise <swise@opengridcomputing.com>
Date: Mon, 14 Jul 2008 23:48:53 -0700
Subject: RDMA/cxgb3: Fixes for zero STag

Handling the zero STag in receive work request requires some extra
logic in the driver:

 - Only set the QP_PRIV bit for kernel mode QPs.

- Add a zero STag build function for recv wrs. The uP needs a PBL
  allocated and passed down in the recv WR so it can construct a HW
  PBL for the zero STag S/G entries.  Note: we need to place a few
  restrictions on zero STag usage because of this:

  1) all SGEs in a recv WR must either be zero STag or not.  No mixing.

  2) an individual SGE length cannot exceed 128MB for a zero-stag SGE.
     This should be OK since it's not really practical to allocate
     such a large chunk of pinned contiguous DMA mapped memory.

- Add an optimized non-zero-STag recv wr format for kernel users.
  This is needed to optimize both zero and non-zero STag cracking in
  the recv path for kernel users.

 - Remove the iwch_ prefix from the static build functions.

 - Bump required FW version.

Signed-off-by: Steve Wise <swise@opengridcomputing.com>
---
 drivers/infiniband/hw/cxgb3/cxio_hal.c      |  12 ++-
 drivers/infiniband/hw/cxgb3/cxio_wr.h       |  13 ++-
 drivers/infiniband/hw/cxgb3/iwch_provider.c |   4 +-
 drivers/infiniband/hw/cxgb3/iwch_qp.c       | 127 ++++++++++++++++++++++++----
 4 files changed, 130 insertions(+), 26 deletions(-)

(limited to 'drivers/infiniband/hw')

diff --git a/drivers/infiniband/hw/cxgb3/cxio_hal.c b/drivers/infiniband/hw/cxgb3/cxio_hal.c
index 340e4181c76..f6d5747153a 100644
--- a/drivers/infiniband/hw/cxgb3/cxio_hal.c
+++ b/drivers/infiniband/hw/cxgb3/cxio_hal.c
@@ -278,7 +278,7 @@ int cxio_create_qp(struct cxio_rdev *rdev_p, u32 kernel_domain,
 	if (!wq->qpid)
 		return -ENOMEM;
 
-	wq->rq = kzalloc(depth * sizeof(u64), GFP_KERNEL);
+	wq->rq = kzalloc(depth * sizeof(struct t3_swrq), GFP_KERNEL);
 	if (!wq->rq)
 		goto err1;
 
@@ -302,6 +302,7 @@ int cxio_create_qp(struct cxio_rdev *rdev_p, u32 kernel_domain,
 	if (!kernel_domain)
 		wq->udb = (u64)rdev_p->rnic_info.udbell_physbase +
 					(wq->qpid << rdev_p->qpshift);
+	wq->rdev = rdev_p;
 	PDBG("%s qpid 0x%x doorbell 0x%p udb 0x%llx\n", __func__,
 	     wq->qpid, wq->doorbell, (unsigned long long) wq->udb);
 	return 0;
@@ -1266,13 +1267,16 @@ proc_cqe:
 		wq->sq_rptr = CQE_WRID_SQ_WPTR(*hw_cqe);
 		PDBG("%s completing sq idx %ld\n", __func__,
 		     Q_PTR2IDX(wq->sq_rptr, wq->sq_size_log2));
-		*cookie = (wq->sq +
-			   Q_PTR2IDX(wq->sq_rptr, wq->sq_size_log2))->wr_id;
+		*cookie = wq->sq[Q_PTR2IDX(wq->sq_rptr, wq->sq_size_log2)].wr_id;
 		wq->sq_rptr++;
 	} else {
 		PDBG("%s completing rq idx %ld\n", __func__,
 		     Q_PTR2IDX(wq->rq_rptr, wq->rq_size_log2));
-		*cookie = *(wq->rq + Q_PTR2IDX(wq->rq_rptr, wq->rq_size_log2));
+		*cookie = wq->rq[Q_PTR2IDX(wq->rq_rptr, wq->rq_size_log2)].wr_id;
+		if (wq->rq[Q_PTR2IDX(wq->rq_rptr, wq->rq_size_log2)].pbl_addr)
+			cxio_hal_pblpool_free(wq->rdev,
+				wq->rq[Q_PTR2IDX(wq->rq_rptr,
+				wq->rq_size_log2)].pbl_addr, T3_STAG0_PBL_SIZE);
 		wq->rq_rptr++;
 	}
 
diff --git a/drivers/infiniband/hw/cxgb3/cxio_wr.h b/drivers/infiniband/hw/cxgb3/cxio_wr.h
index de760e9f1cc..04618f7bfbb 100644
--- a/drivers/infiniband/hw/cxgb3/cxio_wr.h
+++ b/drivers/infiniband/hw/cxgb3/cxio_wr.h
@@ -39,6 +39,9 @@
 
 #define T3_MAX_SGE      4
 #define T3_MAX_INLINE	64
+#define T3_STAG0_PBL_SIZE (2 * T3_MAX_SGE << 3)
+#define T3_STAG0_MAX_PBE_LEN (128 * 1024 * 1024)
+#define T3_STAG0_PAGE_SHIFT 15
 
 #define Q_EMPTY(rptr,wptr) ((rptr)==(wptr))
 #define Q_FULL(rptr,wptr,size_log2)  ( (((wptr)-(rptr))>>(size_log2)) && \
@@ -665,6 +668,11 @@ struct t3_swsq {
 	int			signaled;
 };
 
+struct t3_swrq {
+	__u64			wr_id;
+	__u32			pbl_addr;
+};
+
 /*
  * A T3 WQ implements both the SQ and RQ.
  */
@@ -681,14 +689,15 @@ struct t3_wq {
 	u32 sq_wptr;			/* sq_wptr - sq_rptr == count of */
 	u32 sq_rptr;			/* pending wrs */
 	u32 sq_size_log2;		/* sq size */
-	u64 *rq;			/* SW RQ (holds consumer wr_ids */
+	struct t3_swrq *rq;		/* SW RQ (holds consumer wr_ids */
 	u32 rq_wptr;			/* rq_wptr - rq_rptr == count of */
 	u32 rq_rptr;			/* pending wrs */
-	u64 *rq_oldest_wr;		/* oldest wr on the SW RQ */
+	struct t3_swrq *rq_oldest_wr;	/* oldest wr on the SW RQ */
 	u32 rq_size_log2;		/* rq size */
 	u32 rq_addr;			/* rq adapter address */
 	void __iomem *doorbell;		/* kernel db */
 	u64 udb;			/* user db if any */
+	struct cxio_rdev *rdev;
 };
 
 struct t3_cq {
diff --git a/drivers/infiniband/hw/cxgb3/iwch_provider.c b/drivers/infiniband/hw/cxgb3/iwch_provider.c
index 7ecfd4d638c..b89640aa6e1 100644
--- a/drivers/infiniband/hw/cxgb3/iwch_provider.c
+++ b/drivers/infiniband/hw/cxgb3/iwch_provider.c
@@ -1007,10 +1007,10 @@ static struct ib_qp *iwch_create_qp(struct ib_pd *pd,
 	qhp->ibqp.qp_num = qhp->wq.qpid;
 	init_timer(&(qhp->timer));
 	PDBG("%s sq_num_entries %d, rq_num_entries %d "
-	     "qpid 0x%0x qhp %p dma_addr 0x%llx size %d\n",
+	     "qpid 0x%0x qhp %p dma_addr 0x%llx size %d rq_addr 0x%x\n",
 	     __func__, qhp->attr.sq_num_entries, qhp->attr.rq_num_entries,
 	     qhp->wq.qpid, qhp, (unsigned long long) qhp->wq.dma_addr,
-	     1 << qhp->wq.size_log2);
+	     1 << qhp->wq.size_log2, qhp->wq.rq_addr);
 	return &qhp->ibqp;
 }
 
diff --git a/drivers/infiniband/hw/cxgb3/iwch_qp.c b/drivers/infiniband/hw/cxgb3/iwch_qp.c
index 3b44300a303..9a3be3a9d5d 100644
--- a/drivers/infiniband/hw/cxgb3/iwch_qp.c
+++ b/drivers/infiniband/hw/cxgb3/iwch_qp.c
@@ -33,10 +33,11 @@
 #include "iwch.h"
 #include "iwch_cm.h"
 #include "cxio_hal.h"
+#include "cxio_resource.h"
 
 #define NO_SUPPORT -1
 
-static int iwch_build_rdma_send(union t3_wr *wqe, struct ib_send_wr *wr,
+static int build_rdma_send(union t3_wr *wqe, struct ib_send_wr *wr,
 				u8 * flit_cnt)
 {
 	int i;
@@ -81,7 +82,7 @@ static int iwch_build_rdma_send(union t3_wr *wqe, struct ib_send_wr *wr,
 	return 0;
 }
 
-static int iwch_build_rdma_write(union t3_wr *wqe, struct ib_send_wr *wr,
+static int build_rdma_write(union t3_wr *wqe, struct ib_send_wr *wr,
 				 u8 *flit_cnt)
 {
 	int i;
@@ -122,7 +123,7 @@ static int iwch_build_rdma_write(union t3_wr *wqe, struct ib_send_wr *wr,
 	return 0;
 }
 
-static int iwch_build_rdma_read(union t3_wr *wqe, struct ib_send_wr *wr,
+static int build_rdma_read(union t3_wr *wqe, struct ib_send_wr *wr,
 				u8 *flit_cnt)
 {
 	if (wr->num_sge > 1)
@@ -143,7 +144,7 @@ static int iwch_build_rdma_read(union t3_wr *wqe, struct ib_send_wr *wr,
 	return 0;
 }
 
-static int iwch_build_fastreg(union t3_wr *wqe, struct ib_send_wr *wr,
+static int build_fastreg(union t3_wr *wqe, struct ib_send_wr *wr,
 				u8 *flit_cnt, int *wr_cnt, struct t3_wq *wq)
 {
 	int i;
@@ -185,7 +186,7 @@ static int iwch_build_fastreg(union t3_wr *wqe, struct ib_send_wr *wr,
 	return 0;
 }
 
-static int iwch_build_inv_stag(union t3_wr *wqe, struct ib_send_wr *wr,
+static int build_inv_stag(union t3_wr *wqe, struct ib_send_wr *wr,
 				u8 *flit_cnt)
 {
 	wqe->local_inv.stag = cpu_to_be32(wr->ex.invalidate_rkey);
@@ -244,23 +245,106 @@ static int iwch_sgl2pbl_map(struct iwch_dev *rhp, struct ib_sge *sg_list,
 	return 0;
 }
 
-static int iwch_build_rdma_recv(struct iwch_dev *rhp, union t3_wr *wqe,
+static int build_rdma_recv(struct iwch_qp *qhp, union t3_wr *wqe,
 				struct ib_recv_wr *wr)
 {
-	int i;
-	if (wr->num_sge > T3_MAX_SGE)
-		return -EINVAL;
+	int i, err = 0;
+	u32 pbl_addr[T3_MAX_SGE];
+	u8 page_size[T3_MAX_SGE];
+
+	err = iwch_sgl2pbl_map(qhp->rhp, wr->sg_list, wr->num_sge, pbl_addr,
+			       page_size);
+	if (err)
+		return err;
+	wqe->recv.pagesz[0] = page_size[0];
+	wqe->recv.pagesz[1] = page_size[1];
+	wqe->recv.pagesz[2] = page_size[2];
+	wqe->recv.pagesz[3] = page_size[3];
 	wqe->recv.num_sgle = cpu_to_be32(wr->num_sge);
 	for (i = 0; i < wr->num_sge; i++) {
 		wqe->recv.sgl[i].stag = cpu_to_be32(wr->sg_list[i].lkey);
 		wqe->recv.sgl[i].len = cpu_to_be32(wr->sg_list[i].length);
+
+		/* to in the WQE == the offset into the page */
+		wqe->recv.sgl[i].to = cpu_to_be64(((u32) wr->sg_list[i].addr) %
+				(1UL << (12 + page_size[i])));
+
+		/* pbl_addr is the adapters address in the PBL */
+		wqe->recv.pbl_addr[i] = cpu_to_be32(pbl_addr[i]);
+	}
+	for (; i < T3_MAX_SGE; i++) {
+		wqe->recv.sgl[i].stag = 0;
+		wqe->recv.sgl[i].len = 0;
+		wqe->recv.sgl[i].to = 0;
+		wqe->recv.pbl_addr[i] = 0;
+	}
+	qhp->wq.rq[Q_PTR2IDX(qhp->wq.rq_wptr,
+			     qhp->wq.rq_size_log2)].wr_id = wr->wr_id;
+	qhp->wq.rq[Q_PTR2IDX(qhp->wq.rq_wptr,
+			     qhp->wq.rq_size_log2)].pbl_addr = 0;
+	return 0;
+}
+
+static int build_zero_stag_recv(struct iwch_qp *qhp, union t3_wr *wqe,
+				struct ib_recv_wr *wr)
+{
+	int i;
+	u32 pbl_addr;
+	u32 pbl_offset;
+
+
+	/*
+	 * The T3 HW requires the PBL in the HW recv descriptor to reference
+	 * a PBL entry.  So we allocate the max needed PBL memory here and pass
+	 * it to the uP in the recv WR.  The uP will build the PBL and setup
+	 * the HW recv descriptor.
+	 */
+	pbl_addr = cxio_hal_pblpool_alloc(&qhp->rhp->rdev, T3_STAG0_PBL_SIZE);
+	if (!pbl_addr)
+		return -ENOMEM;
+
+	/*
+	 * Compute the 8B aligned offset.
+	 */
+	pbl_offset = (pbl_addr - qhp->rhp->rdev.rnic_info.pbl_base) >> 3;
+
+	wqe->recv.num_sgle = cpu_to_be32(wr->num_sge);
+
+	for (i = 0; i < wr->num_sge; i++) {
+
+		/*
+		 * Use a 128MB page size. This and an imposed 128MB
+		 * sge length limit allows us to require only a 2-entry HW
+		 * PBL for each SGE.  This restriction is acceptable since
+		 * since it is not possible to allocate 128MB of contiguous
+		 * DMA coherent memory!
+		 */
+		if (wr->sg_list[i].length > T3_STAG0_MAX_PBE_LEN)
+			return -EINVAL;
+		wqe->recv.pagesz[i] = T3_STAG0_PAGE_SHIFT;
+
+		/*
+		 * T3 restricts a recv to all zero-stag or all non-zero-stag.
+		 */
+		if (wr->sg_list[i].lkey != 0)
+			return -EINVAL;
+		wqe->recv.sgl[i].stag = 0;
+		wqe->recv.sgl[i].len = cpu_to_be32(wr->sg_list[i].length);
 		wqe->recv.sgl[i].to = cpu_to_be64(wr->sg_list[i].addr);
+		wqe->recv.pbl_addr[i] = cpu_to_be32(pbl_offset);
+		pbl_offset += 2;
 	}
 	for (; i < T3_MAX_SGE; i++) {
+		wqe->recv.pagesz[i] = 0;
 		wqe->recv.sgl[i].stag = 0;
 		wqe->recv.sgl[i].len = 0;
 		wqe->recv.sgl[i].to = 0;
+		wqe->recv.pbl_addr[i] = 0;
 	}
+	qhp->wq.rq[Q_PTR2IDX(qhp->wq.rq_wptr,
+			     qhp->wq.rq_size_log2)].wr_id = wr->wr_id;
+	qhp->wq.rq[Q_PTR2IDX(qhp->wq.rq_wptr,
+			     qhp->wq.rq_size_log2)].pbl_addr = pbl_addr;
 	return 0;
 }
 
@@ -312,18 +396,18 @@ int iwch_post_send(struct ib_qp *ibqp, struct ib_send_wr *wr,
 			if (wr->send_flags & IB_SEND_FENCE)
 				t3_wr_flags |= T3_READ_FENCE_FLAG;
 			t3_wr_opcode = T3_WR_SEND;
-			err = iwch_build_rdma_send(wqe, wr, &t3_wr_flit_cnt);
+			err = build_rdma_send(wqe, wr, &t3_wr_flit_cnt);
 			break;
 		case IB_WR_RDMA_WRITE:
 		case IB_WR_RDMA_WRITE_WITH_IMM:
 			t3_wr_opcode = T3_WR_WRITE;
-			err = iwch_build_rdma_write(wqe, wr, &t3_wr_flit_cnt);
+			err = build_rdma_write(wqe, wr, &t3_wr_flit_cnt);
 			break;
 		case IB_WR_RDMA_READ:
 		case IB_WR_RDMA_READ_WITH_INV:
 			t3_wr_opcode = T3_WR_READ;
 			t3_wr_flags = 0; /* T3 reads are always signaled */
-			err = iwch_build_rdma_read(wqe, wr, &t3_wr_flit_cnt);
+			err = build_rdma_read(wqe, wr, &t3_wr_flit_cnt);
 			if (err)
 				break;
 			sqp->read_len = wqe->read.local_len;
@@ -332,14 +416,14 @@ int iwch_post_send(struct ib_qp *ibqp, struct ib_send_wr *wr,
 			break;
 		case IB_WR_FAST_REG_MR:
 			t3_wr_opcode = T3_WR_FASTREG;
-			err = iwch_build_fastreg(wqe, wr, &t3_wr_flit_cnt,
+			err = build_fastreg(wqe, wr, &t3_wr_flit_cnt,
 						 &wr_cnt, &qhp->wq);
 			break;
 		case IB_WR_LOCAL_INV:
 			if (wr->send_flags & IB_SEND_FENCE)
 				t3_wr_flags |= T3_LOCAL_FENCE_FLAG;
 			t3_wr_opcode = T3_WR_INV_STAG;
-			err = iwch_build_inv_stag(wqe, wr, &t3_wr_flit_cnt);
+			err = build_inv_stag(wqe, wr, &t3_wr_flit_cnt);
 			break;
 		default:
 			PDBG("%s post of type=%d TBD!\n", __func__,
@@ -398,18 +482,24 @@ int iwch_post_receive(struct ib_qp *ibqp, struct ib_recv_wr *wr,
 		return -EINVAL;
 	}
 	while (wr) {
+		if (wr->num_sge > T3_MAX_SGE) {
+			err = -EINVAL;
+			*bad_wr = wr;
+			break;
+		}
 		idx = Q_PTR2IDX(qhp->wq.wptr, qhp->wq.size_log2);
 		wqe = (union t3_wr *) (qhp->wq.queue + idx);
 		if (num_wrs)
-			err = iwch_build_rdma_recv(qhp->rhp, wqe, wr);
+			if (wr->sg_list[0].lkey)
+				err = build_rdma_recv(qhp, wqe, wr);
+			else
+				err = build_zero_stag_recv(qhp, wqe, wr);
 		else
 			err = -ENOMEM;
 		if (err) {
 			*bad_wr = wr;
 			break;
 		}
-		qhp->wq.rq[Q_PTR2IDX(qhp->wq.rq_wptr, qhp->wq.rq_size_log2)] =
-			wr->wr_id;
 		build_fw_riwrh((void *) wqe, T3_WR_RCV, T3_COMPLETION_FLAG,
 			       Q_GENBIT(qhp->wq.wptr, qhp->wq.size_log2),
 			       0, sizeof(struct t3_receive_wr) >> 3, T3_SOPEOP);
@@ -810,7 +900,8 @@ static int rdma_init(struct iwch_dev *rhp, struct iwch_qp *qhp,
 	init_attr.qp_dma_size = (1UL << qhp->wq.size_log2);
 	init_attr.rqe_count = iwch_rqes_posted(qhp);
 	init_attr.flags = qhp->attr.mpa_attr.initiator ? MPA_INITIATOR : 0;
-	init_attr.flags |= capable(CAP_NET_BIND_SERVICE) ? PRIV_QP : 0;
+	if (!qhp->ibqp.uobject)
+		init_attr.flags |= PRIV_QP;
 	if (peer2peer) {
 		init_attr.rtr_type = RTR_READ;
 		if (init_attr.ord == 0 && qhp->attr.mpa_attr.initiator)
-- 
cgit v1.2.3


From f507d28bff0601f1a8a96b7939fa3855c50d25b6 Mon Sep 17 00:00:00 2001
From: Eli Cohen <eli@mellanox.co.il>
Date: Mon, 14 Jul 2008 23:48:53 -0700
Subject: IB/mlx4: Use kzalloc() for new QPs so flags are initialized to 0

Current code uses kmalloc() and then just does a bitwise OR operation on
qp->flags in create_qp_common(), which means that qp->flags may
potentially have some unintended bits set.  This patch uses kzalloc()
and avoids further explicit clearing of structure members, which also
shrinks the code:

add/remove: 0/0 grow/shrink: 0/1 up/down: 0/-65 (-65)
function                                     old     new   delta
create_qp_common                            2024    1959     -65

Signed-off-by: Eli Cohen <eli@mellanox.co.il>
Signed-off-by: Roland Dreier <rolandd@cisco.com>
---
 drivers/infiniband/hw/mlx4/qp.c | 15 ++-------------
 1 file changed, 2 insertions(+), 13 deletions(-)

(limited to 'drivers/infiniband/hw')

diff --git a/drivers/infiniband/hw/mlx4/qp.c b/drivers/infiniband/hw/mlx4/qp.c
index 91590e7fba0..89eb6cbe592 100644
--- a/drivers/infiniband/hw/mlx4/qp.c
+++ b/drivers/infiniband/hw/mlx4/qp.c
@@ -454,19 +454,8 @@ static int create_qp_common(struct mlx4_ib_dev *dev, struct ib_pd *pd,
 	spin_lock_init(&qp->rq.lock);
 
 	qp->state	 = IB_QPS_RESET;
-	qp->atomic_rd_en = 0;
-	qp->resp_depth   = 0;
-
-	qp->rq.head	    = 0;
-	qp->rq.tail	    = 0;
-	qp->sq.head	    = 0;
-	qp->sq.tail	    = 0;
-	qp->sq_next_wqe     = 0;
-
 	if (init_attr->sq_sig_type == IB_SIGNAL_ALL_WR)
 		qp->sq_signal_bits = cpu_to_be32(MLX4_WQE_CTRL_CQ_UPDATE);
-	else
-		qp->sq_signal_bits = 0;
 
 	err = set_rq_size(dev, &init_attr->cap, !!pd->uobject, !!init_attr->srq, qp);
 	if (err)
@@ -704,7 +693,7 @@ struct ib_qp *mlx4_ib_create_qp(struct ib_pd *pd,
 	case IB_QPT_UC:
 	case IB_QPT_UD:
 	{
-		qp = kmalloc(sizeof *qp, GFP_KERNEL);
+		qp = kzalloc(sizeof *qp, GFP_KERNEL);
 		if (!qp)
 			return ERR_PTR(-ENOMEM);
 
@@ -725,7 +714,7 @@ struct ib_qp *mlx4_ib_create_qp(struct ib_pd *pd,
 		if (pd->uobject)
 			return ERR_PTR(-EINVAL);
 
-		sqp = kmalloc(sizeof *sqp, GFP_KERNEL);
+		sqp = kzalloc(sizeof *sqp, GFP_KERNEL);
 		if (!sqp)
 			return ERR_PTR(-ENOMEM);
 
-- 
cgit v1.2.3