[PATCH 0 of 23] IB/ipath - updates for 2.6.19

[PATCH 0 of 23] IB/ipath - updates for 2.6.19

[Bryan O'Sullivan]

Hi, Roland -

This is a series of patches to bring the ipath driver up to date for 2.6.19.
The patches apply on top of Ralph's mmap patch that you accepted yesterday.

Please apply.

Thanks,

	<b

[PATCH 1 of 23] IB/ipath - More changes to support InfiniPath on PowerPC 970 systems

[Bryan O'Sullivan]

Ordering of writethrough store buffers needs to be forced, and we need
to use ifdef to get writethrough behavior to InfiniPath buffers, because
there is no generic way to specify that at this time (similar to code
in char/drm/drm_vm.c and block/z2ram.c).

Signed-off-by: John Gregor <john.gregor@qlogic.com>
Signed-off-by: Bryan O'Sullivan <bryan.osullivan@qlogic.com>

diff --git a/drivers/infiniband/hw/ipath/Makefile b/drivers/infiniband/hw/ipath/Makefile
--- a/drivers/infiniband/hw/ipath/Makefile	Fri Aug 25 11:19:44 2006 -0700
+++ b/drivers/infiniband/hw/ipath/Makefile	Fri Aug 25 11:19:44 2006 -0700
@@ -20,6 +20,7 @@ ipath_core-y := \
 	ipath_user_pages.o
 
 ipath_core-$(CONFIG_X86_64) += ipath_wc_x86_64.o
+ipath_core-$(CONFIG_PPC64) += ipath_wc_ppc64.o
 
 ib_ipath-y := \
 	ipath_cq.o \
diff --git a/drivers/infiniband/hw/ipath/ipath_driver.c b/drivers/infiniband/hw/ipath/ipath_driver.c
--- a/drivers/infiniband/hw/ipath/ipath_driver.c	Fri Aug 25 11:19:44 2006 -0700
+++ b/drivers/infiniband/hw/ipath/ipath_driver.c	Fri Aug 25 11:19:44 2006 -0700
@@ -440,7 +440,13 @@ static int __devinit ipath_init_one(stru
 	}
 	dd->ipath_pcirev = rev;
 
+#if defined(__powerpc__)
+	/* There isn't a generic way to specify writethrough mappings */
+	dd->ipath_kregbase = __ioremap(addr, len,
+		(_PAGE_NO_CACHE|_PAGE_WRITETHRU));
+#else
 	dd->ipath_kregbase = ioremap_nocache(addr, len);
+#endif
 
 	if (!dd->ipath_kregbase) {
 		ipath_dbg("Unable to map io addr %llx to kvirt, failing\n",
diff --git a/drivers/infiniband/hw/ipath/ipath_file_ops.c b/drivers/infiniband/hw/ipath/ipath_file_ops.c
--- a/drivers/infiniband/hw/ipath/ipath_file_ops.c	Fri Aug 25 11:19:44 2006 -0700
+++ b/drivers/infiniband/hw/ipath/ipath_file_ops.c	Fri Aug 25 11:19:44 2006 -0700
@@ -985,6 +985,13 @@ static int mmap_piobufs(struct vm_area_s
 	 * write combining behavior we want on the PIO buffers!
 	 */
 
+#if defined(__powerpc__)
+	/* There isn't a generic way to specify writethrough mappings */
+	pgprot_val(vma->vm_page_prot) |= _PAGE_NO_CACHE;
+	pgprot_val(vma->vm_page_prot) |= _PAGE_WRITETHRU;
+	pgprot_val(vma->vm_page_prot) &= ~_PAGE_GUARDED;
+#endif
+
 	if (vma->vm_flags & VM_READ) {
 		dev_info(&dd->pcidev->dev,
 			 "Can't map piobufs as readable (flags=%lx)\n",
diff --git a/drivers/infiniband/hw/ipath/ipath_wc_ppc64.c b/drivers/infiniband/hw/ipath/ipath_wc_ppc64.c
new file mode 100644
--- /dev/null	Thu Jan 01 00:00:00 1970 +0000
+++ b/drivers/infiniband/hw/ipath/ipath_wc_ppc64.c	Fri Aug 25 11:19:44 2006 -0700
@@ -0,0 +1,52 @@
+/*
+ * Copyright (c) 2006 QLogic, Inc. All rights reserved.
+ *
+ * This software is available to you under a choice of one of two
+ * licenses.  You may choose to be licensed under the terms of the GNU
+ * General Public License (GPL) Version 2, available from the file
+ * COPYING in the main directory of this source tree, or the
+ * OpenIB.org BSD license below:
+ *
+ *     Redistribution and use in source and binary forms, with or
+ *     without modification, are permitted provided that the following
+ *     conditions are met:
+ *
+ *      - Redistributions of source code must retain the above
+ *        copyright notice, this list of conditions and the following
+ *        disclaimer.
+ *
+ *      - Redistributions in binary form must reproduce the above
+ *        copyright notice, this list of conditions and the following
+ *        disclaimer in the documentation and/or other materials
+ *        provided with the distribution.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
+ * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
+ * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
+ * 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.
+ */
+
+/*
+ * This file is conditionally built on PowerPC only.  Otherwise weak symbol
+ * versions of the functions exported from here are used.
+ */
+
+#include "ipath_kernel.h"
+
+/**
+ * ipath_unordered_wc - indicate whether write combining is ordered
+ *
+ * PowerPC systems (at least those in the 970 processor family)
+ * write partially filled store buffers in address order, but will write
+ * completely filled store buffers in "random" order, and therefore must
+ * have serialization for correctness with current InfiniPath chips.
+ *
+ */
+int ipath_unordered_wc(void)
+{
+	return 1;
+}

Re: [PATCH 1 of 23] IB/ipath - More changes to support InfiniPath on PowerPC 970 systems

[Roland Dreier]

How did you generate these patches?  When I try to apply them with
git, I get errors like

    error: drivers/infiniband/hw/ipath/Makefile	Fri Aug 25 11:19:44 2006 -0700: No such file or directory

because the line

diff --git a/drivers/infiniband/hw/ipath/Makefile b/drivers/infiniband/hw/ipath/Makefile

makes git think it's a git diff, but git doesn't put dates on the
filename lines.  In other words, instead of

--- a/drivers/infiniband/hw/ipath/Makefile	Fri Aug 25 11:19:44 2006 -0700
+++ b/drivers/infiniband/hw/ipath/Makefile	Fri Aug 25 11:19:44 2006 -0700

the patch should just have

--- a/drivers/infiniband/hw/ipath/Makefile
+++ b/drivers/infiniband/hw/ipath/Makefile

before the Makefile chunks.

I fixed this up by deleting the "diff --git" lines, but I'm curious
how you created this in the first place.

 - R.

Re: [PATCH 1 of 23] IB/ipath - More changes to support InfiniPath on PowerPC 970 systems

[Bryan O'Sullivan]

On Fri, 2006-08-25 at 12:45 -0700, Roland Dreier wrote:
> How did you generate these patches?

Using Mercurial.  

> because the line
> 
> diff --git a/drivers/infiniband/hw/ipath/Makefile b/drivers/infiniband/hw/ipath/Makefile
> 
> makes git think it's a git diff, but git doesn't put dates on the
> filename lines.

Ah, interesting.  Looks like a bug in the git-compatible patch
generator, then.  Sorry about that.

	<b

Re: [PATCH 1 of 23] IB/ipath - More changes to support InfiniPath on PowerPC 970 systems

[Brendan Cully]

On Friday, 25 August 2006 at 13:19, Bryan O'Sullivan wrote:
> On Fri, 2006-08-25 at 12:45 -0700, Roland Dreier wrote:
> > How did you generate these patches?
> 
> Using Mercurial.  
> 
> > because the line
> > 
> > diff --git a/drivers/infiniband/hw/ipath/Makefile b/drivers/infiniband/hw/ipath/Makefile
> > 
> > makes git think it's a git diff, but git doesn't put dates on the
> > filename lines.
> 
> Ah, interesting.  Looks like a bug in the git-compatible patch
> generator, then.  Sorry about that.

I've just posted a fix to the mercurial list.

Re: [PATCH 1 of 23] IB/ipath - More changes to support InfiniPath on PowerPC 970 systems

[Roland Dreier]

 > Signed-off-by: John Gregor <john.gregor@qlogic.com>

I assume this patch was actually written by John Gregor?  If so you
should include an extra "From:" line in the body of the email, so that
the authorship information gets put into git correctly.

 - R.

[PATCH 2 of 23] IB/ipath - lock resource limit counters correctly

[Bryan O'Sullivan]

Signed-off-by: Bryan O'Sullivan <bryan.osullivan@qlogic.com>

diff --git a/drivers/infiniband/hw/ipath/ipath_verbs.c b/drivers/infiniband/hw/ipath/ipath_verbs.c
--- a/drivers/infiniband/hw/ipath/ipath_verbs.c	Fri Aug 25 11:19:44 2006 -0700
+++ b/drivers/infiniband/hw/ipath/ipath_verbs.c	Fri Aug 25 11:19:44 2006 -0700
@@ -776,18 +776,22 @@ static struct ib_pd *ipath_alloc_pd(stru
 	 * we allow allocations of more than we report for this value.
 	 */
 
-	if (dev->n_pds_allocated == ib_ipath_max_pds) {
-		ret = ERR_PTR(-ENOMEM);
-		goto bail;
-	}
-
 	pd = kmalloc(sizeof *pd, GFP_KERNEL);
 	if (!pd) {
 		ret = ERR_PTR(-ENOMEM);
 		goto bail;
 	}
 
+	spin_lock(&dev->n_pds_lock);
+	if (dev->n_pds_allocated == ib_ipath_max_pds) {
+		spin_unlock(&dev->n_pds_lock);
+		kfree(pd);
+		ret = ERR_PTR(-ENOMEM);
+		goto bail;
+	}
+
 	dev->n_pds_allocated++;
+	spin_unlock(&dev->n_pds_lock);
 
 	/* ib_alloc_pd() will initialize pd->ibpd. */
 	pd->user = udata != NULL;
@@ -803,7 +807,9 @@ static int ipath_dealloc_pd(struct ib_pd
 	struct ipath_pd *pd = to_ipd(ibpd);
 	struct ipath_ibdev *dev = to_idev(ibpd->device);
 
+	spin_lock(&dev->n_pds_lock);
 	dev->n_pds_allocated--;
+	spin_unlock(&dev->n_pds_lock);
 
 	kfree(pd);
 
@@ -824,11 +830,6 @@ static struct ib_ah *ipath_create_ah(str
 	struct ib_ah *ret;
 	struct ipath_ibdev *dev = to_idev(pd->device);
 
-	if (dev->n_ahs_allocated == ib_ipath_max_ahs) {
-		ret = ERR_PTR(-ENOMEM);
-		goto bail;
-	}
-
 	/* A multicast address requires a GRH (see ch. 8.4.1). */
 	if (ah_attr->dlid >= IPATH_MULTICAST_LID_BASE &&
 	    ah_attr->dlid != IPATH_PERMISSIVE_LID &&
@@ -854,7 +855,16 @@ static struct ib_ah *ipath_create_ah(str
 		goto bail;
 	}
 
+	spin_lock(&dev->n_ahs_lock);
+	if (dev->n_ahs_allocated == ib_ipath_max_ahs) {
+		spin_unlock(&dev->n_ahs_lock);
+		kfree(ah);
+		ret = ERR_PTR(-ENOMEM);
+		goto bail;
+	}
+
 	dev->n_ahs_allocated++;
+	spin_unlock(&dev->n_ahs_lock);
 
 	/* ib_create_ah() will initialize ah->ibah. */
 	ah->attr = *ah_attr;
@@ -876,7 +886,9 @@ static int ipath_destroy_ah(struct ib_ah
 	struct ipath_ibdev *dev = to_idev(ibah->device);
 	struct ipath_ah *ah = to_iah(ibah);
 
+	spin_lock(&dev->n_ahs_lock);
 	dev->n_ahs_allocated--;
+	spin_unlock(&dev->n_ahs_lock);
 
 	kfree(ah);
 
@@ -963,6 +975,12 @@ static void *ipath_register_ib_device(in
 	dev = &idev->ibdev;
 
 	/* Only need to initialize non-zero fields. */
+	spin_lock_init(&idev->n_pds_lock);
+	spin_lock_init(&idev->n_ahs_lock);
+	spin_lock_init(&idev->n_cqs_lock);
+	spin_lock_init(&idev->n_srqs_lock);
+	spin_lock_init(&idev->n_mcast_grps_lock);
+
 	spin_lock_init(&idev->qp_table.lock);
 	spin_lock_init(&idev->lk_table.lock);
 	idev->sm_lid = __constant_be16_to_cpu(IB_LID_PERMISSIVE);
diff --git a/drivers/infiniband/hw/ipath/ipath_verbs.h b/drivers/infiniband/hw/ipath/ipath_verbs.h
--- a/drivers/infiniband/hw/ipath/ipath_verbs.h	Fri Aug 25 11:19:44 2006 -0700
+++ b/drivers/infiniband/hw/ipath/ipath_verbs.h	Fri Aug 25 11:19:44 2006 -0700
@@ -434,11 +434,18 @@ struct ipath_ibdev {
 	__be64 sys_image_guid;	/* in network order */
 	__be64 gid_prefix;	/* in network order */
 	__be64 mkey;
+
 	u32 n_pds_allocated;	/* number of PDs allocated for device */
+	spinlock_t n_pds_lock;
 	u32 n_ahs_allocated;	/* number of AHs allocated for device */
+	spinlock_t n_ahs_lock;
 	u32 n_cqs_allocated;	/* number of CQs allocated for device */
+	spinlock_t n_cqs_lock;
 	u32 n_srqs_allocated;	/* number of SRQs allocated for device */
+	spinlock_t n_srqs_lock;
 	u32 n_mcast_grps_allocated; /* number of mcast groups allocated */
+	spinlock_t n_mcast_grps_lock;
+
 	u64 ipath_sword;	/* total dwords sent (sample result) */
 	u64 ipath_rword;	/* total dwords received (sample result) */
 	u64 ipath_spkts;	/* total packets sent (sample result) */
diff --git a/drivers/infiniband/hw/ipath/ipath_verbs_mcast.c b/drivers/infiniband/hw/ipath/ipath_verbs_mcast.c
--- a/drivers/infiniband/hw/ipath/ipath_verbs_mcast.c	Fri Aug 25 11:19:44 2006 -0700
+++ b/drivers/infiniband/hw/ipath/ipath_verbs_mcast.c	Fri Aug 25 11:19:44 2006 -0700
@@ -207,12 +207,15 @@ static int ipath_mcast_add(struct ipath_
 		goto bail;
 	}
 
+	spin_lock(&dev->n_mcast_grps_lock);
 	if (dev->n_mcast_grps_allocated == ib_ipath_max_mcast_grps) {
+		spin_unlock(&dev->n_mcast_grps_lock);
 		ret = ENOMEM;
 		goto bail;
 	}
 
 	dev->n_mcast_grps_allocated++;
+	spin_unlock(&dev->n_mcast_grps_lock);
 
 	list_add_tail_rcu(&mqp->list, &mcast->qp_list);
 
@@ -343,7 +346,9 @@ int ipath_multicast_detach(struct ib_qp 
 		atomic_dec(&mcast->refcount);
 		wait_event(mcast->wait, !atomic_read(&mcast->refcount));
 		ipath_mcast_free(mcast);
+		spin_lock(&dev->n_mcast_grps_lock);
 		dev->n_mcast_grps_allocated--;
+		spin_unlock(&dev->n_mcast_grps_lock);
 	}
 
 	ret = 0;

[PATCH 3 of 23] IB/ipath - fix for crash on module unload, if cfgports < portcnt

[Bryan O'Sullivan]

Allocate enough pointers for all possible ports, to avoid problems in
cleanup/unload.

Signed-off-by: Bryan O'Sullivan <bryan.osullivan@qlogic.com>

diff --git a/drivers/infiniband/hw/ipath/ipath_init_chip.c b/drivers/infiniband/hw/ipath/ipath_init_chip.c
--- a/drivers/infiniband/hw/ipath/ipath_init_chip.c	Fri Aug 25 11:19:44 2006 -0700
+++ b/drivers/infiniband/hw/ipath/ipath_init_chip.c	Fri Aug 25 11:19:44 2006 -0700
@@ -240,7 +240,11 @@ static int init_chip_first(struct ipath_
 			  "only supports %u\n", ipath_cfgports,
 			  dd->ipath_portcnt);
 	}
-	dd->ipath_pd = kzalloc(sizeof(*dd->ipath_pd) * dd->ipath_cfgports,
+	/*
+	 * Allocate full portcnt array, rather than just cfgports, because
+	 * cleanup iterates across all possible ports.
+	 */
+	dd->ipath_pd = kzalloc(sizeof(*dd->ipath_pd) * dd->ipath_portcnt,
 			       GFP_KERNEL);
 
 	if (!dd->ipath_pd) {

[PATCH 4 of 23] IB/ipath - fix handling of kpiobufs

[Bryan O'Sullivan]

Change comment: no longer imply that user can set ipath_kpiobufs to zero.
Actually set ipath_kpiobufs from parameter. Previously only altered
per-device ipath_lastport_piobuf, which was over-written in chip init.

Signed-off-by: Bryan O'Sullivan <bryan.osullivan@qlogic.com>

diff --git a/drivers/infiniband/hw/ipath/ipath_init_chip.c b/drivers/infiniband/hw/ipath/ipath_init_chip.c
--- a/drivers/infiniband/hw/ipath/ipath_init_chip.c	Fri Aug 25 11:19:44 2006 -0700
+++ b/drivers/infiniband/hw/ipath/ipath_init_chip.c	Fri Aug 25 11:19:44 2006 -0700
@@ -691,7 +691,7 @@ int ipath_init_chip(struct ipath_devdata
 	dd->ipath_pioavregs = ALIGN(val, sizeof(u64) * BITS_PER_BYTE / 2)
 		/ (sizeof(u64) * BITS_PER_BYTE / 2);
 	if (ipath_kpiobufs == 0) {
-		/* not set by user, or set explictly to default  */
+		/* not set by user (this is default) */
 		if ((dd->ipath_piobcnt2k + dd->ipath_piobcnt4k) > 128)
 			kpiobufs = 32;
 		else
@@ -950,6 +950,7 @@ static int ipath_set_kpiobufs(const char
 			dd->ipath_piobcnt2k + dd->ipath_piobcnt4k - val;
 	}
 
+	ipath_kpiobufs = val;
 	ret = 0;
 bail:
 	spin_unlock_irqrestore(&ipath_devs_lock, flags);

[PATCH 5 of 23] IB/ipath - drop requirement that PIO buffers be mmaped write-only

[Bryan O'Sullivan]

Some userlands try to mmap these pages read-write, so accommodate them.

Signed-off-by: Bryan O'Sullivan <bryan.osullivan@qlogic.com>

diff --git a/drivers/infiniband/hw/ipath/ipath_file_ops.c b/drivers/infiniband/hw/ipath/ipath_file_ops.c
--- a/drivers/infiniband/hw/ipath/ipath_file_ops.c	Fri Aug 25 11:19:44 2006 -0700
+++ b/drivers/infiniband/hw/ipath/ipath_file_ops.c	Fri Aug 25 11:19:44 2006 -0700
@@ -992,15 +992,10 @@ static int mmap_piobufs(struct vm_area_s
 	pgprot_val(vma->vm_page_prot) &= ~_PAGE_GUARDED;
 #endif
 
-	if (vma->vm_flags & VM_READ) {
-		dev_info(&dd->pcidev->dev,
-			 "Can't map piobufs as readable (flags=%lx)\n",
-			 vma->vm_flags);
-		ret = -EPERM;
-		goto bail;
-	}
-
-	/* don't allow them to later change to readable with mprotect */
+	/*
+	 * don't allow them to later change to readable with mprotect (for when
+	 * not initially mapped readable, as is normally the case)
+	 */
 	vma->vm_flags &= ~VM_MAYREAD;
 	vma->vm_flags |= VM_DONTCOPY | VM_DONTEXPAND;
 

[PATCH 6 of 23] IB/ipath - merge ipath_core and ib_ipath drivers

[Bryan O'Sullivan]

There is little point in keeping the two drivers separate, so we are
merging them.

Signed-off-by: Bryan O'Sullivan <bryan.osullivan@qlogic.com>

diff --git a/drivers/infiniband/Makefile b/drivers/infiniband/Makefile
--- a/drivers/infiniband/Makefile	Fri Aug 25 11:19:44 2006 -0700
+++ b/drivers/infiniband/Makefile	Fri Aug 25 11:19:45 2006 -0700
@@ -1,6 +1,6 @@ obj-$(CONFIG_INFINIBAND)		+= core/
 obj-$(CONFIG_INFINIBAND)		+= core/
 obj-$(CONFIG_INFINIBAND_MTHCA)		+= hw/mthca/
-obj-$(CONFIG_IPATH_CORE)		+= hw/ipath/
+obj-$(CONFIG_INFINIBAND_IPATH)		+= hw/ipath/
 obj-$(CONFIG_INFINIBAND_IPOIB)		+= ulp/ipoib/
 obj-$(CONFIG_INFINIBAND_SRP)		+= ulp/srp/
 obj-$(CONFIG_INFINIBAND_ISER)		+= ulp/iser/
diff --git a/drivers/infiniband/hw/ipath/Kconfig b/drivers/infiniband/hw/ipath/Kconfig
--- a/drivers/infiniband/hw/ipath/Kconfig	Fri Aug 25 11:19:44 2006 -0700
+++ b/drivers/infiniband/hw/ipath/Kconfig	Fri Aug 25 11:19:45 2006 -0700
@@ -1,16 +1,9 @@ config IPATH_CORE
-config IPATH_CORE
+config INFINIBAND_IPATH
 	tristate "QLogic InfiniPath Driver"
-	depends on 64BIT && PCI_MSI && NET
+	depends on PCI_MSI && 64BIT && INFINIBAND
 	---help---
-	This is a low-level driver for QLogic InfiniPath host channel
-	adapters (HCAs) based on the HT-400 and PE-800 chips.
-
-config INFINIBAND_IPATH
-	tristate "QLogic InfiniPath Verbs Driver"
-	depends on IPATH_CORE && INFINIBAND
-	---help---
-	This is a driver that provides InfiniBand verbs support for
-	QLogic InfiniPath host channel adapters (HCAs).  This
-	allows these devices to be used with both kernel upper level
-	protocols such as IP-over-InfiniBand as well as with userspace
-	applications (in conjunction with InfiniBand userspace access).
+	This is a driver for QLogic InfiniPath host channel adapters,
+	including InfiniBand verbs support.  This driver allows these
+	devices to be used with both kernel upper level protocols such
+	as IP-over-InfiniBand as well as with userspace applications
+	(in conjunction with InfiniBand userspace access).
diff --git a/drivers/infiniband/hw/ipath/Makefile b/drivers/infiniband/hw/ipath/Makefile
--- a/drivers/infiniband/hw/ipath/Makefile	Fri Aug 25 11:19:44 2006 -0700
+++ b/drivers/infiniband/hw/ipath/Makefile	Fri Aug 25 11:19:45 2006 -0700
@@ -1,10 +1,10 @@ EXTRA_CFLAGS += -DIPATH_IDSTR='"QLogic k
 EXTRA_CFLAGS += -DIPATH_IDSTR='"QLogic kernel.org driver"' \
 	-DIPATH_KERN_TYPE=0
 
-obj-$(CONFIG_IPATH_CORE) += ipath_core.o
 obj-$(CONFIG_INFINIBAND_IPATH) += ib_ipath.o
 
-ipath_core-y := \
+ib_ipath-y := \
+	ipath_cq.o \
 	ipath_diag.o \
 	ipath_driver.o \
 	ipath_eeprom.o \
@@ -13,26 +13,23 @@ ipath_core-y := \
 	ipath_ht400.o \
 	ipath_init_chip.o \
 	ipath_intr.o \
+	ipath_keys.o \
 	ipath_layer.o \
-	ipath_pe800.o \
-	ipath_stats.o \
-	ipath_sysfs.o \
-	ipath_user_pages.o
-
-ipath_core-$(CONFIG_X86_64) += ipath_wc_x86_64.o
-ipath_core-$(CONFIG_PPC64) += ipath_wc_ppc64.o
-
-ib_ipath-y := \
-	ipath_cq.o \
-	ipath_keys.o \
 	ipath_mad.o \
 	ipath_mmap.o \
 	ipath_mr.o \
+	ipath_pe800.o \
 	ipath_qp.o \
 	ipath_rc.o \
 	ipath_ruc.o \
 	ipath_srq.o \
+	ipath_stats.o \
+	ipath_sysfs.o \
 	ipath_uc.o \
 	ipath_ud.o \
-	ipath_verbs.o \
-	ipath_verbs_mcast.o
+	ipath_user_pages.o \
+	ipath_verbs_mcast.o \
+	ipath_verbs.o
+
+ib_ipath-$(CONFIG_X86_64) += ipath_wc_x86_64.o
+ib_ipath-$(CONFIG_PPC64) += ipath_wc_ppc64.o
diff --git a/drivers/infiniband/hw/ipath/ipath_driver.c b/drivers/infiniband/hw/ipath/ipath_driver.c
--- a/drivers/infiniband/hw/ipath/ipath_driver.c	Fri Aug 25 11:19:44 2006 -0700
+++ b/drivers/infiniband/hw/ipath/ipath_driver.c	Fri Aug 25 11:19:45 2006 -0700
@@ -40,6 +40,7 @@
 
 #include "ipath_kernel.h"
 #include "ipath_layer.h"
+#include "ipath_verbs.h"
 #include "ipath_common.h"
 
 static void ipath_update_pio_bufs(struct ipath_devdata *);
@@ -50,8 +51,6 @@ const char *ipath_get_unit_name(int unit
 	snprintf(iname, sizeof iname, "infinipath%u", unit);
 	return iname;
 }
-
-EXPORT_SYMBOL_GPL(ipath_get_unit_name);
 
 #define DRIVER_LOAD_MSG "QLogic " IPATH_DRV_NAME " loaded: "
 #define PFX IPATH_DRV_NAME ": "
@@ -510,6 +509,7 @@ static int __devinit ipath_init_one(stru
 	ipath_user_add(dd);
 	ipath_diag_add(dd);
 	ipath_layer_add(dd);
+	ipath_register_ib_device(dd);
 
 	goto bail;
 
@@ -538,6 +538,7 @@ static void __devexit ipath_remove_one(s
 		return;
 
 	dd = pci_get_drvdata(pdev);
+	ipath_unregister_ib_device(dd->verbs_dev);
 	ipath_layer_remove(dd);
 	ipath_diag_remove(dd);
 	ipath_user_remove(dd);
@@ -978,12 +979,8 @@ reloop:
 		if (unlikely(eflags))
 			ipath_rcv_hdrerr(dd, eflags, l, etail, rc);
 		else if (etype == RCVHQ_RCV_TYPE_NON_KD) {
-				int ret = __ipath_verbs_rcv(dd, rc + 1,
-							    ebuf, tlen);
-				if (ret == -ENODEV)
-					ipath_cdbg(VERBOSE,
-						   "received IB packet, "
-						   "not SMA (QP=%x)\n", qp);
+				ipath_ib_rcv(dd->verbs_dev, rc + 1, ebuf,
+					     tlen);
 				if (dd->ipath_lli_counter)
 					dd->ipath_lli_counter--;
 
diff --git a/drivers/infiniband/hw/ipath/ipath_intr.c b/drivers/infiniband/hw/ipath/ipath_intr.c
--- a/drivers/infiniband/hw/ipath/ipath_intr.c	Fri Aug 25 11:19:44 2006 -0700
+++ b/drivers/infiniband/hw/ipath/ipath_intr.c	Fri Aug 25 11:19:45 2006 -0700
@@ -35,6 +35,7 @@
 
 #include "ipath_kernel.h"
 #include "ipath_layer.h"
+#include "ipath_verbs.h"
 #include "ipath_common.h"
 
 /* These are all rcv-related errors which we want to count for stats */
@@ -712,7 +713,7 @@ static void handle_layer_pioavail(struct
 	if (ret > 0)
 		goto set;
 
-	ret = __ipath_verbs_piobufavail(dd);
+	ret = ipath_ib_piobufavail(dd->verbs_dev);
 	if (ret > 0)
 		goto set;
 
diff --git a/drivers/infiniband/hw/ipath/ipath_kernel.h b/drivers/infiniband/hw/ipath/ipath_kernel.h
--- a/drivers/infiniband/hw/ipath/ipath_kernel.h	Fri Aug 25 11:19:44 2006 -0700
+++ b/drivers/infiniband/hw/ipath/ipath_kernel.h	Fri Aug 25 11:19:45 2006 -0700
@@ -132,12 +132,6 @@ struct _ipath_layer {
 	void *l_arg;
 };
 
-/* Verbs layer interface */
-struct _verbs_layer {
-	void *l_arg;
-	struct timer_list l_timer;
-};
-
 struct ipath_devdata {
 	struct list_head ipath_list;
 
@@ -198,7 +192,8 @@ struct ipath_devdata {
 	void (*ipath_f_setextled)(struct ipath_devdata *, u64, u64);
 	/* fill out chip-specific fields */
 	int (*ipath_f_get_base_info)(struct ipath_portdata *, void *);
-	struct _verbs_layer verbs_layer;
+	struct ipath_ibdev *verbs_dev;
+	struct timer_list verbs_timer;
 	/* total dwords sent (summed from counter) */
 	u64 ipath_sword;
 	/* total dwords rcvd (summed from counter) */
@@ -529,8 +524,6 @@ extern int __ipath_layer_rcv(struct ipat
 extern int __ipath_layer_rcv(struct ipath_devdata *, void *,
 			     struct sk_buff *);
 extern int __ipath_layer_rcv_lid(struct ipath_devdata *, void *);
-extern int __ipath_verbs_piobufavail(struct ipath_devdata *);
-extern int __ipath_verbs_rcv(struct ipath_devdata *, void *, void *, u32);
 
 void ipath_layer_add(struct ipath_devdata *);
 void ipath_layer_remove(struct ipath_devdata *);
diff --git a/drivers/infiniband/hw/ipath/ipath_layer.c b/drivers/infiniband/hw/ipath/ipath_layer.c
--- a/drivers/infiniband/hw/ipath/ipath_layer.c	Fri Aug 25 11:19:44 2006 -0700
+++ b/drivers/infiniband/hw/ipath/ipath_layer.c	Fri Aug 25 11:19:45 2006 -0700
@@ -42,26 +42,20 @@
 
 #include "ipath_kernel.h"
 #include "ipath_layer.h"
+#include "ipath_verbs.h"
 #include "ipath_common.h"
 
 /* Acquire before ipath_devs_lock. */
 static DEFINE_MUTEX(ipath_layer_mutex);
-
-static int ipath_verbs_registered;
 
 u16 ipath_layer_rcv_opcode;
 
 static int (*layer_intr)(void *, u32);
 static int (*layer_rcv)(void *, void *, struct sk_buff *);
 static int (*layer_rcv_lid)(void *, void *);
-static int (*verbs_piobufavail)(void *);
-static void (*verbs_rcv)(void *, void *, void *, u32);
 
 static void *(*layer_add_one)(int, struct ipath_devdata *);
 static void (*layer_remove_one)(void *);
-static void *(*verbs_add_one)(int, struct ipath_devdata *);
-static void (*verbs_remove_one)(void *);
-static void (*verbs_timer_cb)(void *);
 
 int __ipath_layer_intr(struct ipath_devdata *dd, u32 arg)
 {
@@ -103,29 +97,6 @@ int __ipath_layer_rcv_lid(struct ipath_d
 
 	if (dd->ipath_layer.l_arg && layer_rcv_lid)
 		ret = layer_rcv_lid(dd->ipath_layer.l_arg, hdr);
-
-	return ret;
-}
-
-int __ipath_verbs_piobufavail(struct ipath_devdata *dd)
-{
-	int ret = -ENODEV;
-
-	if (dd->verbs_layer.l_arg && verbs_piobufavail)
-		ret = verbs_piobufavail(dd->verbs_layer.l_arg);
-
-	return ret;
-}
-
-int __ipath_verbs_rcv(struct ipath_devdata *dd, void *rc, void *ebuf,
-		      u32 tlen)
-{
-	int ret = -ENODEV;
-
-	if (dd->verbs_layer.l_arg && verbs_rcv) {
-		verbs_rcv(dd->verbs_layer.l_arg, rc, ebuf, tlen);
-		ret = 0;
-	}
 
 	return ret;
 }
@@ -211,8 +182,6 @@ bail:
 bail:
 	return ret;
 }
-
-EXPORT_SYMBOL_GPL(ipath_layer_set_linkstate);
 
 /**
  * ipath_layer_set_mtu - set the MTU
@@ -298,8 +267,6 @@ bail:
 	return ret;
 }
 
-EXPORT_SYMBOL_GPL(ipath_layer_set_mtu);
-
 int ipath_set_lid(struct ipath_devdata *dd, u32 arg, u8 lmc)
 {
 	dd->ipath_lid = arg;
@@ -315,8 +282,6 @@ int ipath_set_lid(struct ipath_devdata *
 	return 0;
 }
 
-EXPORT_SYMBOL_GPL(ipath_set_lid);
-
 int ipath_layer_set_guid(struct ipath_devdata *dd, __be64 guid)
 {
 	/* XXX - need to inform anyone who cares this just happened. */
@@ -324,84 +289,55 @@ int ipath_layer_set_guid(struct ipath_de
 	return 0;
 }
 
-EXPORT_SYMBOL_GPL(ipath_layer_set_guid);
-
 __be64 ipath_layer_get_guid(struct ipath_devdata *dd)
 {
 	return dd->ipath_guid;
 }
 
-EXPORT_SYMBOL_GPL(ipath_layer_get_guid);
-
-u32 ipath_layer_get_nguid(struct ipath_devdata *dd)
-{
-	return dd->ipath_nguid;
-}
-
-EXPORT_SYMBOL_GPL(ipath_layer_get_nguid);
-
 u32 ipath_layer_get_majrev(struct ipath_devdata *dd)
 {
 	return dd->ipath_majrev;
 }
 
-EXPORT_SYMBOL_GPL(ipath_layer_get_majrev);
-
 u32 ipath_layer_get_minrev(struct ipath_devdata *dd)
 {
 	return dd->ipath_minrev;
 }
 
-EXPORT_SYMBOL_GPL(ipath_layer_get_minrev);
-
 u32 ipath_layer_get_pcirev(struct ipath_devdata *dd)
 {
 	return dd->ipath_pcirev;
 }
 
-EXPORT_SYMBOL_GPL(ipath_layer_get_pcirev);
-
 u32 ipath_layer_get_flags(struct ipath_devdata *dd)
 {
 	return dd->ipath_flags;
 }
 
-EXPORT_SYMBOL_GPL(ipath_layer_get_flags);
-
 struct device *ipath_layer_get_device(struct ipath_devdata *dd)
 {
 	return &dd->pcidev->dev;
 }
 
-EXPORT_SYMBOL_GPL(ipath_layer_get_device);
-
 u16 ipath_layer_get_deviceid(struct ipath_devdata *dd)
 {
 	return dd->ipath_deviceid;
 }
 
-EXPORT_SYMBOL_GPL(ipath_layer_get_deviceid);
-
 u32 ipath_layer_get_vendorid(struct ipath_devdata *dd)
 {
 	return dd->ipath_vendorid;
 }
 
-EXPORT_SYMBOL_GPL(ipath_layer_get_vendorid);
-
 u64 ipath_layer_get_lastibcstat(struct ipath_devdata *dd)
 {
 	return dd->ipath_lastibcstat;
 }
 
-EXPORT_SYMBOL_GPL(ipath_layer_get_lastibcstat);
-
 u32 ipath_layer_get_ibmtu(struct ipath_devdata *dd)
 {
 	return dd->ipath_ibmtu;
 }
-
-EXPORT_SYMBOL_GPL(ipath_layer_get_ibmtu);
 
 void ipath_layer_add(struct ipath_devdata *dd)
 {
@@ -411,10 +347,6 @@ void ipath_layer_add(struct ipath_devdat
 		dd->ipath_layer.l_arg =
 			layer_add_one(dd->ipath_unit, dd);
 
-	if (verbs_add_one)
-		dd->verbs_layer.l_arg =
-			verbs_add_one(dd->ipath_unit, dd);
-
 	mutex_unlock(&ipath_layer_mutex);
 }
 
@@ -425,11 +357,6 @@ void ipath_layer_remove(struct ipath_dev
 	if (dd->ipath_layer.l_arg && layer_remove_one) {
 		layer_remove_one(dd->ipath_layer.l_arg);
 		dd->ipath_layer.l_arg = NULL;
-	}
-
-	if (dd->verbs_layer.l_arg && verbs_remove_one) {
-		verbs_remove_one(dd->verbs_layer.l_arg);
-		dd->verbs_layer.l_arg = NULL;
 	}
 
 	mutex_unlock(&ipath_layer_mutex);
@@ -521,94 +448,9 @@ static void __ipath_verbs_timer(unsigned
 		ipath_kreceive(dd);
 
 	/* Handle verbs layer timeouts. */
-	if (dd->verbs_layer.l_arg && verbs_timer_cb)
-		verbs_timer_cb(dd->verbs_layer.l_arg);
-
-	mod_timer(&dd->verbs_layer.l_timer, jiffies + 1);
-}
-
-/**
- * ipath_verbs_register - verbs layer registration
- * @l_piobufavail: callback for when PIO buffers become available
- * @l_rcv: callback for receiving a packet
- * @l_timer_cb: timer callback
- * @ipath_devdata: device data structure is put here
- */
-int ipath_verbs_register(void *(*l_add)(int, struct ipath_devdata *),
-			 void (*l_remove)(void *arg),
-			 int (*l_piobufavail) (void *arg),
-			 void (*l_rcv) (void *arg, void *rhdr,
-					void *data, u32 tlen),
-			 void (*l_timer_cb) (void *arg))
-{
-	struct ipath_devdata *dd, *tmp;
-	unsigned long flags;
-
-	mutex_lock(&ipath_layer_mutex);
-
-	verbs_add_one = l_add;
-	verbs_remove_one = l_remove;
-	verbs_piobufavail = l_piobufavail;
-	verbs_rcv = l_rcv;
-	verbs_timer_cb = l_timer_cb;
-
-	spin_lock_irqsave(&ipath_devs_lock, flags);
-
-	list_for_each_entry_safe(dd, tmp, &ipath_dev_list, ipath_list) {
-		if (!(dd->ipath_flags & IPATH_INITTED))
-			continue;
-
-		if (dd->verbs_layer.l_arg)
-			continue;
-
-		spin_unlock_irqrestore(&ipath_devs_lock, flags);
-		dd->verbs_layer.l_arg = l_add(dd->ipath_unit, dd);
-		spin_lock_irqsave(&ipath_devs_lock, flags);
-	}
-
-	spin_unlock_irqrestore(&ipath_devs_lock, flags);
-	mutex_unlock(&ipath_layer_mutex);
-
-	ipath_verbs_registered = 1;
-
-	return 0;
-}
-
-EXPORT_SYMBOL_GPL(ipath_verbs_register);
-
-void ipath_verbs_unregister(void)
-{
-	struct ipath_devdata *dd, *tmp;
-	unsigned long flags;
-
-	mutex_lock(&ipath_layer_mutex);
-	spin_lock_irqsave(&ipath_devs_lock, flags);
-
-	list_for_each_entry_safe(dd, tmp, &ipath_dev_list, ipath_list) {
-		*dd->ipath_statusp &= ~IPATH_STATUS_OIB_SMA;
-
-		if (dd->verbs_layer.l_arg && verbs_remove_one) {
-			spin_unlock_irqrestore(&ipath_devs_lock, flags);
-			verbs_remove_one(dd->verbs_layer.l_arg);
-			spin_lock_irqsave(&ipath_devs_lock, flags);
-			dd->verbs_layer.l_arg = NULL;
-		}
-	}
-
-	spin_unlock_irqrestore(&ipath_devs_lock, flags);
-
-	verbs_add_one = NULL;
-	verbs_remove_one = NULL;
-	verbs_piobufavail = NULL;
-	verbs_rcv = NULL;
-	verbs_timer_cb = NULL;
-
-	ipath_verbs_registered = 0;
-
-	mutex_unlock(&ipath_layer_mutex);
-}
-
-EXPORT_SYMBOL_GPL(ipath_verbs_unregister);
+	ipath_ib_timer(dd->verbs_dev);
+	mod_timer(&dd->verbs_timer, jiffies + 1);
+}
 
 int ipath_layer_open(struct ipath_devdata *dd, u32 * pktmax)
 {
@@ -702,8 +544,6 @@ u32 ipath_layer_get_cr_errpkey(struct ip
 {
 	return ipath_read_creg32(dd, dd->ipath_cregs->cr_errpkey);
 }
-
-EXPORT_SYMBOL_GPL(ipath_layer_get_cr_errpkey);
 
 static void update_sge(struct ipath_sge_state *ss, u32 length)
 {
@@ -981,8 +821,6 @@ bail:
 	return ret;
 }
 
-EXPORT_SYMBOL_GPL(ipath_verbs_send);
-
 int ipath_layer_snapshot_counters(struct ipath_devdata *dd, u64 *swords,
 				  u64 *rwords, u64 *spkts, u64 *rpkts,
 				  u64 *xmit_wait)
@@ -1006,8 +844,6 @@ bail:
 bail:
 	return ret;
 }
-
-EXPORT_SYMBOL_GPL(ipath_layer_snapshot_counters);
 
 /**
  * ipath_layer_get_counters - get various chip counters
@@ -1069,8 +905,6 @@ bail:
 	return ret;
 }
 
-EXPORT_SYMBOL_GPL(ipath_layer_get_counters);
-
 int ipath_layer_want_buffer(struct ipath_devdata *dd)
 {
 	set_bit(IPATH_S_PIOINTBUFAVAIL, &dd->ipath_sendctrl);
@@ -1079,8 +913,6 @@ int ipath_layer_want_buffer(struct ipath
 
 	return 0;
 }
-
-EXPORT_SYMBOL_GPL(ipath_layer_want_buffer);
 
 int ipath_layer_send_hdr(struct ipath_devdata *dd, struct ether_header *hdr)
 {
@@ -1174,16 +1006,14 @@ int ipath_layer_enable_timer(struct ipat
 				 (u64) (1 << 2));
 	}
 
-	init_timer(&dd->verbs_layer.l_timer);
-	dd->verbs_layer.l_timer.function = __ipath_verbs_timer;
-	dd->verbs_layer.l_timer.data = (unsigned long)dd;
-	dd->verbs_layer.l_timer.expires = jiffies + 1;
-	add_timer(&dd->verbs_layer.l_timer);
-
-	return 0;
-}
-
-EXPORT_SYMBOL_GPL(ipath_layer_enable_timer);
+	init_timer(&dd->verbs_timer);
+	dd->verbs_timer.function = __ipath_verbs_timer;
+	dd->verbs_timer.data = (unsigned long)dd;
+	dd->verbs_timer.expires = jiffies + 1;
+	add_timer(&dd->verbs_timer);
+
+	return 0;
+}
 
 int ipath_layer_disable_timer(struct ipath_devdata *dd)
 {
@@ -1191,12 +1021,10 @@ int ipath_layer_disable_timer(struct ipa
 	if (dd->ipath_flags & IPATH_GPIO_INTR)
 		ipath_write_kreg(dd, dd->ipath_kregs->kr_gpio_mask, 0);
 
-	del_timer_sync(&dd->verbs_layer.l_timer);
-
-	return 0;
-}
-
-EXPORT_SYMBOL_GPL(ipath_layer_disable_timer);
+	del_timer_sync(&dd->verbs_timer);
+
+	return 0;
+}
 
 /**
  * ipath_layer_set_verbs_flags - set the verbs layer flags
@@ -1225,8 +1053,6 @@ int ipath_layer_set_verbs_flags(struct i
 	return 0;
 }
 
-EXPORT_SYMBOL_GPL(ipath_layer_set_verbs_flags);
-
 /**
  * ipath_layer_get_npkeys - return the size of the PKEY table for port 0
  * @dd: the infinipath device
@@ -1235,8 +1061,6 @@ unsigned ipath_layer_get_npkeys(struct i
 {
 	return ARRAY_SIZE(dd->ipath_pd[0]->port_pkeys);
 }
-
-EXPORT_SYMBOL_GPL(ipath_layer_get_npkeys);
 
 /**
  * ipath_layer_get_pkey - return the indexed PKEY from the port 0 PKEY table
@@ -1255,8 +1079,6 @@ unsigned ipath_layer_get_pkey(struct ipa
 	return ret;
 }
 
-EXPORT_SYMBOL_GPL(ipath_layer_get_pkey);
-
 /**
  * ipath_layer_get_pkeys - return the PKEY table for port 0
  * @dd: the infinipath device
@@ -1270,8 +1092,6 @@ int ipath_layer_get_pkeys(struct ipath_d
 
 	return 0;
 }
-
-EXPORT_SYMBOL_GPL(ipath_layer_get_pkeys);
 
 /**
  * rm_pkey - decrecment the reference count for the given PKEY
@@ -1419,8 +1239,6 @@ int ipath_layer_set_pkeys(struct ipath_d
 	return 0;
 }
 
-EXPORT_SYMBOL_GPL(ipath_layer_set_pkeys);
-
 /**
  * ipath_layer_get_linkdowndefaultstate - get the default linkdown state
  * @dd: the infinipath device
@@ -1431,8 +1249,6 @@ int ipath_layer_get_linkdowndefaultstate
 {
 	return !!(dd->ipath_ibcctrl & INFINIPATH_IBCC_LINKDOWNDEFAULTSTATE);
 }
-
-EXPORT_SYMBOL_GPL(ipath_layer_get_linkdowndefaultstate);
 
 /**
  * ipath_layer_set_linkdowndefaultstate - set the default linkdown state
@@ -1453,16 +1269,12 @@ int ipath_layer_set_linkdowndefaultstate
 	return 0;
 }
 
-EXPORT_SYMBOL_GPL(ipath_layer_set_linkdowndefaultstate);
-
 int ipath_layer_get_phyerrthreshold(struct ipath_devdata *dd)
 {
 	return (dd->ipath_ibcctrl >>
 		INFINIPATH_IBCC_PHYERRTHRESHOLD_SHIFT) &
 		INFINIPATH_IBCC_PHYERRTHRESHOLD_MASK;
 }
-
-EXPORT_SYMBOL_GPL(ipath_layer_get_phyerrthreshold);
 
 /**
  * ipath_layer_set_phyerrthreshold - set the physical error threshold
@@ -1489,16 +1301,12 @@ int ipath_layer_set_phyerrthreshold(stru
 	return 0;
 }
 
-EXPORT_SYMBOL_GPL(ipath_layer_set_phyerrthreshold);
-
 int ipath_layer_get_overrunthreshold(struct ipath_devdata *dd)
 {
 	return (dd->ipath_ibcctrl >>
 		INFINIPATH_IBCC_OVERRUNTHRESHOLD_SHIFT) &
 		INFINIPATH_IBCC_OVERRUNTHRESHOLD_MASK;
 }
-
-EXPORT_SYMBOL_GPL(ipath_layer_get_overrunthreshold);
 
 /**
  * ipath_layer_set_overrunthreshold - set the overrun threshold
@@ -1525,17 +1333,13 @@ int ipath_layer_set_overrunthreshold(str
 	return 0;
 }
 
-EXPORT_SYMBOL_GPL(ipath_layer_set_overrunthreshold);
-
 int ipath_layer_get_boardname(struct ipath_devdata *dd, char *name,
 			      size_t namelen)
 {
 	return dd->ipath_f_get_boardname(dd, name, namelen);
 }
-EXPORT_SYMBOL_GPL(ipath_layer_get_boardname);
 
 u32 ipath_layer_get_rcvhdrentsize(struct ipath_devdata *dd)
 {
 	return dd->ipath_rcvhdrentsize;
 }
-EXPORT_SYMBOL_GPL(ipath_layer_get_rcvhdrentsize);
diff --git a/drivers/infiniband/hw/ipath/ipath_layer.h b/drivers/infiniband/hw/ipath/ipath_layer.h
--- a/drivers/infiniband/hw/ipath/ipath_layer.h	Fri Aug 25 11:19:44 2006 -0700
+++ b/drivers/infiniband/hw/ipath/ipath_layer.h	Fri Aug 25 11:19:45 2006 -0700
@@ -114,14 +114,7 @@ int ipath_layer_register(void *(*l_add)(
 				      struct sk_buff *),
 			 u16 rcv_opcode,
 			 int (*l_rcv_lid)(void *, void *));
-int ipath_verbs_register(void *(*l_add)(int, struct ipath_devdata *),
-			 void (*l_remove)(void *arg),
-			 int (*l_piobufavail)(void *arg),
-			 void (*l_rcv)(void *arg, void *rhdr,
-				       void *data, u32 tlen),
-			 void (*l_timer_cb)(void *arg));
 void ipath_layer_unregister(void);
-void ipath_verbs_unregister(void);
 int ipath_layer_open(struct ipath_devdata *, u32 * pktmax);
 u16 ipath_layer_get_lid(struct ipath_devdata *dd);
 int ipath_layer_get_mac(struct ipath_devdata *dd, u8 *);
@@ -145,7 +138,6 @@ int ipath_layer_want_buffer(struct ipath
 int ipath_layer_want_buffer(struct ipath_devdata *dd);
 int ipath_layer_set_guid(struct ipath_devdata *, __be64 guid);
 __be64 ipath_layer_get_guid(struct ipath_devdata *);
-u32 ipath_layer_get_nguid(struct ipath_devdata *);
 u32 ipath_layer_get_majrev(struct ipath_devdata *);
 u32 ipath_layer_get_minrev(struct ipath_devdata *);
 u32 ipath_layer_get_pcirev(struct ipath_devdata *);
diff --git a/drivers/infiniband/hw/ipath/ipath_verbs.c b/drivers/infiniband/hw/ipath/ipath_verbs.c
--- a/drivers/infiniband/hw/ipath/ipath_verbs.c	Fri Aug 25 11:19:44 2006 -0700
+++ b/drivers/infiniband/hw/ipath/ipath_verbs.c	Fri Aug 25 11:19:45 2006 -0700
@@ -368,7 +368,7 @@ static void ipath_qp_rcv(struct ipath_ib
 }
 
 /**
- * ipath_ib_rcv - process and incoming packet
+ * ipath_ib_rcv - process an incoming packet
  * @arg: the device pointer
  * @rhdr: the header of the packet
  * @data: the packet data
@@ -377,9 +377,9 @@ static void ipath_qp_rcv(struct ipath_ib
  * This is called from ipath_kreceive() to process an incoming packet at
  * interrupt level. Tlen is the length of the header + data + CRC in bytes.
  */
-static void ipath_ib_rcv(void *arg, void *rhdr, void *data, u32 tlen)
-{
-	struct ipath_ibdev *dev = (struct ipath_ibdev *) arg;
+void ipath_ib_rcv(struct ipath_ibdev *dev, void *rhdr, void *data,
+		  u32 tlen)
+{
 	struct ipath_ib_header *hdr = rhdr;
 	struct ipath_other_headers *ohdr;
 	struct ipath_qp *qp;
@@ -468,9 +468,8 @@ bail:;
  * This is called from ipath_do_rcv_timer() at interrupt level to check for
  * QPs which need retransmits and to collect performance numbers.
  */
-static void ipath_ib_timer(void *arg)
-{
-	struct ipath_ibdev *dev = (struct ipath_ibdev *) arg;
+void ipath_ib_timer(struct ipath_ibdev *dev)
+{
 	struct ipath_qp *resend = NULL;
 	struct list_head *last;
 	struct ipath_qp *qp;
@@ -564,9 +563,8 @@ static void ipath_ib_timer(void *arg)
  * QPs waiting for buffers (for now, just do a tasklet_hi_schedule and
  * return zero).
  */
-static int ipath_ib_piobufavail(void *arg)
-{
-	struct ipath_ibdev *dev = (struct ipath_ibdev *) arg;
+int ipath_ib_piobufavail(struct ipath_ibdev *dev)
+{
 	struct ipath_qp *qp;
 	unsigned long flags;
 
@@ -957,11 +955,10 @@ static int ipath_verbs_register_sysfs(st
 
 /**
  * ipath_register_ib_device - register our device with the infiniband core
- * @unit: the device number to register
  * @dd: the device data structure
  * Return the allocated ipath_ibdev pointer or NULL on error.
  */
-static void *ipath_register_ib_device(int unit, struct ipath_devdata *dd)
+int ipath_register_ib_device(struct ipath_devdata *dd)
 {
 	struct ipath_layer_counters cntrs;
 	struct ipath_ibdev *idev;
@@ -969,8 +966,10 @@ static void *ipath_register_ib_device(in
 	int ret;
 
 	idev = (struct ipath_ibdev *)ib_alloc_device(sizeof *idev);
-	if (idev == NULL)
-		goto bail;
+	if (idev == NULL) {
+		ret = -ENOMEM;
+		goto bail;
+	}
 
 	dev = &idev->ibdev;
 
@@ -1047,7 +1046,7 @@ static void *ipath_register_ib_device(in
 	if (!sys_image_guid)
 		sys_image_guid = ipath_layer_get_guid(dd);
 	idev->sys_image_guid = sys_image_guid;
-	idev->ib_unit = unit;
+	idev->ib_unit = dd->ipath_unit;
 	idev->dd = dd;
 
 	strlcpy(dev->name, "ipath%d", IB_DEVICE_NAME_MAX);
@@ -1153,16 +1152,16 @@ err_qp:
 err_qp:
 	ib_dealloc_device(dev);
 	_VERBS_ERROR("ib_ipath%d cannot register verbs (%d)!\n",
-		     unit, -ret);
+		     dd->ipath_unit, -ret);
 	idev = NULL;
 
 bail:
-	return idev;
-}
-
-static void ipath_unregister_ib_device(void *arg)
-{
-	struct ipath_ibdev *dev = (struct ipath_ibdev *) arg;
+	dd->verbs_dev = idev;
+	return ret;
+}
+
+void ipath_unregister_ib_device(struct ipath_ibdev *dev)
+{
 	struct ib_device *ibdev = &dev->ibdev;
 
 	ipath_layer_disable_timer(dev->dd);
@@ -1193,19 +1192,6 @@ static void ipath_unregister_ib_device(v
 	ib_dealloc_device(ibdev);
 }
 
-static int __init ipath_verbs_init(void)
-{
-	return ipath_verbs_register(ipath_register_ib_device,
-				    ipath_unregister_ib_device,
-				    ipath_ib_piobufavail, ipath_ib_rcv,
-				    ipath_ib_timer);
-}
-
-static void __exit ipath_verbs_cleanup(void)
-{
-	ipath_verbs_unregister();
-}
-
 static ssize_t show_rev(struct class_device *cdev, char *buf)
 {
 	struct ipath_ibdev *dev =
@@ -1297,6 +1283,3 @@ bail:
 bail:
 	return ret;
 }
-
-module_init(ipath_verbs_init);
-module_exit(ipath_verbs_cleanup);
diff --git a/drivers/infiniband/hw/ipath/ipath_verbs.h b/drivers/infiniband/hw/ipath/ipath_verbs.h
--- a/drivers/infiniband/hw/ipath/ipath_verbs.h	Fri Aug 25 11:19:44 2006 -0700
+++ b/drivers/infiniband/hw/ipath/ipath_verbs.h	Fri Aug 25 11:19:45 2006 -0700
@@ -711,6 +711,16 @@ int ipath_make_uc_req(struct ipath_qp *q
 int ipath_make_uc_req(struct ipath_qp *qp, struct ipath_other_headers *ohdr,
 		      u32 pmtu, u32 *bth0p, u32 *bth2p);
 
+int ipath_register_ib_device(struct ipath_devdata *);
+
+void ipath_unregister_ib_device(struct ipath_ibdev *);
+
+void ipath_ib_rcv(struct ipath_ibdev *, void *, void *, u32);
+
+int ipath_ib_piobufavail(struct ipath_ibdev *);
+
+void ipath_ib_timer(struct ipath_ibdev *);
+
 extern const enum ib_wc_opcode ib_ipath_wc_opcode[];
 
 extern const u8 ipath_cvt_physportstate[];

[PATCH 7 of 23] IB/ipath - simplify layering code

[Bryan O'Sullivan]

A lot of ipath layer code was only called in one place. Now that the
ipath_core and ib_ipath drivers are merged, it's more sensible to simply
inline the simple stuff that the layer code was doing.

Signed-off-by: Bryan O'Sullivan <bryan.osullivan@qlogic.com>

diff --git a/drivers/infiniband/hw/ipath/ipath_diag.c b/drivers/infiniband/hw/ipath/ipath_diag.c
--- a/drivers/infiniband/hw/ipath/ipath_diag.c	Fri Aug 25 11:19:45 2006 -0700
+++ b/drivers/infiniband/hw/ipath/ipath_diag.c	Fri Aug 25 11:19:45 2006 -0700
@@ -45,7 +45,6 @@
 #include <asm/uaccess.h>
 
 #include "ipath_kernel.h"
-#include "ipath_layer.h"
 #include "ipath_common.h"
 
 int ipath_diag_inuse;
diff --git a/drivers/infiniband/hw/ipath/ipath_driver.c b/drivers/infiniband/hw/ipath/ipath_driver.c
--- a/drivers/infiniband/hw/ipath/ipath_driver.c	Fri Aug 25 11:19:45 2006 -0700
+++ b/drivers/infiniband/hw/ipath/ipath_driver.c	Fri Aug 25 11:19:45 2006 -0700
@@ -39,7 +39,6 @@
 #include <linux/vmalloc.h>
 
 #include "ipath_kernel.h"
-#include "ipath_layer.h"
 #include "ipath_verbs.h"
 #include "ipath_common.h"
 
@@ -508,7 +507,6 @@ static int __devinit ipath_init_one(stru
 	ipathfs_add_device(dd);
 	ipath_user_add(dd);
 	ipath_diag_add(dd);
-	ipath_layer_add(dd);
 	ipath_register_ib_device(dd);
 
 	goto bail;
@@ -539,7 +537,6 @@ static void __devexit ipath_remove_one(s
 
 	dd = pci_get_drvdata(pdev);
 	ipath_unregister_ib_device(dd->verbs_dev);
-	ipath_layer_remove(dd);
 	ipath_diag_remove(dd);
 	ipath_user_remove(dd);
 	ipathfs_remove_device(dd);
@@ -614,11 +611,12 @@ void ipath_disarm_piobufs(struct ipath_d
  *
  * wait up to msecs milliseconds for IB link state change to occur for
  * now, take the easy polling route.  Currently used only by
- * ipath_layer_set_linkstate.  Returns 0 if state reached, otherwise
+ * ipath_set_linkstate.  Returns 0 if state reached, otherwise
  * -ETIMEDOUT state can have multiple states set, for any of several
  * transitions.
  */
-int ipath_wait_linkstate(struct ipath_devdata *dd, u32 state, int msecs)
+static int ipath_wait_linkstate(struct ipath_devdata *dd, u32 state,
+				int msecs)
 {
 	dd->ipath_sma_state_wanted = state;
 	wait_event_interruptible_timeout(ipath_sma_state_wait,
@@ -814,58 +812,6 @@ bail:
 	return skb;
 }
 
-/**
- * ipath_rcv_layer - receive a packet for the layered (ethernet) driver
- * @dd: the infinipath device
- * @etail: the sk_buff number
- * @tlen: the total packet length
- * @hdr: the ethernet header
- *
- * Separate routine for better overall optimization
- */
-static void ipath_rcv_layer(struct ipath_devdata *dd, u32 etail,
-			    u32 tlen, struct ether_header *hdr)
-{
-	u32 elen;
-	u8 pad, *bthbytes;
-	struct sk_buff *skb, *nskb;
-
-	if (dd->ipath_port0_skbs &&
-			hdr->sub_opcode == IPATH_ITH4X_OPCODE_ENCAP) {
-		/*
-		 * Allocate a new sk_buff to replace the one we give
-		 * to the network stack.
-		 */
-		nskb = ipath_alloc_skb(dd, GFP_ATOMIC);
-		if (!nskb) {
-			/* count OK packets that we drop */
-			ipath_stats.sps_krdrops++;
-			return;
-		}
-
-		bthbytes = (u8 *) hdr->bth;
-		pad = (bthbytes[1] >> 4) & 3;
-		/* +CRC32 */
-		elen = tlen - (sizeof(*hdr) + pad + sizeof(u32));
-
-		skb = dd->ipath_port0_skbs[etail];
-		dd->ipath_port0_skbs[etail] = nskb;
-		skb_put(skb, elen);
-
-		dd->ipath_f_put_tid(dd, etail + (u64 __iomem *)
-				    ((char __iomem *) dd->ipath_kregbase
-				     + dd->ipath_rcvegrbase), 0,
-				    virt_to_phys(nskb->data));
-
-		__ipath_layer_rcv(dd, hdr, skb);
-
-		/* another ether packet received */
-		ipath_stats.sps_ether_rpkts++;
-	}
-	else if (hdr->sub_opcode == IPATH_ITH4X_OPCODE_LID_ARP)
-		__ipath_layer_rcv_lid(dd, hdr);
-}
-
 static void ipath_rcv_hdrerr(struct ipath_devdata *dd,
 			     u32 eflags,
 			     u32 l,
@@ -979,22 +925,17 @@ reloop:
 		if (unlikely(eflags))
 			ipath_rcv_hdrerr(dd, eflags, l, etail, rc);
 		else if (etype == RCVHQ_RCV_TYPE_NON_KD) {
-				ipath_ib_rcv(dd->verbs_dev, rc + 1, ebuf,
-					     tlen);
-				if (dd->ipath_lli_counter)
-					dd->ipath_lli_counter--;
-
-		} else if (etype == RCVHQ_RCV_TYPE_EAGER) {
-			if (qp == IPATH_KD_QP &&
-			    bthbytes[0] == ipath_layer_rcv_opcode &&
-			    ebuf)
-				ipath_rcv_layer(dd, etail, tlen,
-						(struct ether_header *)hdr);
-			else
-				ipath_cdbg(PKT, "typ %x, opcode %x (eager, "
-					   "qp=%x), len %x; ignored\n",
-					   etype, bthbytes[0], qp, tlen);
-		}
+			ipath_ib_rcv(dd->verbs_dev, rc + 1, ebuf, tlen);
+			if (dd->ipath_lli_counter)
+				dd->ipath_lli_counter--;
+			ipath_cdbg(PKT, "typ %x, opcode %x (eager, "
+				   "qp=%x), len %x; ignored\n",
+				   etype, bthbytes[0], qp, tlen);
+		}
+		else if (etype == RCVHQ_RCV_TYPE_EAGER)
+			ipath_cdbg(PKT, "typ %x, opcode %x (eager, "
+				   "qp=%x), len %x; ignored\n",
+				   etype, bthbytes[0], qp, tlen);
 		else if (etype == RCVHQ_RCV_TYPE_EXPECTED)
 			ipath_dbg("Bug: Expected TID, opcode %x; ignored\n",
 				  be32_to_cpu(hdr->bth[0]) & 0xff);
@@ -1320,13 +1261,6 @@ rescan:
 		goto bail;
 	}
 
-	if (updated)
-		/*
-		 * ran out of bufs, now some (at least this one we just
-		 * got) are now available, so tell the layered driver.
-		 */
-		__ipath_layer_intr(dd, IPATH_LAYER_INT_SEND_CONTINUE);
-
 	/*
 	 * set next starting place.  Since it's just an optimization,
 	 * it doesn't matter who wins on this, so no locking
@@ -1503,7 +1437,7 @@ int ipath_waitfor_mdio_cmdready(struct i
 	return ret;
 }
 
-void ipath_set_ib_lstate(struct ipath_devdata *dd, int which)
+static void ipath_set_ib_lstate(struct ipath_devdata *dd, int which)
 {
 	static const char *what[4] = {
 		[0] = "DOWN",
@@ -1537,6 +1471,180 @@ void ipath_set_ib_lstate(struct ipath_de
 			 dd->ipath_ibcctrl | which);
 }
 
+int ipath_set_linkstate(struct ipath_devdata *dd, u8 newstate)
+{
+	u32 lstate;
+	int ret;
+
+	switch (newstate) {
+	case IPATH_IB_LINKDOWN:
+		ipath_set_ib_lstate(dd, INFINIPATH_IBCC_LINKINITCMD_POLL <<
+				    INFINIPATH_IBCC_LINKINITCMD_SHIFT);
+		/* don't wait */
+		ret = 0;
+		goto bail;
+
+	case IPATH_IB_LINKDOWN_SLEEP:
+		ipath_set_ib_lstate(dd, INFINIPATH_IBCC_LINKINITCMD_SLEEP <<
+				    INFINIPATH_IBCC_LINKINITCMD_SHIFT);
+		/* don't wait */
+		ret = 0;
+		goto bail;
+
+	case IPATH_IB_LINKDOWN_DISABLE:
+		ipath_set_ib_lstate(dd,
+				    INFINIPATH_IBCC_LINKINITCMD_DISABLE <<
+				    INFINIPATH_IBCC_LINKINITCMD_SHIFT);
+		/* don't wait */
+		ret = 0;
+		goto bail;
+
+	case IPATH_IB_LINKINIT:
+		if (dd->ipath_flags & IPATH_LINKINIT) {
+			ret = 0;
+			goto bail;
+		}
+		ipath_set_ib_lstate(dd, INFINIPATH_IBCC_LINKCMD_INIT <<
+				    INFINIPATH_IBCC_LINKCMD_SHIFT);
+		lstate = IPATH_LINKINIT;
+		break;
+
+	case IPATH_IB_LINKARM:
+		if (dd->ipath_flags & IPATH_LINKARMED) {
+			ret = 0;
+			goto bail;
+		}
+		if (!(dd->ipath_flags &
+		      (IPATH_LINKINIT | IPATH_LINKACTIVE))) {
+			ret = -EINVAL;
+			goto bail;
+		}
+		ipath_set_ib_lstate(dd, INFINIPATH_IBCC_LINKCMD_ARMED <<
+				    INFINIPATH_IBCC_LINKCMD_SHIFT);
+		/*
+		 * Since the port can transition to ACTIVE by receiving
+		 * a non VL 15 packet, wait for either state.
+		 */
+		lstate = IPATH_LINKARMED | IPATH_LINKACTIVE;
+		break;
+
+	case IPATH_IB_LINKACTIVE:
+		if (dd->ipath_flags & IPATH_LINKACTIVE) {
+			ret = 0;
+			goto bail;
+		}
+		if (!(dd->ipath_flags & IPATH_LINKARMED)) {
+			ret = -EINVAL;
+			goto bail;
+		}
+		ipath_set_ib_lstate(dd, INFINIPATH_IBCC_LINKCMD_ACTIVE <<
+				    INFINIPATH_IBCC_LINKCMD_SHIFT);
+		lstate = IPATH_LINKACTIVE;
+		break;
+
+	default:
+		ipath_dbg("Invalid linkstate 0x%x requested\n", newstate);
+		ret = -EINVAL;
+		goto bail;
+	}
+	ret = ipath_wait_linkstate(dd, lstate, 2000);
+
+bail:
+	return ret;
+}
+
+/**
+ * ipath_set_mtu - set the MTU
+ * @dd: the infinipath device
+ * @arg: the new MTU
+ *
+ * we can handle "any" incoming size, the issue here is whether we
+ * need to restrict our outgoing size.   For now, we don't do any
+ * sanity checking on this, and we don't deal with what happens to
+ * programs that are already running when the size changes.
+ * NOTE: changing the MTU will usually cause the IBC to go back to
+ * link initialize (IPATH_IBSTATE_INIT) state...
+ */
+int ipath_set_mtu(struct ipath_devdata *dd, u16 arg)
+{
+	u32 piosize;
+	int changed = 0;
+	int ret;
+
+	/*
+	 * mtu is IB data payload max.  It's the largest power of 2 less
+	 * than piosize (or even larger, since it only really controls the
+	 * largest we can receive; we can send the max of the mtu and
+	 * piosize).  We check that it's one of the valid IB sizes.
+	 */
+	if (arg != 256 && arg != 512 && arg != 1024 && arg != 2048 &&
+	    arg != 4096) {
+		ipath_dbg("Trying to set invalid mtu %u, failing\n", arg);
+		ret = -EINVAL;
+		goto bail;
+	}
+	if (dd->ipath_ibmtu == arg) {
+		ret = 0;        /* same as current */
+		goto bail;
+	}
+
+	piosize = dd->ipath_ibmaxlen;
+	dd->ipath_ibmtu = arg;
+
+	if (arg >= (piosize - IPATH_PIO_MAXIBHDR)) {
+		/* Only if it's not the initial value (or reset to it) */
+		if (piosize != dd->ipath_init_ibmaxlen) {
+			dd->ipath_ibmaxlen = piosize;
+			changed = 1;
+		}
+	} else if ((arg + IPATH_PIO_MAXIBHDR) != dd->ipath_ibmaxlen) {
+		piosize = arg + IPATH_PIO_MAXIBHDR;
+		ipath_cdbg(VERBOSE, "ibmaxlen was 0x%x, setting to 0x%x "
+			   "(mtu 0x%x)\n", dd->ipath_ibmaxlen, piosize,
+			   arg);
+		dd->ipath_ibmaxlen = piosize;
+		changed = 1;
+	}
+
+	if (changed) {
+		/*
+		 * set the IBC maxpktlength to the size of our pio
+		 * buffers in words
+		 */
+		u64 ibc = dd->ipath_ibcctrl;
+		ibc &= ~(INFINIPATH_IBCC_MAXPKTLEN_MASK <<
+			 INFINIPATH_IBCC_MAXPKTLEN_SHIFT);
+
+		piosize = piosize - 2 * sizeof(u32);    /* ignore pbc */
+		dd->ipath_ibmaxlen = piosize;
+		piosize /= sizeof(u32); /* in words */
+		/*
+		 * for ICRC, which we only send in diag test pkt mode, and
+		 * we don't need to worry about that for mtu
+		 */
+		piosize += 1;
+
+		ibc |= piosize << INFINIPATH_IBCC_MAXPKTLEN_SHIFT;
+		dd->ipath_ibcctrl = ibc;
+		ipath_write_kreg(dd, dd->ipath_kregs->kr_ibcctrl,
+				 dd->ipath_ibcctrl);
+		dd->ipath_f_tidtemplate(dd);
+	}
+
+	ret = 0;
+
+bail:
+	return ret;
+}
+
+int ipath_set_lid(struct ipath_devdata *dd, u32 arg, u8 lmc)
+{
+	dd->ipath_lid = arg;
+	dd->ipath_lmc = lmc;
+
+	return 0;
+}
+
 /**
  * ipath_read_kreg64_port - read a device's per-port 64-bit kernel register
  * @dd: the infinipath device
@@ -1639,13 +1747,6 @@ void ipath_shutdown_device(struct ipath_
 
 	ipath_set_ib_lstate(dd, INFINIPATH_IBCC_LINKINITCMD_DISABLE <<
 			    INFINIPATH_IBCC_LINKINITCMD_SHIFT);
-
-	/*
-	 * we are shutting down, so tell the layered driver.  We don't do
-	 * this on just a link state change, much like ethernet, a cable
-	 * unplug, etc. doesn't change driver state
-	 */
-	ipath_layer_intr(dd, IPATH_LAYER_INT_IF_DOWN);
 
 	/* disable IBC */
 	dd->ipath_control &= ~INFINIPATH_C_LINKENABLE;
diff --git a/drivers/infiniband/hw/ipath/ipath_file_ops.c b/drivers/infiniband/hw/ipath/ipath_file_ops.c
--- a/drivers/infiniband/hw/ipath/ipath_file_ops.c	Fri Aug 25 11:19:45 2006 -0700
+++ b/drivers/infiniband/hw/ipath/ipath_file_ops.c	Fri Aug 25 11:19:45 2006 -0700
@@ -39,7 +39,6 @@
 #include <asm/pgtable.h>
 
 #include "ipath_kernel.h"
-#include "ipath_layer.h"
 #include "ipath_common.h"
 
 static int ipath_open(struct inode *, struct file *);
diff --git a/drivers/infiniband/hw/ipath/ipath_intr.c b/drivers/infiniband/hw/ipath/ipath_intr.c
--- a/drivers/infiniband/hw/ipath/ipath_intr.c	Fri Aug 25 11:19:45 2006 -0700
+++ b/drivers/infiniband/hw/ipath/ipath_intr.c	Fri Aug 25 11:19:45 2006 -0700
@@ -34,7 +34,6 @@
 #include <linux/pci.h>
 
 #include "ipath_kernel.h"
-#include "ipath_layer.h"
 #include "ipath_verbs.h"
 #include "ipath_common.h"
 
@@ -290,8 +289,6 @@ static void handle_e_ibstatuschanged(str
 		*dd->ipath_statusp |=
 			IPATH_STATUS_IB_READY | IPATH_STATUS_IB_CONF;
 		dd->ipath_f_setextled(dd, lstate, ltstate);
-
-		__ipath_layer_intr(dd, IPATH_LAYER_INT_IF_UP);
 	} else if ((val & IPATH_IBSTATE_MASK) == IPATH_IBSTATE_INIT) {
 		/*
 		 * set INIT and DOWN.  Down is checked by most of the other
@@ -709,10 +706,6 @@ static void handle_layer_pioavail(struct
 {
 	int ret;
 
-	ret = __ipath_layer_intr(dd, IPATH_LAYER_INT_SEND_CONTINUE);
-	if (ret > 0)
-		goto set;
-
 	ret = ipath_ib_piobufavail(dd->verbs_dev);
 	if (ret > 0)
 		goto set;
diff --git a/drivers/infiniband/hw/ipath/ipath_kernel.h b/drivers/infiniband/hw/ipath/ipath_kernel.h
--- a/drivers/infiniband/hw/ipath/ipath_kernel.h	Fri Aug 25 11:19:45 2006 -0700
+++ b/drivers/infiniband/hw/ipath/ipath_kernel.h	Fri Aug 25 11:19:45 2006 -0700
@@ -518,16 +518,6 @@ extern spinlock_t ipath_devs_lock;
 extern spinlock_t ipath_devs_lock;
 extern struct ipath_devdata *ipath_lookup(int unit);
 
-extern u16 ipath_layer_rcv_opcode;
-extern int __ipath_layer_intr(struct ipath_devdata *, u32);
-extern int ipath_layer_intr(struct ipath_devdata *, u32);
-extern int __ipath_layer_rcv(struct ipath_devdata *, void *,
-			     struct sk_buff *);
-extern int __ipath_layer_rcv_lid(struct ipath_devdata *, void *);
-
-void ipath_layer_add(struct ipath_devdata *);
-void ipath_layer_remove(struct ipath_devdata *);
-
 int ipath_init_chip(struct ipath_devdata *, int);
 int ipath_enable_wc(struct ipath_devdata *dd);
 void ipath_disable_wc(struct ipath_devdata *dd);
@@ -575,12 +565,13 @@ void ipath_free_pddata(struct ipath_devd
 
 int ipath_parse_ushort(const char *str, unsigned short *valp);
 
-int ipath_wait_linkstate(struct ipath_devdata *, u32, int);
-void ipath_set_ib_lstate(struct ipath_devdata *, int);
 void ipath_kreceive(struct ipath_devdata *);
 int ipath_setrcvhdrsize(struct ipath_devdata *, unsigned);
 int ipath_reset_device(int);
 void ipath_get_faststats(unsigned long);
+int ipath_set_linkstate(struct ipath_devdata *, u8);
+int ipath_set_mtu(struct ipath_devdata *, u16);
+int ipath_set_lid(struct ipath_devdata *, u32, u8);
 
 /* for use in system calls, where we want to know device type, etc. */
 #define port_fp(fp) ((struct ipath_portdata *) (fp)->private_data)
diff --git a/drivers/infiniband/hw/ipath/ipath_layer.c b/drivers/infiniband/hw/ipath/ipath_layer.c
--- a/drivers/infiniband/hw/ipath/ipath_layer.c	Fri Aug 25 11:19:45 2006 -0700
+++ b/drivers/infiniband/hw/ipath/ipath_layer.c	Fri Aug 25 11:19:45 2006 -0700
@@ -101,242 +101,14 @@ int __ipath_layer_rcv_lid(struct ipath_d
 	return ret;
 }
 
-int ipath_layer_set_linkstate(struct ipath_devdata *dd, u8 newstate)
-{
-	u32 lstate;
-	int ret;
-
-	switch (newstate) {
-	case IPATH_IB_LINKDOWN:
-		ipath_set_ib_lstate(dd, INFINIPATH_IBCC_LINKINITCMD_POLL <<
-				    INFINIPATH_IBCC_LINKINITCMD_SHIFT);
-		/* don't wait */
-		ret = 0;
-		goto bail;
-
-	case IPATH_IB_LINKDOWN_SLEEP:
-		ipath_set_ib_lstate(dd, INFINIPATH_IBCC_LINKINITCMD_SLEEP <<
-				    INFINIPATH_IBCC_LINKINITCMD_SHIFT);
-		/* don't wait */
-		ret = 0;
-		goto bail;
-
-	case IPATH_IB_LINKDOWN_DISABLE:
-		ipath_set_ib_lstate(dd,
-				    INFINIPATH_IBCC_LINKINITCMD_DISABLE <<
-				    INFINIPATH_IBCC_LINKINITCMD_SHIFT);
-		/* don't wait */
-		ret = 0;
-		goto bail;
-
-	case IPATH_IB_LINKINIT:
-		if (dd->ipath_flags & IPATH_LINKINIT) {
-			ret = 0;
-			goto bail;
-		}
-		ipath_set_ib_lstate(dd, INFINIPATH_IBCC_LINKCMD_INIT <<
-				    INFINIPATH_IBCC_LINKCMD_SHIFT);
-		lstate = IPATH_LINKINIT;
-		break;
-
-	case IPATH_IB_LINKARM:
-		if (dd->ipath_flags & IPATH_LINKARMED) {
-			ret = 0;
-			goto bail;
-		}
-		if (!(dd->ipath_flags &
-		      (IPATH_LINKINIT | IPATH_LINKACTIVE))) {
-			ret = -EINVAL;
-			goto bail;
-		}
-		ipath_set_ib_lstate(dd, INFINIPATH_IBCC_LINKCMD_ARMED <<
-				    INFINIPATH_IBCC_LINKCMD_SHIFT);
-		/*
-		 * Since the port can transition to ACTIVE by receiving
-		 * a non VL 15 packet, wait for either state.
-		 */
-		lstate = IPATH_LINKARMED | IPATH_LINKACTIVE;
-		break;
-
-	case IPATH_IB_LINKACTIVE:
-		if (dd->ipath_flags & IPATH_LINKACTIVE) {
-			ret = 0;
-			goto bail;
-		}
-		if (!(dd->ipath_flags & IPATH_LINKARMED)) {
-			ret = -EINVAL;
-			goto bail;
-		}
-		ipath_set_ib_lstate(dd, INFINIPATH_IBCC_LINKCMD_ACTIVE <<
-				    INFINIPATH_IBCC_LINKCMD_SHIFT);
-		lstate = IPATH_LINKACTIVE;
-		break;
-
-	default:
-		ipath_dbg("Invalid linkstate 0x%x requested\n", newstate);
-		ret = -EINVAL;
-		goto bail;
-	}
-	ret = ipath_wait_linkstate(dd, lstate, 2000);
-
-bail:
-	return ret;
-}
-
-/**
- * ipath_layer_set_mtu - set the MTU
- * @dd: the infinipath device
- * @arg: the new MTU
- *
- * we can handle "any" incoming size, the issue here is whether we
- * need to restrict our outgoing size.   For now, we don't do any
- * sanity checking on this, and we don't deal with what happens to
- * programs that are already running when the size changes.
- * NOTE: changing the MTU will usually cause the IBC to go back to
- * link initialize (IPATH_IBSTATE_INIT) state...
- */
-int ipath_layer_set_mtu(struct ipath_devdata *dd, u16 arg)
-{
-	u32 piosize;
-	int changed = 0;
-	int ret;
-
-	/*
-	 * mtu is IB data payload max.  It's the largest power of 2 less
-	 * than piosize (or even larger, since it only really controls the
-	 * largest we can receive; we can send the max of the mtu and
-	 * piosize).  We check that it's one of the valid IB sizes.
-	 */
-	if (arg != 256 && arg != 512 && arg != 1024 && arg != 2048 &&
-	    arg != 4096) {
-		ipath_dbg("Trying to set invalid mtu %u, failing\n", arg);
-		ret = -EINVAL;
-		goto bail;
-	}
-	if (dd->ipath_ibmtu == arg) {
-		ret = 0;	/* same as current */
-		goto bail;
-	}
-
-	piosize = dd->ipath_ibmaxlen;
-	dd->ipath_ibmtu = arg;
-
-	if (arg >= (piosize - IPATH_PIO_MAXIBHDR)) {
-		/* Only if it's not the initial value (or reset to it) */
-		if (piosize != dd->ipath_init_ibmaxlen) {
-			dd->ipath_ibmaxlen = piosize;
-			changed = 1;
-		}
-	} else if ((arg + IPATH_PIO_MAXIBHDR) != dd->ipath_ibmaxlen) {
-		piosize = arg + IPATH_PIO_MAXIBHDR;
-		ipath_cdbg(VERBOSE, "ibmaxlen was 0x%x, setting to 0x%x "
-			   "(mtu 0x%x)\n", dd->ipath_ibmaxlen, piosize,
-			   arg);
-		dd->ipath_ibmaxlen = piosize;
-		changed = 1;
-	}
-
-	if (changed) {
-		/*
-		 * set the IBC maxpktlength to the size of our pio
-		 * buffers in words
-		 */
-		u64 ibc = dd->ipath_ibcctrl;
-		ibc &= ~(INFINIPATH_IBCC_MAXPKTLEN_MASK <<
-			 INFINIPATH_IBCC_MAXPKTLEN_SHIFT);
-
-		piosize = piosize - 2 * sizeof(u32);	/* ignore pbc */
-		dd->ipath_ibmaxlen = piosize;
-		piosize /= sizeof(u32);	/* in words */
-		/*
-		 * for ICRC, which we only send in diag test pkt mode, and
-		 * we don't need to worry about that for mtu
-		 */
-		piosize += 1;
-
-		ibc |= piosize << INFINIPATH_IBCC_MAXPKTLEN_SHIFT;
-		dd->ipath_ibcctrl = ibc;
-		ipath_write_kreg(dd, dd->ipath_kregs->kr_ibcctrl,
-				 dd->ipath_ibcctrl);
-		dd->ipath_f_tidtemplate(dd);
-	}
-
-	ret = 0;
-
-bail:
-	return ret;
-}
-
-int ipath_set_lid(struct ipath_devdata *dd, u32 arg, u8 lmc)
-{
-	dd->ipath_lid = arg;
-	dd->ipath_lmc = lmc;
-
+void ipath_layer_lid_changed(struct ipath_devdata *dd)
+{
 	mutex_lock(&ipath_layer_mutex);
 
 	if (dd->ipath_layer.l_arg && layer_intr)
 		layer_intr(dd->ipath_layer.l_arg, IPATH_LAYER_INT_LID);
 
 	mutex_unlock(&ipath_layer_mutex);
-
-	return 0;
-}
-
-int ipath_layer_set_guid(struct ipath_devdata *dd, __be64 guid)
-{
-	/* XXX - need to inform anyone who cares this just happened. */
-	dd->ipath_guid = guid;
-	return 0;
-}
-
-__be64 ipath_layer_get_guid(struct ipath_devdata *dd)
-{
-	return dd->ipath_guid;
-}
-
-u32 ipath_layer_get_majrev(struct ipath_devdata *dd)
-{
-	return dd->ipath_majrev;
-}
-
-u32 ipath_layer_get_minrev(struct ipath_devdata *dd)
-{
-	return dd->ipath_minrev;
-}
-
-u32 ipath_layer_get_pcirev(struct ipath_devdata *dd)
-{
-	return dd->ipath_pcirev;
-}
-
-u32 ipath_layer_get_flags(struct ipath_devdata *dd)
-{
-	return dd->ipath_flags;
-}
-
-struct device *ipath_layer_get_device(struct ipath_devdata *dd)
-{
-	return &dd->pcidev->dev;
-}
-
-u16 ipath_layer_get_deviceid(struct ipath_devdata *dd)
-{
-	return dd->ipath_deviceid;
-}
-
-u32 ipath_layer_get_vendorid(struct ipath_devdata *dd)
-{
-	return dd->ipath_vendorid;
-}
-
-u64 ipath_layer_get_lastibcstat(struct ipath_devdata *dd)
-{
-	return dd->ipath_lastibcstat;
-}
-
-u32 ipath_layer_get_ibmtu(struct ipath_devdata *dd)
-{
-	return dd->ipath_ibmtu;
 }
 
 void ipath_layer_add(struct ipath_devdata *dd)
@@ -435,22 +207,6 @@ void ipath_layer_unregister(void)
 }
 
 EXPORT_SYMBOL_GPL(ipath_layer_unregister);
-
-static void __ipath_verbs_timer(unsigned long arg)
-{
-	struct ipath_devdata *dd = (struct ipath_devdata *) arg;
-
-	/*
-	 * If port 0 receive packet interrupts are not available, or
-	 * can be missed, poll the receive queue
-	 */
-	if (dd->ipath_flags & IPATH_POLL_RX_INTR)
-		ipath_kreceive(dd);
-
-	/* Handle verbs layer timeouts. */
-	ipath_ib_timer(dd->verbs_dev);
-	mod_timer(&dd->verbs_timer, jiffies + 1);
-}
 
 int ipath_layer_open(struct ipath_devdata *dd, u32 * pktmax)
 {
@@ -539,380 +295,6 @@ u16 ipath_layer_get_bcast(struct ipath_d
 }
 
 EXPORT_SYMBOL_GPL(ipath_layer_get_bcast);
-
-u32 ipath_layer_get_cr_errpkey(struct ipath_devdata *dd)
-{
-	return ipath_read_creg32(dd, dd->ipath_cregs->cr_errpkey);
-}
-
-static void update_sge(struct ipath_sge_state *ss, u32 length)
-{
-	struct ipath_sge *sge = &ss->sge;
-
-	sge->vaddr += length;
-	sge->length -= length;
-	sge->sge_length -= length;
-	if (sge->sge_length == 0) {
-		if (--ss->num_sge)
-			*sge = *ss->sg_list++;
-	} else if (sge->length == 0 && sge->mr != NULL) {
-		if (++sge->n >= IPATH_SEGSZ) {
-			if (++sge->m >= sge->mr->mapsz)
-				return;
-			sge->n = 0;
-		}
-		sge->vaddr = sge->mr->map[sge->m]->segs[sge->n].vaddr;
-		sge->length = sge->mr->map[sge->m]->segs[sge->n].length;
-	}
-}
-
-#ifdef __LITTLE_ENDIAN
-static inline u32 get_upper_bits(u32 data, u32 shift)
-{
-	return data >> shift;
-}
-
-static inline u32 set_upper_bits(u32 data, u32 shift)
-{
-	return data << shift;
-}
-
-static inline u32 clear_upper_bytes(u32 data, u32 n, u32 off)
-{
-	data <<= ((sizeof(u32) - n) * BITS_PER_BYTE);
-	data >>= ((sizeof(u32) - n - off) * BITS_PER_BYTE);
-	return data;
-}
-#else
-static inline u32 get_upper_bits(u32 data, u32 shift)
-{
-	return data << shift;
-}
-
-static inline u32 set_upper_bits(u32 data, u32 shift)
-{
-	return data >> shift;
-}
-
-static inline u32 clear_upper_bytes(u32 data, u32 n, u32 off)
-{
-	data >>= ((sizeof(u32) - n) * BITS_PER_BYTE);
-	data <<= ((sizeof(u32) - n - off) * BITS_PER_BYTE);
-	return data;
-}
-#endif
-
-static void copy_io(u32 __iomem *piobuf, struct ipath_sge_state *ss,
-		    u32 length)
-{
-	u32 extra = 0;
-	u32 data = 0;
-	u32 last;
-
-	while (1) {
-		u32 len = ss->sge.length;
-		u32 off;
-
-		BUG_ON(len == 0);
-		if (len > length)
-			len = length;
-		if (len > ss->sge.sge_length)
-			len = ss->sge.sge_length;
-		/* If the source address is not aligned, try to align it. */
-		off = (unsigned long)ss->sge.vaddr & (sizeof(u32) - 1);
-		if (off) {
-			u32 *addr = (u32 *)((unsigned long)ss->sge.vaddr &
-					    ~(sizeof(u32) - 1));
-			u32 v = get_upper_bits(*addr, off * BITS_PER_BYTE);
-			u32 y;
-
-			y = sizeof(u32) - off;
-			if (len > y)
-				len = y;
-			if (len + extra >= sizeof(u32)) {
-				data |= set_upper_bits(v, extra *
-						       BITS_PER_BYTE);
-				len = sizeof(u32) - extra;
-				if (len == length) {
-					last = data;
-					break;
-				}
-				__raw_writel(data, piobuf);
-				piobuf++;
-				extra = 0;
-				data = 0;
-			} else {
-				/* Clear unused upper bytes */
-				data |= clear_upper_bytes(v, len, extra);
-				if (len == length) {
-					last = data;
-					break;
-				}
-				extra += len;
-			}
-		} else if (extra) {
-			/* Source address is aligned. */
-			u32 *addr = (u32 *) ss->sge.vaddr;
-			int shift = extra * BITS_PER_BYTE;
-			int ushift = 32 - shift;
-			u32 l = len;
-
-			while (l >= sizeof(u32)) {
-				u32 v = *addr;
-
-				data |= set_upper_bits(v, shift);
-				__raw_writel(data, piobuf);
-				data = get_upper_bits(v, ushift);
-				piobuf++;
-				addr++;
-				l -= sizeof(u32);
-			}
-			/*
-			 * We still have 'extra' number of bytes leftover.
-			 */
-			if (l) {
-				u32 v = *addr;
-
-				if (l + extra >= sizeof(u32)) {
-					data |= set_upper_bits(v, shift);
-					len -= l + extra - sizeof(u32);
-					if (len == length) {
-						last = data;
-						break;
-					}
-					__raw_writel(data, piobuf);
-					piobuf++;
-					extra = 0;
-					data = 0;
-				} else {
-					/* Clear unused upper bytes */
-					data |= clear_upper_bytes(v, l,
-								  extra);
-					if (len == length) {
-						last = data;
-						break;
-					}
-					extra += l;
-				}
-			} else if (len == length) {
-				last = data;
-				break;
-			}
-		} else if (len == length) {
-			u32 w;
-
-			/*
-			 * Need to round up for the last dword in the
-			 * packet.
-			 */
-			w = (len + 3) >> 2;
-			__iowrite32_copy(piobuf, ss->sge.vaddr, w - 1);
-			piobuf += w - 1;
-			last = ((u32 *) ss->sge.vaddr)[w - 1];
-			break;
-		} else {
-			u32 w = len >> 2;
-
-			__iowrite32_copy(piobuf, ss->sge.vaddr, w);
-			piobuf += w;
-
-			extra = len & (sizeof(u32) - 1);
-			if (extra) {
-				u32 v = ((u32 *) ss->sge.vaddr)[w];
-
-				/* Clear unused upper bytes */
-				data = clear_upper_bytes(v, extra, 0);
-			}
-		}
-		update_sge(ss, len);
-		length -= len;
-	}
-	/* Update address before sending packet. */
-	update_sge(ss, length);
-	/* must flush early everything before trigger word */
-	ipath_flush_wc();
-	__raw_writel(last, piobuf);
-	/* be sure trigger word is written */
-	ipath_flush_wc();
-}
-
-/**
- * ipath_verbs_send - send a packet from the verbs layer
- * @dd: the infinipath device
- * @hdrwords: the number of words in the header
- * @hdr: the packet header
- * @len: the length of the packet in bytes
- * @ss: the SGE to send
- *
- * This is like ipath_sma_send_pkt() in that we need to be able to send
- * packets after the chip is initialized (MADs) but also like
- * ipath_layer_send_hdr() since its used by the verbs layer.
- */
-int ipath_verbs_send(struct ipath_devdata *dd, u32 hdrwords,
-		     u32 *hdr, u32 len, struct ipath_sge_state *ss)
-{
-	u32 __iomem *piobuf;
-	u32 plen;
-	int ret;
-
-	/* +1 is for the qword padding of pbc */
-	plen = hdrwords + ((len + 3) >> 2) + 1;
-	if (unlikely((plen << 2) > dd->ipath_ibmaxlen)) {
-		ipath_dbg("packet len 0x%x too long, failing\n", plen);
-		ret = -EINVAL;
-		goto bail;
-	}
-
-	/* Get a PIO buffer to use. */
-	piobuf = ipath_getpiobuf(dd, NULL);
-	if (unlikely(piobuf == NULL)) {
-		ret = -EBUSY;
-		goto bail;
-	}
-
-	/*
-	 * Write len to control qword, no flags.
-	 * We have to flush after the PBC for correctness on some cpus
-	 * or WC buffer can be written out of order.
-	 */
-	writeq(plen, piobuf);
-	ipath_flush_wc();
-	piobuf += 2;
-	if (len == 0) {
-		/*
-		 * If there is just the header portion, must flush before
-		 * writing last word of header for correctness, and after
-		 * the last header word (trigger word).
-		 */
-		__iowrite32_copy(piobuf, hdr, hdrwords - 1);
-		ipath_flush_wc();
-		__raw_writel(hdr[hdrwords - 1], piobuf + hdrwords - 1);
-		ipath_flush_wc();
-		ret = 0;
-		goto bail;
-	}
-
-	__iowrite32_copy(piobuf, hdr, hdrwords);
-	piobuf += hdrwords;
-
-	/* The common case is aligned and contained in one segment. */
-	if (likely(ss->num_sge == 1 && len <= ss->sge.length &&
-		   !((unsigned long)ss->sge.vaddr & (sizeof(u32) - 1)))) {
-		u32 w;
-		u32 *addr = (u32 *) ss->sge.vaddr;
-
-		/* Update address before sending packet. */
-		update_sge(ss, len);
-		/* Need to round up for the last dword in the packet. */
-		w = (len + 3) >> 2;
-		__iowrite32_copy(piobuf, addr, w - 1);
-		/* must flush early everything before trigger word */
-		ipath_flush_wc();
-		__raw_writel(addr[w - 1], piobuf + w - 1);
-		/* be sure trigger word is written */
-		ipath_flush_wc();
-		ret = 0;
-		goto bail;
-	}
-	copy_io(piobuf, ss, len);
-	ret = 0;
-
-bail:
-	return ret;
-}
-
-int ipath_layer_snapshot_counters(struct ipath_devdata *dd, u64 *swords,
-				  u64 *rwords, u64 *spkts, u64 *rpkts,
-				  u64 *xmit_wait)
-{
-	int ret;
-
-	if (!(dd->ipath_flags & IPATH_INITTED)) {
-		/* no hardware, freeze, etc. */
-		ipath_dbg("unit %u not usable\n", dd->ipath_unit);
-		ret = -EINVAL;
-		goto bail;
-	}
-	*swords = ipath_snap_cntr(dd, dd->ipath_cregs->cr_wordsendcnt);
-	*rwords = ipath_snap_cntr(dd, dd->ipath_cregs->cr_wordrcvcnt);
-	*spkts = ipath_snap_cntr(dd, dd->ipath_cregs->cr_pktsendcnt);
-	*rpkts = ipath_snap_cntr(dd, dd->ipath_cregs->cr_pktrcvcnt);
-	*xmit_wait = ipath_snap_cntr(dd, dd->ipath_cregs->cr_sendstallcnt);
-
-	ret = 0;
-
-bail:
-	return ret;
-}
-
-/**
- * ipath_layer_get_counters - get various chip counters
- * @dd: the infinipath device
- * @cntrs: counters are placed here
- *
- * Return the counters needed by recv_pma_get_portcounters().
- */
-int ipath_layer_get_counters(struct ipath_devdata *dd,
-			      struct ipath_layer_counters *cntrs)
-{
-	int ret;
-
-	if (!(dd->ipath_flags & IPATH_INITTED)) {
-		/* no hardware, freeze, etc. */
-		ipath_dbg("unit %u not usable\n", dd->ipath_unit);
-		ret = -EINVAL;
-		goto bail;
-	}
-	cntrs->symbol_error_counter =
-		ipath_snap_cntr(dd, dd->ipath_cregs->cr_ibsymbolerrcnt);
-	cntrs->link_error_recovery_counter =
-		ipath_snap_cntr(dd, dd->ipath_cregs->cr_iblinkerrrecovcnt);
-	/*
-	 * The link downed counter counts when the other side downs the
-	 * connection.  We add in the number of times we downed the link
-	 * due to local link integrity errors to compensate.
-	 */
-	cntrs->link_downed_counter =
-		ipath_snap_cntr(dd, dd->ipath_cregs->cr_iblinkdowncnt);
-	cntrs->port_rcv_errors =
-		ipath_snap_cntr(dd, dd->ipath_cregs->cr_rxdroppktcnt) +
-		ipath_snap_cntr(dd, dd->ipath_cregs->cr_rcvovflcnt) +
-		ipath_snap_cntr(dd, dd->ipath_cregs->cr_portovflcnt) +
-		ipath_snap_cntr(dd, dd->ipath_cregs->cr_err_rlencnt) +
-		ipath_snap_cntr(dd, dd->ipath_cregs->cr_invalidrlencnt) +
-		ipath_snap_cntr(dd, dd->ipath_cregs->cr_erricrccnt) +
-		ipath_snap_cntr(dd, dd->ipath_cregs->cr_errvcrccnt) +
-		ipath_snap_cntr(dd, dd->ipath_cregs->cr_errlpcrccnt) +
-		ipath_snap_cntr(dd, dd->ipath_cregs->cr_badformatcnt);
-	cntrs->port_rcv_remphys_errors =
-		ipath_snap_cntr(dd, dd->ipath_cregs->cr_rcvebpcnt);
-	cntrs->port_xmit_discards =
-		ipath_snap_cntr(dd, dd->ipath_cregs->cr_unsupvlcnt);
-	cntrs->port_xmit_data =
-		ipath_snap_cntr(dd, dd->ipath_cregs->cr_wordsendcnt);
-	cntrs->port_rcv_data =
-		ipath_snap_cntr(dd, dd->ipath_cregs->cr_wordrcvcnt);
-	cntrs->port_xmit_packets =
-		ipath_snap_cntr(dd, dd->ipath_cregs->cr_pktsendcnt);
-	cntrs->port_rcv_packets =
-		ipath_snap_cntr(dd, dd->ipath_cregs->cr_pktrcvcnt);
-	cntrs->local_link_integrity_errors = dd->ipath_lli_errors;
-	cntrs->excessive_buffer_overrun_errors = 0; /* XXX */
-
-	ret = 0;
-
-bail:
-	return ret;
-}
-
-int ipath_layer_want_buffer(struct ipath_devdata *dd)
-{
-	set_bit(IPATH_S_PIOINTBUFAVAIL, &dd->ipath_sendctrl);
-	ipath_write_kreg(dd, dd->ipath_kregs->kr_sendctrl,
-			 dd->ipath_sendctrl);
-
-	return 0;
-}
 
 int ipath_layer_send_hdr(struct ipath_devdata *dd, struct ether_header *hdr)
 {
@@ -985,361 +367,3 @@ int ipath_layer_set_piointbufavail_int(s
 }
 
 EXPORT_SYMBOL_GPL(ipath_layer_set_piointbufavail_int);
-
-int ipath_layer_enable_timer(struct ipath_devdata *dd)
-{
-	/*
-	 * HT-400 has a design flaw where the chip and kernel idea
-	 * of the tail register don't always agree, and therefore we won't
-	 * get an interrupt on the next packet received.
-	 * If the board supports per packet receive interrupts, use it.
-	 * Otherwise, the timer function periodically checks for packets
-	 * to cover this case.
-	 * Either way, the timer is needed for verbs layer related
-	 * processing.
-	 */
-	if (dd->ipath_flags & IPATH_GPIO_INTR) {
-		ipath_write_kreg(dd, dd->ipath_kregs->kr_debugportselect,
-				 0x2074076542310ULL);
-		/* Enable GPIO bit 2 interrupt */
-		ipath_write_kreg(dd, dd->ipath_kregs->kr_gpio_mask,
-				 (u64) (1 << 2));
-	}
-
-	init_timer(&dd->verbs_timer);
-	dd->verbs_timer.function = __ipath_verbs_timer;
-	dd->verbs_timer.data = (unsigned long)dd;
-	dd->verbs_timer.expires = jiffies + 1;
-	add_timer(&dd->verbs_timer);
-
-	return 0;
-}
-
-int ipath_layer_disable_timer(struct ipath_devdata *dd)
-{
-	/* Disable GPIO bit 2 interrupt */
-	if (dd->ipath_flags & IPATH_GPIO_INTR)
-		ipath_write_kreg(dd, dd->ipath_kregs->kr_gpio_mask, 0);
-
-	del_timer_sync(&dd->verbs_timer);
-
-	return 0;
-}
-
-/**
- * ipath_layer_set_verbs_flags - set the verbs layer flags
- * @dd: the infinipath device
- * @flags: the flags to set
- */
-int ipath_layer_set_verbs_flags(struct ipath_devdata *dd, unsigned flags)
-{
-	struct ipath_devdata *ss;
-	unsigned long lflags;
-
-	spin_lock_irqsave(&ipath_devs_lock, lflags);
-
-	list_for_each_entry(ss, &ipath_dev_list, ipath_list) {
-		if (!(ss->ipath_flags & IPATH_INITTED))
-			continue;
-		if ((flags & IPATH_VERBS_KERNEL_SMA) &&
-		    !(*ss->ipath_statusp & IPATH_STATUS_SMA))
-			*ss->ipath_statusp |= IPATH_STATUS_OIB_SMA;
-		else
-			*ss->ipath_statusp &= ~IPATH_STATUS_OIB_SMA;
-	}
-
-	spin_unlock_irqrestore(&ipath_devs_lock, lflags);
-
-	return 0;
-}
-
-/**
- * ipath_layer_get_npkeys - return the size of the PKEY table for port 0
- * @dd: the infinipath device
- */
-unsigned ipath_layer_get_npkeys(struct ipath_devdata *dd)
-{
-	return ARRAY_SIZE(dd->ipath_pd[0]->port_pkeys);
-}
-
-/**
- * ipath_layer_get_pkey - return the indexed PKEY from the port 0 PKEY table
- * @dd: the infinipath device
- * @index: the PKEY index
- */
-unsigned ipath_layer_get_pkey(struct ipath_devdata *dd, unsigned index)
-{
-	unsigned ret;
-
-	if (index >= ARRAY_SIZE(dd->ipath_pd[0]->port_pkeys))
-		ret = 0;
-	else
-		ret = dd->ipath_pd[0]->port_pkeys[index];
-
-	return ret;
-}
-
-/**
- * ipath_layer_get_pkeys - return the PKEY table for port 0
- * @dd: the infinipath device
- * @pkeys: the pkey table is placed here
- */
-int ipath_layer_get_pkeys(struct ipath_devdata *dd, u16 * pkeys)
-{
-	struct ipath_portdata *pd = dd->ipath_pd[0];
-
-	memcpy(pkeys, pd->port_pkeys, sizeof(pd->port_pkeys));
-
-	return 0;
-}
-
-/**
- * rm_pkey - decrecment the reference count for the given PKEY
- * @dd: the infinipath device
- * @key: the PKEY index
- *
- * Return true if this was the last reference and the hardware table entry
- * needs to be changed.
- */
-static int rm_pkey(struct ipath_devdata *dd, u16 key)
-{
-	int i;
-	int ret;
-
-	for (i = 0; i < ARRAY_SIZE(dd->ipath_pkeys); i++) {
-		if (dd->ipath_pkeys[i] != key)
-			continue;
-		if (atomic_dec_and_test(&dd->ipath_pkeyrefs[i])) {
-			dd->ipath_pkeys[i] = 0;
-			ret = 1;
-			goto bail;
-		}
-		break;
-	}
-
-	ret = 0;
-
-bail:
-	return ret;
-}
-
-/**
- * add_pkey - add the given PKEY to the hardware table
- * @dd: the infinipath device
- * @key: the PKEY
- *
- * Return an error code if unable to add the entry, zero if no change,
- * or 1 if the hardware PKEY register needs to be updated.
- */
-static int add_pkey(struct ipath_devdata *dd, u16 key)
-{
-	int i;
-	u16 lkey = key & 0x7FFF;
-	int any = 0;
-	int ret;
-
-	if (lkey == 0x7FFF) {
-		ret = 0;
-		goto bail;
-	}
-
-	/* Look for an empty slot or a matching PKEY. */
-	for (i = 0; i < ARRAY_SIZE(dd->ipath_pkeys); i++) {
-		if (!dd->ipath_pkeys[i]) {
-			any++;
-			continue;
-		}
-		/* If it matches exactly, try to increment the ref count */
-		if (dd->ipath_pkeys[i] == key) {
-			if (atomic_inc_return(&dd->ipath_pkeyrefs[i]) > 1) {
-				ret = 0;
-				goto bail;
-			}
-			/* Lost the race. Look for an empty slot below. */
-			atomic_dec(&dd->ipath_pkeyrefs[i]);
-			any++;
-		}
-		/*
-		 * It makes no sense to have both the limited and unlimited
-		 * PKEY set at the same time since the unlimited one will
-		 * disable the limited one.
-		 */
-		if ((dd->ipath_pkeys[i] & 0x7FFF) == lkey) {
-			ret = -EEXIST;
-			goto bail;
-		}
-	}
-	if (!any) {
-		ret = -EBUSY;
-		goto bail;
-	}
-	for (i = 0; i < ARRAY_SIZE(dd->ipath_pkeys); i++) {
-		if (!dd->ipath_pkeys[i] &&
-		    atomic_inc_return(&dd->ipath_pkeyrefs[i]) == 1) {
-			/* for ipathstats, etc. */
-			ipath_stats.sps_pkeys[i] = lkey;
-			dd->ipath_pkeys[i] = key;
-			ret = 1;
-			goto bail;
-		}
-	}
-	ret = -EBUSY;
-
-bail:
-	return ret;
-}
-
-/**
- * ipath_layer_set_pkeys - set the PKEY table for port 0
- * @dd: the infinipath device
- * @pkeys: the PKEY table
- */
-int ipath_layer_set_pkeys(struct ipath_devdata *dd, u16 * pkeys)
-{
-	struct ipath_portdata *pd;
-	int i;
-	int changed = 0;
-
-	pd = dd->ipath_pd[0];
-
-	for (i = 0; i < ARRAY_SIZE(pd->port_pkeys); i++) {
-		u16 key = pkeys[i];
-		u16 okey = pd->port_pkeys[i];
-
-		if (key == okey)
-			continue;
-		/*
-		 * The value of this PKEY table entry is changing.
-		 * Remove the old entry in the hardware's array of PKEYs.
-		 */
-		if (okey & 0x7FFF)
-			changed |= rm_pkey(dd, okey);
-		if (key & 0x7FFF) {
-			int ret = add_pkey(dd, key);
-
-			if (ret < 0)
-				key = 0;
-			else
-				changed |= ret;
-		}
-		pd->port_pkeys[i] = key;
-	}
-	if (changed) {
-		u64 pkey;
-
-		pkey = (u64) dd->ipath_pkeys[0] |
-			((u64) dd->ipath_pkeys[1] << 16) |
-			((u64) dd->ipath_pkeys[2] << 32) |
-			((u64) dd->ipath_pkeys[3] << 48);
-		ipath_cdbg(VERBOSE, "p0 new pkey reg %llx\n",
-			   (unsigned long long) pkey);
-		ipath_write_kreg(dd, dd->ipath_kregs->kr_partitionkey,
-				 pkey);
-	}
-	return 0;
-}
-
-/**
- * ipath_layer_get_linkdowndefaultstate - get the default linkdown state
- * @dd: the infinipath device
- *
- * Returns zero if the default is POLL, 1 if the default is SLEEP.
- */
-int ipath_layer_get_linkdowndefaultstate(struct ipath_devdata *dd)
-{
-	return !!(dd->ipath_ibcctrl & INFINIPATH_IBCC_LINKDOWNDEFAULTSTATE);
-}
-
-/**
- * ipath_layer_set_linkdowndefaultstate - set the default linkdown state
- * @dd: the infinipath device
- * @sleep: the new state
- *
- * Note that this will only take effect when the link state changes.
- */
-int ipath_layer_set_linkdowndefaultstate(struct ipath_devdata *dd,
-					 int sleep)
-{
-	if (sleep)
-		dd->ipath_ibcctrl |= INFINIPATH_IBCC_LINKDOWNDEFAULTSTATE;
-	else
-		dd->ipath_ibcctrl &= ~INFINIPATH_IBCC_LINKDOWNDEFAULTSTATE;
-	ipath_write_kreg(dd, dd->ipath_kregs->kr_ibcctrl,
-			 dd->ipath_ibcctrl);
-	return 0;
-}
-
-int ipath_layer_get_phyerrthreshold(struct ipath_devdata *dd)
-{
-	return (dd->ipath_ibcctrl >>
-		INFINIPATH_IBCC_PHYERRTHRESHOLD_SHIFT) &
-		INFINIPATH_IBCC_PHYERRTHRESHOLD_MASK;
-}
-
-/**
- * ipath_layer_set_phyerrthreshold - set the physical error threshold
- * @dd: the infinipath device
- * @n: the new threshold
- *
- * Note that this will only take effect when the link state changes.
- */
-int ipath_layer_set_phyerrthreshold(struct ipath_devdata *dd, unsigned n)
-{
-	unsigned v;
-
-	v = (dd->ipath_ibcctrl >> INFINIPATH_IBCC_PHYERRTHRESHOLD_SHIFT) &
-		INFINIPATH_IBCC_PHYERRTHRESHOLD_MASK;
-	if (v != n) {
-		dd->ipath_ibcctrl &=
-			~(INFINIPATH_IBCC_PHYERRTHRESHOLD_MASK <<
-			  INFINIPATH_IBCC_PHYERRTHRESHOLD_SHIFT);
-		dd->ipath_ibcctrl |=
-			(u64) n << INFINIPATH_IBCC_PHYERRTHRESHOLD_SHIFT;
-		ipath_write_kreg(dd, dd->ipath_kregs->kr_ibcctrl,
-				 dd->ipath_ibcctrl);
-	}
-	return 0;
-}
-
-int ipath_layer_get_overrunthreshold(struct ipath_devdata *dd)
-{
-	return (dd->ipath_ibcctrl >>
-		INFINIPATH_IBCC_OVERRUNTHRESHOLD_SHIFT) &
-		INFINIPATH_IBCC_OVERRUNTHRESHOLD_MASK;
-}
-
-/**
- * ipath_layer_set_overrunthreshold - set the overrun threshold
- * @dd: the infinipath device
- * @n: the new threshold
- *
- * Note that this will only take effect when the link state changes.
- */
-int ipath_layer_set_overrunthreshold(struct ipath_devdata *dd, unsigned n)
-{
-	unsigned v;
-
-	v = (dd->ipath_ibcctrl >> INFINIPATH_IBCC_OVERRUNTHRESHOLD_SHIFT) &
-		INFINIPATH_IBCC_OVERRUNTHRESHOLD_MASK;
-	if (v != n) {
-		dd->ipath_ibcctrl &=
-			~(INFINIPATH_IBCC_OVERRUNTHRESHOLD_MASK <<
-			  INFINIPATH_IBCC_OVERRUNTHRESHOLD_SHIFT);
-		dd->ipath_ibcctrl |=
-			(u64) n << INFINIPATH_IBCC_OVERRUNTHRESHOLD_SHIFT;
-		ipath_write_kreg(dd, dd->ipath_kregs->kr_ibcctrl,
-				 dd->ipath_ibcctrl);
-	}
-	return 0;
-}
-
-int ipath_layer_get_boardname(struct ipath_devdata *dd, char *name,
-			      size_t namelen)
-{
-	return dd->ipath_f_get_boardname(dd, name, namelen);
-}
-
-u32 ipath_layer_get_rcvhdrentsize(struct ipath_devdata *dd)
-{
-	return dd->ipath_rcvhdrentsize;
-}
diff --git a/drivers/infiniband/hw/ipath/ipath_layer.h b/drivers/infiniband/hw/ipath/ipath_layer.h
--- a/drivers/infiniband/hw/ipath/ipath_layer.h	Fri Aug 25 11:19:45 2006 -0700
+++ b/drivers/infiniband/hw/ipath/ipath_layer.h	Fri Aug 25 11:19:45 2006 -0700
@@ -40,72 +40,8 @@
  */
 
 struct sk_buff;
-struct ipath_sge_state;
 struct ipath_devdata;
 struct ether_header;
-
-struct ipath_layer_counters {
-	u64 symbol_error_counter;
-	u64 link_error_recovery_counter;
-	u64 link_downed_counter;
-	u64 port_rcv_errors;
-	u64 port_rcv_remphys_errors;
-	u64 port_xmit_discards;
-	u64 port_xmit_data;
-	u64 port_rcv_data;
-	u64 port_xmit_packets;
-	u64 port_rcv_packets;
-	u32 local_link_integrity_errors;
-	u32 excessive_buffer_overrun_errors;
-};
-
-/*
- * A segment is a linear region of low physical memory.
- * XXX Maybe we should use phys addr here and kmap()/kunmap().
- * Used by the verbs layer.
- */
-struct ipath_seg {
-	void *vaddr;
-	size_t length;
-};
-
-/* The number of ipath_segs that fit in a page. */
-#define IPATH_SEGSZ     (PAGE_SIZE / sizeof (struct ipath_seg))
-
-struct ipath_segarray {
-	struct ipath_seg segs[IPATH_SEGSZ];
-};
-
-struct ipath_mregion {
-	u64 user_base;		/* User's address for this region */
-	u64 iova;		/* IB start address of this region */
-	size_t length;
-	u32 lkey;
-	u32 offset;		/* offset (bytes) to start of region */
-	int access_flags;
-	u32 max_segs;		/* number of ipath_segs in all the arrays */
-	u32 mapsz;		/* size of the map array */
-	struct ipath_segarray *map[0];	/* the segments */
-};
-
-/*
- * These keep track of the copy progress within a memory region.
- * Used by the verbs layer.
- */
-struct ipath_sge {
-	struct ipath_mregion *mr;
-	void *vaddr;		/* current pointer into the segment */
-	u32 sge_length;		/* length of the SGE */
-	u32 length;		/* remaining length of the segment */
-	u16 m;			/* current index: mr->map[m] */
-	u16 n;			/* current index: mr->map[m]->segs[n] */
-};
-
-struct ipath_sge_state {
-	struct ipath_sge *sg_list;	/* next SGE to be used if any */
-	struct ipath_sge sge;	/* progress state for the current SGE */
-	u8 num_sge;
-};
 
 int ipath_layer_register(void *(*l_add)(int, struct ipath_devdata *),
 			 void (*l_remove)(void *),
@@ -119,49 +55,9 @@ u16 ipath_layer_get_lid(struct ipath_dev
 u16 ipath_layer_get_lid(struct ipath_devdata *dd);
 int ipath_layer_get_mac(struct ipath_devdata *dd, u8 *);
 u16 ipath_layer_get_bcast(struct ipath_devdata *dd);
-u32 ipath_layer_get_cr_errpkey(struct ipath_devdata *dd);
-int ipath_layer_set_linkstate(struct ipath_devdata *dd, u8 state);
-int ipath_layer_set_mtu(struct ipath_devdata *, u16);
-int ipath_set_lid(struct ipath_devdata *, u32, u8);
 int ipath_layer_send_hdr(struct ipath_devdata *dd,
 			 struct ether_header *hdr);
-int ipath_verbs_send(struct ipath_devdata *dd, u32 hdrwords,
-		     u32 * hdr, u32 len, struct ipath_sge_state *ss);
 int ipath_layer_set_piointbufavail_int(struct ipath_devdata *dd);
-int ipath_layer_get_boardname(struct ipath_devdata *dd, char *name,
-			      size_t namelen);
-int ipath_layer_snapshot_counters(struct ipath_devdata *dd, u64 *swords,
-				  u64 *rwords, u64 *spkts, u64 *rpkts,
-				  u64 *xmit_wait);
-int ipath_layer_get_counters(struct ipath_devdata *dd,
-			     struct ipath_layer_counters *cntrs);
-int ipath_layer_want_buffer(struct ipath_devdata *dd);
-int ipath_layer_set_guid(struct ipath_devdata *, __be64 guid);
-__be64 ipath_layer_get_guid(struct ipath_devdata *);
-u32 ipath_layer_get_majrev(struct ipath_devdata *);
-u32 ipath_layer_get_minrev(struct ipath_devdata *);
-u32 ipath_layer_get_pcirev(struct ipath_devdata *);
-u32 ipath_layer_get_flags(struct ipath_devdata *dd);
-struct device *ipath_layer_get_device(struct ipath_devdata *dd);
-u16 ipath_layer_get_deviceid(struct ipath_devdata *dd);
-u32 ipath_layer_get_vendorid(struct ipath_devdata *);
-u64 ipath_layer_get_lastibcstat(struct ipath_devdata *dd);
-u32 ipath_layer_get_ibmtu(struct ipath_devdata *dd);
-int ipath_layer_enable_timer(struct ipath_devdata *dd);
-int ipath_layer_disable_timer(struct ipath_devdata *dd);
-int ipath_layer_set_verbs_flags(struct ipath_devdata *dd, unsigned flags);
-unsigned ipath_layer_get_npkeys(struct ipath_devdata *dd);
-unsigned ipath_layer_get_pkey(struct ipath_devdata *dd, unsigned index);
-int ipath_layer_get_pkeys(struct ipath_devdata *dd, u16 *pkeys);
-int ipath_layer_set_pkeys(struct ipath_devdata *dd, u16 *pkeys);
-int ipath_layer_get_linkdowndefaultstate(struct ipath_devdata *dd);
-int ipath_layer_set_linkdowndefaultstate(struct ipath_devdata *dd,
-					 int sleep);
-int ipath_layer_get_phyerrthreshold(struct ipath_devdata *dd);
-int ipath_layer_set_phyerrthreshold(struct ipath_devdata *dd, unsigned n);
-int ipath_layer_get_overrunthreshold(struct ipath_devdata *dd);
-int ipath_layer_set_overrunthreshold(struct ipath_devdata *dd, unsigned n);
-u32 ipath_layer_get_rcvhdrentsize(struct ipath_devdata *dd);
 
 /* ipath_ether interrupt values */
 #define IPATH_LAYER_INT_IF_UP 0x2
diff --git a/drivers/infiniband/hw/ipath/ipath_mad.c b/drivers/infiniband/hw/ipath/ipath_mad.c
--- a/drivers/infiniband/hw/ipath/ipath_mad.c	Fri Aug 25 11:19:45 2006 -0700
+++ b/drivers/infiniband/hw/ipath/ipath_mad.c	Fri Aug 25 11:19:45 2006 -0700
@@ -101,15 +101,15 @@ static int recv_subn_get_nodeinfo(struct
 	nip->num_ports = ibdev->phys_port_cnt;
 	/* This is already in network order */
 	nip->sys_guid = to_idev(ibdev)->sys_image_guid;
-	nip->node_guid = ipath_layer_get_guid(dd);
+	nip->node_guid = dd->ipath_guid;
 	nip->port_guid = nip->sys_guid;
-	nip->partition_cap = cpu_to_be16(ipath_layer_get_npkeys(dd));
-	nip->device_id = cpu_to_be16(ipath_layer_get_deviceid(dd));
-	majrev = ipath_layer_get_majrev(dd);
-	minrev = ipath_layer_get_minrev(dd);
+	nip->partition_cap = cpu_to_be16(ipath_get_npkeys(dd));
+	nip->device_id = cpu_to_be16(dd->ipath_deviceid);
+	majrev = dd->ipath_majrev;
+	minrev = dd->ipath_minrev;
 	nip->revision = cpu_to_be32((majrev << 16) | minrev);
 	nip->local_port_num = port;
-	vendor = ipath_layer_get_vendorid(dd);
+	vendor = dd->ipath_vendorid;
 	nip->vendor_id[0] = 0;
 	nip->vendor_id[1] = vendor >> 8;
 	nip->vendor_id[2] = vendor;
@@ -133,11 +133,87 @@ static int recv_subn_get_guidinfo(struct
 	 */
 	if (startgx == 0)
 		/* The first is a copy of the read-only HW GUID. */
-		*p = ipath_layer_get_guid(to_idev(ibdev)->dd);
+		*p = to_idev(ibdev)->dd->ipath_guid;
 	else
 		smp->status |= IB_SMP_INVALID_FIELD;
 
 	return reply(smp);
+}
+
+
+static int get_overrunthreshold(struct ipath_devdata *dd)
+{
+	return (dd->ipath_ibcctrl >>
+		INFINIPATH_IBCC_OVERRUNTHRESHOLD_SHIFT) &
+		INFINIPATH_IBCC_OVERRUNTHRESHOLD_MASK;
+}
+
+/**
+ * set_overrunthreshold - set the overrun threshold
+ * @dd: the infinipath device
+ * @n: the new threshold
+ *
+ * Note that this will only take effect when the link state changes.
+ */
+static int set_overrunthreshold(struct ipath_devdata *dd, unsigned n)
+{
+	unsigned v;
+
+	v = (dd->ipath_ibcctrl >> INFINIPATH_IBCC_OVERRUNTHRESHOLD_SHIFT) &
+		INFINIPATH_IBCC_OVERRUNTHRESHOLD_MASK;
+	if (v != n) {
+		dd->ipath_ibcctrl &=
+			~(INFINIPATH_IBCC_OVERRUNTHRESHOLD_MASK <<
+			  INFINIPATH_IBCC_OVERRUNTHRESHOLD_SHIFT);
+		dd->ipath_ibcctrl |=
+			(u64) n << INFINIPATH_IBCC_OVERRUNTHRESHOLD_SHIFT;
+		ipath_write_kreg(dd, dd->ipath_kregs->kr_ibcctrl,
+				 dd->ipath_ibcctrl);
+	}
+	return 0;
+}
+
+static int get_phyerrthreshold(struct ipath_devdata *dd)
+{
+	return (dd->ipath_ibcctrl >>
+		INFINIPATH_IBCC_PHYERRTHRESHOLD_SHIFT) &
+		INFINIPATH_IBCC_PHYERRTHRESHOLD_MASK;
+}
+
+/**
+ * set_phyerrthreshold - set the physical error threshold
+ * @dd: the infinipath device
+ * @n: the new threshold
+ *
+ * Note that this will only take effect when the link state changes.
+ */
+static int set_phyerrthreshold(struct ipath_devdata *dd, unsigned n)
+{
+	unsigned v;
+
+	v = (dd->ipath_ibcctrl >> INFINIPATH_IBCC_PHYERRTHRESHOLD_SHIFT) &
+		INFINIPATH_IBCC_PHYERRTHRESHOLD_MASK;
+	if (v != n) {
+		dd->ipath_ibcctrl &=
+			~(INFINIPATH_IBCC_PHYERRTHRESHOLD_MASK <<
+			  INFINIPATH_IBCC_PHYERRTHRESHOLD_SHIFT);
+		dd->ipath_ibcctrl |=
+			(u64) n << INFINIPATH_IBCC_PHYERRTHRESHOLD_SHIFT;
+		ipath_write_kreg(dd, dd->ipath_kregs->kr_ibcctrl,
+				 dd->ipath_ibcctrl);
+	}
+	return 0;
+}
+
+/**
+ * get_linkdowndefaultstate - get the default linkdown state
+ * @dd: the infinipath device
+ *
+ * Returns zero if the default is POLL, 1 if the default is SLEEP.
+ */
+static int get_linkdowndefaultstate(struct ipath_devdata *dd)
+{
+	return !!(dd->ipath_ibcctrl & INFINIPATH_IBCC_LINKDOWNDEFAULTSTATE);
 }
 
 static int recv_subn_get_portinfo(struct ib_smp *smp,
@@ -166,7 +242,7 @@ static int recv_subn_get_portinfo(struct
 	    (dev->mkeyprot_resv_lmc >> 6) == 0)
 		pip->mkey = dev->mkey;
 	pip->gid_prefix = dev->gid_prefix;
-	lid = ipath_layer_get_lid(dev->dd);
+	lid = dev->dd->ipath_lid;
 	pip->lid = lid ? cpu_to_be16(lid) : IB_LID_PERMISSIVE;
 	pip->sm_lid = cpu_to_be16(dev->sm_lid);
 	pip->cap_mask = cpu_to_be32(dev->port_cap_flags);
@@ -177,14 +253,14 @@ static int recv_subn_get_portinfo(struct
 	pip->link_width_supported = 3;	/* 1x or 4x */
 	pip->link_width_active = 2;	/* 4x */
 	pip->linkspeed_portstate = 0x10;	/* 2.5Gbps */
-	ibcstat = ipath_layer_get_lastibcstat(dev->dd);
+	ibcstat = dev->dd->ipath_lastibcstat;
 	pip->linkspeed_portstate |= ((ibcstat >> 4) & 0x3) + 1;
 	pip->portphysstate_linkdown =
 		(ipath_cvt_physportstate[ibcstat & 0xf] << 4) |
-		(ipath_layer_get_linkdowndefaultstate(dev->dd) ? 1 : 2);
+		(get_linkdowndefaultstate(dev->dd) ? 1 : 2);
 	pip->mkeyprot_resv_lmc = dev->mkeyprot_resv_lmc;
 	pip->linkspeedactive_enabled = 0x11;	/* 2.5Gbps, 2.5Gbps */
-	switch (ipath_layer_get_ibmtu(dev->dd)) {
+	switch (dev->dd->ipath_ibmtu) {
 	case 4096:
 		mtu = IB_MTU_4096;
 		break;
@@ -217,7 +293,7 @@ static int recv_subn_get_portinfo(struct
 	pip->mkey_violations = cpu_to_be16(dev->mkey_violations);
 	/* P_KeyViolations are counted by hardware. */
 	pip->pkey_violations =
-		cpu_to_be16((ipath_layer_get_cr_errpkey(dev->dd) -
+		cpu_to_be16((ipath_get_cr_errpkey(dev->dd) -
 			     dev->z_pkey_violations) & 0xFFFF);
 	pip->qkey_violations = cpu_to_be16(dev->qkey_violations);
 	/* Only the hardware GUID is supported for now */
@@ -226,8 +302,8 @@ static int recv_subn_get_portinfo(struct
 	/* 32.768 usec. response time (guessing) */
 	pip->resv_resptimevalue = 3;
 	pip->localphyerrors_overrunerrors =
-		(ipath_layer_get_phyerrthreshold(dev->dd) << 4) |
-		ipath_layer_get_overrunthreshold(dev->dd);
+		(get_phyerrthreshold(dev->dd) << 4) |
+		get_overrunthreshold(dev->dd);
 	/* pip->max_credit_hint; */
 	/* pip->link_roundtrip_latency[3]; */
 
@@ -235,6 +311,20 @@ static int recv_subn_get_portinfo(struct
 
 bail:
 	return ret;
+}
+
+/**
+ * get_pkeys - return the PKEY table for port 0
+ * @dd: the infinipath device
+ * @pkeys: the pkey table is placed here
+ */
+static int get_pkeys(struct ipath_devdata *dd, u16 * pkeys)
+{
+	struct ipath_portdata *pd = dd->ipath_pd[0];
+
+	memcpy(pkeys, pd->port_pkeys, sizeof(pd->port_pkeys));
+
+	return 0;
 }
 
 static int recv_subn_get_pkeytable(struct ib_smp *smp,
@@ -249,9 +339,9 @@ static int recv_subn_get_pkeytable(struc
 	memset(smp->data, 0, sizeof(smp->data));
 	if (startpx == 0) {
 		struct ipath_ibdev *dev = to_idev(ibdev);
-		unsigned i, n = ipath_layer_get_npkeys(dev->dd);
-
-		ipath_layer_get_pkeys(dev->dd, p);
+		unsigned i, n = ipath_get_npkeys(dev->dd);
+
+		get_pkeys(dev->dd, p);
 
 		for (i = 0; i < n; i++)
 			q[i] = cpu_to_be16(p[i]);
@@ -266,6 +356,24 @@ static int recv_subn_set_guidinfo(struct
 {
 	/* The only GUID we support is the first read-only entry. */
 	return recv_subn_get_guidinfo(smp, ibdev);
+}
+
+/**
+ * set_linkdowndefaultstate - set the default linkdown state
+ * @dd: the infinipath device
+ * @sleep: the new state
+ *
+ * Note that this will only take effect when the link state changes.
+ */
+static int set_linkdowndefaultstate(struct ipath_devdata *dd, int sleep)
+{
+	if (sleep)
+		dd->ipath_ibcctrl |= INFINIPATH_IBCC_LINKDOWNDEFAULTSTATE;
+	else
+		dd->ipath_ibcctrl &= ~INFINIPATH_IBCC_LINKDOWNDEFAULTSTATE;
+	ipath_write_kreg(dd, dd->ipath_kregs->kr_ibcctrl,
+			 dd->ipath_ibcctrl);
+	return 0;
 }
 
 /**
@@ -290,7 +398,7 @@ static int recv_subn_set_portinfo(struct
 	u8 state;
 	u16 lstate;
 	u32 mtu;
-	int ret;
+	int ret, ore;
 
 	if (be32_to_cpu(smp->attr_mod) > ibdev->phys_port_cnt)
 		goto err;
@@ -304,7 +412,7 @@ static int recv_subn_set_portinfo(struct
 	dev->mkey_lease_period = be16_to_cpu(pip->mkey_lease_period);
 
 	lid = be16_to_cpu(pip->lid);
-	if (lid != ipath_layer_get_lid(dev->dd)) {
+	if (lid != dev->dd->ipath_lid) {
 		/* Must be a valid unicast LID address. */
 		if (lid == 0 || lid >= IPATH_MULTICAST_LID_BASE)
 			goto err;
@@ -342,11 +450,11 @@ static int recv_subn_set_portinfo(struct
 	case 0: /* NOP */
 		break;
 	case 1: /* SLEEP */
-		if (ipath_layer_set_linkdowndefaultstate(dev->dd, 1))
+		if (set_linkdowndefaultstate(dev->dd, 1))
 			goto err;
 		break;
 	case 2: /* POLL */
-		if (ipath_layer_set_linkdowndefaultstate(dev->dd, 0))
+		if (set_linkdowndefaultstate(dev->dd, 0))
 			goto err;
 		break;
 	default:
@@ -376,7 +484,7 @@ static int recv_subn_set_portinfo(struct
 		/* XXX We have already partially updated our state! */
 		goto err;
 	}
-	ipath_layer_set_mtu(dev->dd, mtu);
+	ipath_set_mtu(dev->dd, mtu);
 
 	dev->sm_sl = pip->neighbormtu_mastersmsl & 0xF;
 
@@ -392,20 +500,16 @@ static int recv_subn_set_portinfo(struct
 	 * later.
 	 */
 	if (pip->pkey_violations == 0)
-		dev->z_pkey_violations =
-			ipath_layer_get_cr_errpkey(dev->dd);
+		dev->z_pkey_violations = ipath_get_cr_errpkey(dev->dd);
 
 	if (pip->qkey_violations == 0)
 		dev->qkey_violations = 0;
 
-	if (ipath_layer_set_phyerrthreshold(
-		    dev->dd,
-		    (pip->localphyerrors_overrunerrors >> 4) & 0xF))
+	ore = pip->localphyerrors_overrunerrors;
+	if (set_phyerrthreshold(dev->dd, (ore >> 4) & 0xF))
 		goto err;
 
-	if (ipath_layer_set_overrunthreshold(
-		    dev->dd,
-		    (pip->localphyerrors_overrunerrors & 0xF)))
+	if (set_overrunthreshold(dev->dd, (ore & 0xF)))
 		goto err;
 
 	dev->subnet_timeout = pip->clientrereg_resv_subnetto & 0x1F;
@@ -423,7 +527,7 @@ static int recv_subn_set_portinfo(struct
 	 * is down or is being set to down.
 	 */
 	state = pip->linkspeed_portstate & 0xF;
-	flags = ipath_layer_get_flags(dev->dd);
+	flags = dev->dd->ipath_flags;
 	lstate = (pip->portphysstate_linkdown >> 4) & 0xF;
 	if (lstate && !(state == IB_PORT_DOWN || state == IB_PORT_NOP))
 		goto err;
@@ -439,7 +543,7 @@ static int recv_subn_set_portinfo(struct
 		/* FALLTHROUGH */
 	case IB_PORT_DOWN:
 		if (lstate == 0)
-			if (ipath_layer_get_linkdowndefaultstate(dev->dd))
+			if (get_linkdowndefaultstate(dev->dd))
 				lstate = IPATH_IB_LINKDOWN_SLEEP;
 			else
 				lstate = IPATH_IB_LINKDOWN;
@@ -451,7 +555,7 @@ static int recv_subn_set_portinfo(struct
 			lstate = IPATH_IB_LINKDOWN_DISABLE;
 		else
 			goto err;
-		ipath_layer_set_linkstate(dev->dd, lstate);
+		ipath_set_linkstate(dev->dd, lstate);
 		if (flags & IPATH_LINKACTIVE) {
 			event.event = IB_EVENT_PORT_ERR;
 			ib_dispatch_event(&event);
@@ -460,7 +564,7 @@ static int recv_subn_set_portinfo(struct
 	case IB_PORT_ARMED:
 		if (!(flags & (IPATH_LINKINIT | IPATH_LINKACTIVE)))
 			break;
-		ipath_layer_set_linkstate(dev->dd, IPATH_IB_LINKARM);
+		ipath_set_linkstate(dev->dd, IPATH_IB_LINKARM);
 		if (flags & IPATH_LINKACTIVE) {
 			event.event = IB_EVENT_PORT_ERR;
 			ib_dispatch_event(&event);
@@ -469,7 +573,7 @@ static int recv_subn_set_portinfo(struct
 	case IB_PORT_ACTIVE:
 		if (!(flags & IPATH_LINKARMED))
 			break;
-		ipath_layer_set_linkstate(dev->dd, IPATH_IB_LINKACTIVE);
+		ipath_set_linkstate(dev->dd, IPATH_IB_LINKACTIVE);
 		event.event = IB_EVENT_PORT_ACTIVE;
 		ib_dispatch_event(&event);
 		break;
@@ -491,6 +595,152 @@ err:
 
 done:
 	return ret;
+}
+
+/**
+ * rm_pkey - decrecment the reference count for the given PKEY
+ * @dd: the infinipath device
+ * @key: the PKEY index
+ *
+ * Return true if this was the last reference and the hardware table entry
+ * needs to be changed.
+ */
+static int rm_pkey(struct ipath_devdata *dd, u16 key)
+{
+	int i;
+	int ret;
+
+	for (i = 0; i < ARRAY_SIZE(dd->ipath_pkeys); i++) {
+		if (dd->ipath_pkeys[i] != key)
+			continue;
+		if (atomic_dec_and_test(&dd->ipath_pkeyrefs[i])) {
+			dd->ipath_pkeys[i] = 0;
+			ret = 1;
+			goto bail;
+		}
+		break;
+	}
+
+	ret = 0;
+
+bail:
+	return ret;
+}
+
+/**
+ * add_pkey - add the given PKEY to the hardware table
+ * @dd: the infinipath device
+ * @key: the PKEY
+ *
+ * Return an error code if unable to add the entry, zero if no change,
+ * or 1 if the hardware PKEY register needs to be updated.
+ */
+static int add_pkey(struct ipath_devdata *dd, u16 key)
+{
+	int i;
+	u16 lkey = key & 0x7FFF;
+	int any = 0;
+	int ret;
+
+	if (lkey == 0x7FFF) {
+		ret = 0;
+		goto bail;
+	}
+
+	/* Look for an empty slot or a matching PKEY. */
+	for (i = 0; i < ARRAY_SIZE(dd->ipath_pkeys); i++) {
+		if (!dd->ipath_pkeys[i]) {
+			any++;
+			continue;
+		}
+		/* If it matches exactly, try to increment the ref count */
+		if (dd->ipath_pkeys[i] == key) {
+			if (atomic_inc_return(&dd->ipath_pkeyrefs[i]) > 1) {
+				ret = 0;
+				goto bail;
+			}
+			/* Lost the race. Look for an empty slot below. */
+			atomic_dec(&dd->ipath_pkeyrefs[i]);
+			any++;
+		}
+		/*
+		 * It makes no sense to have both the limited and unlimited
+		 * PKEY set at the same time since the unlimited one will
+		 * disable the limited one.
+		 */
+		if ((dd->ipath_pkeys[i] & 0x7FFF) == lkey) {
+			ret = -EEXIST;
+			goto bail;
+		}
+	}
+	if (!any) {
+		ret = -EBUSY;
+		goto bail;
+	}
+	for (i = 0; i < ARRAY_SIZE(dd->ipath_pkeys); i++) {
+		if (!dd->ipath_pkeys[i] &&
+		    atomic_inc_return(&dd->ipath_pkeyrefs[i]) == 1) {
+			/* for ipathstats, etc. */
+			ipath_stats.sps_pkeys[i] = lkey;
+			dd->ipath_pkeys[i] = key;
+			ret = 1;
+			goto bail;
+		}
+	}
+	ret = -EBUSY;
+
+bail:
+	return ret;
+}
+
+/**
+ * set_pkeys - set the PKEY table for port 0
+ * @dd: the infinipath device
+ * @pkeys: the PKEY table
+ */
+static int set_pkeys(struct ipath_devdata *dd, u16 *pkeys)
+{
+	struct ipath_portdata *pd;
+	int i;
+	int changed = 0;
+
+	pd = dd->ipath_pd[0];
+
+	for (i = 0; i < ARRAY_SIZE(pd->port_pkeys); i++) {
+		u16 key = pkeys[i];
+		u16 okey = pd->port_pkeys[i];
+
+		if (key == okey)
+			continue;
+		/*
+		 * The value of this PKEY table entry is changing.
+		 * Remove the old entry in the hardware's array of PKEYs.
+		 */
+		if (okey & 0x7FFF)
+			changed |= rm_pkey(dd, okey);
+		if (key & 0x7FFF) {
+			int ret = add_pkey(dd, key);
+
+			if (ret < 0)
+				key = 0;
+			else
+				changed |= ret;
+		}
+		pd->port_pkeys[i] = key;
+	}
+	if (changed) {
+		u64 pkey;
+
+		pkey = (u64) dd->ipath_pkeys[0] |
+			((u64) dd->ipath_pkeys[1] << 16) |
+			((u64) dd->ipath_pkeys[2] << 32) |
+			((u64) dd->ipath_pkeys[3] << 48);
+		ipath_cdbg(VERBOSE, "p0 new pkey reg %llx\n",
+			   (unsigned long long) pkey);
+		ipath_write_kreg(dd, dd->ipath_kregs->kr_partitionkey,
+				 pkey);
+	}
+	return 0;
 }
 
 static int recv_subn_set_pkeytable(struct ib_smp *smp,
@@ -500,13 +750,12 @@ static int recv_subn_set_pkeytable(struc
 	__be16 *p = (__be16 *) smp->data;
 	u16 *q = (u16 *) smp->data;
 	struct ipath_ibdev *dev = to_idev(ibdev);
-	unsigned i, n = ipath_layer_get_npkeys(dev->dd);
+	unsigned i, n = ipath_get_npkeys(dev->dd);
 
 	for (i = 0; i < n; i++)
 		q[i] = be16_to_cpu(p[i]);
 
-	if (startpx != 0 ||
-	    ipath_layer_set_pkeys(dev->dd, q) != 0)
+	if (startpx != 0 || set_pkeys(dev->dd, q) != 0)
 		smp->status |= IB_SMP_INVALID_FIELD;
 
 	return recv_subn_get_pkeytable(smp, ibdev);
@@ -844,10 +1093,10 @@ static int recv_pma_get_portcounters(str
 	struct ib_pma_portcounters *p = (struct ib_pma_portcounters *)
 		pmp->data;
 	struct ipath_ibdev *dev = to_idev(ibdev);
-	struct ipath_layer_counters cntrs;
+	struct ipath_verbs_counters cntrs;
 	u8 port_select = p->port_select;
 
-	ipath_layer_get_counters(dev->dd, &cntrs);
+	ipath_get_counters(dev->dd, &cntrs);
 
 	/* Adjust counters for any resets done. */
 	cntrs.symbol_error_counter -= dev->z_symbol_error_counter;
@@ -944,8 +1193,8 @@ static int recv_pma_get_portcounters_ext
 	u64 swords, rwords, spkts, rpkts, xwait;
 	u8 port_select = p->port_select;
 
-	ipath_layer_snapshot_counters(dev->dd, &swords, &rwords, &spkts,
-				      &rpkts, &xwait);
+	ipath_snapshot_counters(dev->dd, &swords, &rwords, &spkts,
+				&rpkts, &xwait);
 
 	/* Adjust counters for any resets done. */
 	swords -= dev->z_port_xmit_data;
@@ -978,13 +1227,13 @@ static int recv_pma_set_portcounters(str
 	struct ib_pma_portcounters *p = (struct ib_pma_portcounters *)
 		pmp->data;
 	struct ipath_ibdev *dev = to_idev(ibdev);
-	struct ipath_layer_counters cntrs;
+	struct ipath_verbs_counters cntrs;
 
 	/*
 	 * Since the HW doesn't support clearing counters, we save the
 	 * current count and subtract it from future responses.
 	 */
-	ipath_layer_get_counters(dev->dd, &cntrs);
+	ipath_get_counters(dev->dd, &cntrs);
 
 	if (p->counter_select & IB_PMA_SEL_SYMBOL_ERROR)
 		dev->z_symbol_error_counter = cntrs.symbol_error_counter;
@@ -1041,8 +1290,8 @@ static int recv_pma_set_portcounters_ext
 	struct ipath_ibdev *dev = to_idev(ibdev);
 	u64 swords, rwords, spkts, rpkts, xwait;
 
-	ipath_layer_snapshot_counters(dev->dd, &swords, &rwords, &spkts,
-				      &rpkts, &xwait);
+	ipath_snapshot_counters(dev->dd, &swords, &rwords, &spkts,
+				&rpkts, &xwait);
 
 	if (p->counter_select & IB_PMA_SELX_PORT_XMIT_DATA)
 		dev->z_port_xmit_data = swords;
diff --git a/drivers/infiniband/hw/ipath/ipath_mr.c b/drivers/infiniband/hw/ipath/ipath_mr.c
--- a/drivers/infiniband/hw/ipath/ipath_mr.c	Fri Aug 25 11:19:45 2006 -0700
+++ b/drivers/infiniband/hw/ipath/ipath_mr.c	Fri Aug 25 11:19:45 2006 -0700
@@ -35,6 +35,18 @@
 #include <rdma/ib_smi.h>
 
 #include "ipath_verbs.h"
+
+/* Fast memory region */
+struct ipath_fmr {
+	struct ib_fmr ibfmr;
+	u8 page_shift;
+	struct ipath_mregion mr;        /* must be last */
+};
+
+static inline struct ipath_fmr *to_ifmr(struct ib_fmr *ibfmr)
+{
+	return container_of(ibfmr, struct ipath_fmr, ibfmr);
+}
 
 /**
  * ipath_get_dma_mr - get a DMA memory region
diff --git a/drivers/infiniband/hw/ipath/ipath_qp.c b/drivers/infiniband/hw/ipath/ipath_qp.c
--- a/drivers/infiniband/hw/ipath/ipath_qp.c	Fri Aug 25 11:19:45 2006 -0700
+++ b/drivers/infiniband/hw/ipath/ipath_qp.c	Fri Aug 25 11:19:45 2006 -0700
@@ -461,7 +461,7 @@ int ipath_modify_qp(struct ib_qp *ibqp, 
 			goto inval;
 
 	if (attr_mask & IB_QP_PKEY_INDEX)
-		if (attr->pkey_index >= ipath_layer_get_npkeys(dev->dd))
+		if (attr->pkey_index >= ipath_get_npkeys(dev->dd))
 			goto inval;
 
 	if (attr_mask & IB_QP_MIN_RNR_TIMER)
@@ -645,6 +645,33 @@ __be32 ipath_compute_aeth(struct ipath_q
 }
 
 /**
+ * set_verbs_flags - set the verbs layer flags
+ * @dd: the infinipath device
+ * @flags: the flags to set
+ */
+static int set_verbs_flags(struct ipath_devdata *dd, unsigned flags)
+{
+	struct ipath_devdata *ss;
+	unsigned long lflags;
+
+	spin_lock_irqsave(&ipath_devs_lock, lflags);
+
+	list_for_each_entry(ss, &ipath_dev_list, ipath_list) {
+		if (!(ss->ipath_flags & IPATH_INITTED))
+			continue;
+		if ((flags & IPATH_VERBS_KERNEL_SMA) &&
+		    !(*ss->ipath_statusp & IPATH_STATUS_SMA))
+			*ss->ipath_statusp |= IPATH_STATUS_OIB_SMA;
+		else
+			*ss->ipath_statusp &= ~IPATH_STATUS_OIB_SMA;
+	}
+
+	spin_unlock_irqrestore(&ipath_devs_lock, lflags);
+
+	return 0;
+}
+
+/**
  * ipath_create_qp - create a queue pair for a device
  * @ibpd: the protection domain who's device we create the queue pair for
  * @init_attr: the attributes of the queue pair
@@ -760,8 +787,7 @@ struct ib_qp *ipath_create_qp(struct ib_
 
 		/* Tell the core driver that the kernel SMA is present. */
 		if (init_attr->qp_type == IB_QPT_SMI)
-			ipath_layer_set_verbs_flags(dev->dd,
-						    IPATH_VERBS_KERNEL_SMA);
+			set_verbs_flags(dev->dd, IPATH_VERBS_KERNEL_SMA);
 		break;
 
 	default:
@@ -838,7 +864,7 @@ int ipath_destroy_qp(struct ib_qp *ibqp)
 
 	/* Tell the core driver that the kernel SMA is gone. */
 	if (qp->ibqp.qp_type == IB_QPT_SMI)
-		ipath_layer_set_verbs_flags(dev->dd, 0);
+		set_verbs_flags(dev->dd, 0);
 
 	spin_lock_irqsave(&qp->s_lock, flags);
 	qp->state = IB_QPS_ERR;
diff --git a/drivers/infiniband/hw/ipath/ipath_rc.c b/drivers/infiniband/hw/ipath/ipath_rc.c
--- a/drivers/infiniband/hw/ipath/ipath_rc.c	Fri Aug 25 11:19:45 2006 -0700
+++ b/drivers/infiniband/hw/ipath/ipath_rc.c	Fri Aug 25 11:19:45 2006 -0700
@@ -32,7 +32,7 @@
  */
 
 #include "ipath_verbs.h"
-#include "ipath_common.h"
+#include "ipath_kernel.h"
 
 /* cut down ridiculously long IB macro names */
 #define OP(x) IB_OPCODE_RC_##x
@@ -540,7 +540,7 @@ static void send_rc_ack(struct ipath_qp 
 		lrh0 = IPATH_LRH_GRH;
 	}
 	/* read pkey_index w/o lock (its atomic) */
-	bth0 = ipath_layer_get_pkey(dev->dd, qp->s_pkey_index);
+	bth0 = ipath_get_pkey(dev->dd, qp->s_pkey_index);
 	if (qp->r_nak_state)
 		ohdr->u.aeth = cpu_to_be32((qp->r_msn & IPATH_MSN_MASK) |
 					    (qp->r_nak_state <<
@@ -557,7 +557,7 @@ static void send_rc_ack(struct ipath_qp 
 	hdr.lrh[0] = cpu_to_be16(lrh0);
 	hdr.lrh[1] = cpu_to_be16(qp->remote_ah_attr.dlid);
 	hdr.lrh[2] = cpu_to_be16(hwords + SIZE_OF_CRC);
-	hdr.lrh[3] = cpu_to_be16(ipath_layer_get_lid(dev->dd));
+	hdr.lrh[3] = cpu_to_be16(dev->dd->ipath_lid);
 	ohdr->bth[0] = cpu_to_be32(bth0);
 	ohdr->bth[1] = cpu_to_be32(qp->remote_qpn);
 	ohdr->bth[2] = cpu_to_be32(qp->r_ack_psn & IPATH_PSN_MASK);
@@ -1323,8 +1323,7 @@ void ipath_rc_rcv(struct ipath_ibdev *de
 		 * the eager header buffer size to 56 bytes so the last 4
 		 * bytes of the BTH header (PSN) is in the data buffer.
 		 */
-		header_in_data =
-			ipath_layer_get_rcvhdrentsize(dev->dd) == 16;
+		header_in_data = dev->dd->ipath_rcvhdrentsize == 16;
 		if (header_in_data) {
 			psn = be32_to_cpu(((__be32 *) data)[0]);
 			data += sizeof(__be32);
diff --git a/drivers/infiniband/hw/ipath/ipath_ruc.c b/drivers/infiniband/hw/ipath/ipath_ruc.c
--- a/drivers/infiniband/hw/ipath/ipath_ruc.c	Fri Aug 25 11:19:45 2006 -0700
+++ b/drivers/infiniband/hw/ipath/ipath_ruc.c	Fri Aug 25 11:19:45 2006 -0700
@@ -470,6 +470,15 @@ done:
 		wake_up(&qp->wait);
 }
 
+static int want_buffer(struct ipath_devdata *dd)
+{
+	set_bit(IPATH_S_PIOINTBUFAVAIL, &dd->ipath_sendctrl);
+	ipath_write_kreg(dd, dd->ipath_kregs->kr_sendctrl,
+			 dd->ipath_sendctrl);
+
+	return 0;
+}
+
 /**
  * ipath_no_bufs_available - tell the layer driver we need buffers
  * @qp: the QP that caused the problem
@@ -486,7 +495,7 @@ void ipath_no_bufs_available(struct ipat
 		list_add_tail(&qp->piowait, &dev->piowait);
 	spin_unlock_irqrestore(&dev->pending_lock, flags);
 	/*
-	 * Note that as soon as ipath_layer_want_buffer() is called and
+	 * Note that as soon as want_buffer() is called and
 	 * possibly before it returns, ipath_ib_piobufavail()
 	 * could be called.  If we are still in the tasklet function,
 	 * tasklet_hi_schedule() will not call us until the next time
@@ -496,7 +505,7 @@ void ipath_no_bufs_available(struct ipat
 	 */
 	clear_bit(IPATH_S_BUSY, &qp->s_flags);
 	tasklet_unlock(&qp->s_task);
-	ipath_layer_want_buffer(dev->dd);
+	want_buffer(dev->dd);
 	dev->n_piowait++;
 }
 
@@ -611,7 +620,7 @@ u32 ipath_make_grh(struct ipath_ibdev *d
 	hdr->hop_limit = grh->hop_limit;
 	/* The SGID is 32-bit aligned. */
 	hdr->sgid.global.subnet_prefix = dev->gid_prefix;
-	hdr->sgid.global.interface_id = ipath_layer_get_guid(dev->dd);
+	hdr->sgid.global.interface_id = dev->dd->ipath_guid;
 	hdr->dgid = grh->dgid;
 
 	/* GRH header size in 32-bit words. */
@@ -643,8 +652,7 @@ void ipath_do_ruc_send(unsigned long dat
 	if (test_and_set_bit(IPATH_S_BUSY, &qp->s_flags))
 		goto bail;
 
-	if (unlikely(qp->remote_ah_attr.dlid ==
-		     ipath_layer_get_lid(dev->dd))) {
+	if (unlikely(qp->remote_ah_attr.dlid == dev->dd->ipath_lid)) {
 		ipath_ruc_loopback(qp);
 		goto clear;
 	}
@@ -711,8 +719,8 @@ again:
 	qp->s_hdr.lrh[1] = cpu_to_be16(qp->remote_ah_attr.dlid);
 	qp->s_hdr.lrh[2] = cpu_to_be16(qp->s_hdrwords + nwords +
 				       SIZE_OF_CRC);
-	qp->s_hdr.lrh[3] = cpu_to_be16(ipath_layer_get_lid(dev->dd));
-	bth0 |= ipath_layer_get_pkey(dev->dd, qp->s_pkey_index);
+	qp->s_hdr.lrh[3] = cpu_to_be16(dev->dd->ipath_lid);
+	bth0 |= ipath_get_pkey(dev->dd, qp->s_pkey_index);
 	bth0 |= extra_bytes << 20;
 	ohdr->bth[0] = cpu_to_be32(bth0);
 	ohdr->bth[1] = cpu_to_be32(qp->remote_qpn);
diff --git a/drivers/infiniband/hw/ipath/ipath_sysfs.c b/drivers/infiniband/hw/ipath/ipath_sysfs.c
--- a/drivers/infiniband/hw/ipath/ipath_sysfs.c	Fri Aug 25 11:19:45 2006 -0700
+++ b/drivers/infiniband/hw/ipath/ipath_sysfs.c	Fri Aug 25 11:19:45 2006 -0700
@@ -35,7 +35,6 @@
 #include <linux/pci.h>
 
 #include "ipath_kernel.h"
-#include "ipath_layer.h"
 #include "ipath_common.h"
 
 /**
@@ -227,7 +226,6 @@ static ssize_t store_mlid(struct device 
 	unit = dd->ipath_unit;
 
 	dd->ipath_mlid = mlid;
-	ipath_layer_intr(dd, IPATH_LAYER_INT_BCAST);
 
 	goto bail;
 invalid:
@@ -467,7 +465,7 @@ static ssize_t store_link_state(struct d
 	if (ret < 0)
 		goto invalid;
 
-	r = ipath_layer_set_linkstate(dd, state);
+	r = ipath_set_linkstate(dd, state);
 	if (r < 0) {
 		ret = r;
 		goto bail;
@@ -502,7 +500,7 @@ static ssize_t store_mtu(struct device *
 	if (ret < 0)
 		goto invalid;
 
-	r = ipath_layer_set_mtu(dd, mtu);
+	r = ipath_set_mtu(dd, mtu);
 	if (r < 0)
 		ret = r;
 
diff --git a/drivers/infiniband/hw/ipath/ipath_uc.c b/drivers/infiniband/hw/ipath/ipath_uc.c
--- a/drivers/infiniband/hw/ipath/ipath_uc.c	Fri Aug 25 11:19:45 2006 -0700
+++ b/drivers/infiniband/hw/ipath/ipath_uc.c	Fri Aug 25 11:19:45 2006 -0700
@@ -32,7 +32,7 @@
  */
 
 #include "ipath_verbs.h"
-#include "ipath_common.h"
+#include "ipath_kernel.h"
 
 /* cut down ridiculously long IB macro names */
 #define OP(x) IB_OPCODE_UC_##x
@@ -261,8 +261,7 @@ void ipath_uc_rcv(struct ipath_ibdev *de
 		 * size to 56 bytes so the last 4 bytes of
 		 * the BTH header (PSN) is in the data buffer.
 		 */
-		header_in_data =
-			ipath_layer_get_rcvhdrentsize(dev->dd) == 16;
+		header_in_data = dev->dd->ipath_rcvhdrentsize == 16;
 		if (header_in_data) {
 			psn = be32_to_cpu(((__be32 *) data)[0]);
 			data += sizeof(__be32);
diff --git a/drivers/infiniband/hw/ipath/ipath_ud.c b/drivers/infiniband/hw/ipath/ipath_ud.c
--- a/drivers/infiniband/hw/ipath/ipath_ud.c	Fri Aug 25 11:19:45 2006 -0700
+++ b/drivers/infiniband/hw/ipath/ipath_ud.c	Fri Aug 25 11:19:45 2006 -0700
@@ -353,7 +353,7 @@ int ipath_post_ud_send(struct ipath_qp *
 		ss.num_sge++;
 	}
 	/* Check for invalid packet size. */
-	if (len > ipath_layer_get_ibmtu(dev->dd)) {
+	if (len > dev->dd->ipath_ibmtu) {
 		ret = -EINVAL;
 		goto bail;
 	}
@@ -375,7 +375,7 @@ int ipath_post_ud_send(struct ipath_qp *
 		dev->n_unicast_xmit++;
 		lid = ah_attr->dlid &
 			~((1 << (dev->mkeyprot_resv_lmc & 7)) - 1);
-		if (unlikely(lid == ipath_layer_get_lid(dev->dd))) {
+		if (unlikely(lid == dev->dd->ipath_lid)) {
 			/*
 			 * Pass in an uninitialized ib_wc to save stack
 			 * space.
@@ -404,7 +404,7 @@ int ipath_post_ud_send(struct ipath_qp *
 		qp->s_hdr.u.l.grh.sgid.global.subnet_prefix =
 			dev->gid_prefix;
 		qp->s_hdr.u.l.grh.sgid.global.interface_id =
-			ipath_layer_get_guid(dev->dd);
+			dev->dd->ipath_guid;
 		qp->s_hdr.u.l.grh.dgid = ah_attr->grh.dgid;
 		/*
 		 * Don't worry about sending to locally attached multicast
@@ -434,7 +434,7 @@ int ipath_post_ud_send(struct ipath_qp *
 	qp->s_hdr.lrh[0] = cpu_to_be16(lrh0);
 	qp->s_hdr.lrh[1] = cpu_to_be16(ah_attr->dlid);	/* DEST LID */
 	qp->s_hdr.lrh[2] = cpu_to_be16(hwords + nwords + SIZE_OF_CRC);
-	lid = ipath_layer_get_lid(dev->dd);
+	lid = dev->dd->ipath_lid;
 	if (lid) {
 		lid |= ah_attr->src_path_bits &
 			((1 << (dev->mkeyprot_resv_lmc & 7)) - 1);
@@ -445,7 +445,7 @@ int ipath_post_ud_send(struct ipath_qp *
 		bth0 |= 1 << 23;
 	bth0 |= extra_bytes << 20;
 	bth0 |= qp->ibqp.qp_type == IB_QPT_SMI ? IPATH_DEFAULT_P_KEY :
-		ipath_layer_get_pkey(dev->dd, qp->s_pkey_index);
+		ipath_get_pkey(dev->dd, qp->s_pkey_index);
 	ohdr->bth[0] = cpu_to_be32(bth0);
 	/*
 	 * Use the multicast QP if the destination LID is a multicast LID.
@@ -531,8 +531,7 @@ void ipath_ud_rcv(struct ipath_ibdev *de
 		 * the eager header buffer size to 56 bytes so the last 12
 		 * bytes of the IB header is in the data buffer.
 		 */
-		header_in_data =
-			ipath_layer_get_rcvhdrentsize(dev->dd) == 16;
+		header_in_data = dev->dd->ipath_rcvhdrentsize == 16;
 		if (header_in_data) {
 			qkey = be32_to_cpu(((__be32 *) data)[1]);
 			src_qp = be32_to_cpu(((__be32 *) data)[2]);
diff --git a/drivers/infiniband/hw/ipath/ipath_verbs.c b/drivers/infiniband/hw/ipath/ipath_verbs.c
--- a/drivers/infiniband/hw/ipath/ipath_verbs.c	Fri Aug 25 11:19:45 2006 -0700
+++ b/drivers/infiniband/hw/ipath/ipath_verbs.c	Fri Aug 25 11:19:45 2006 -0700
@@ -33,14 +33,12 @@
 
 #include <rdma/ib_mad.h>
 #include <rdma/ib_user_verbs.h>
+#include <linux/io.h>
 #include <linux/utsname.h>
 
 #include "ipath_kernel.h"
 #include "ipath_verbs.h"
 #include "ipath_common.h"
-
-/* Not static, because we don't want the compiler removing it */
-const char ipath_verbs_version[] = "ipath_verbs " IPATH_IDSTR;
 
 static unsigned int ib_ipath_qp_table_size = 251;
 module_param_named(qp_table_size, ib_ipath_qp_table_size, uint, S_IRUGO);
@@ -108,10 +106,6 @@ module_param_named(max_srq_wrs, ib_ipath
 module_param_named(max_srq_wrs, ib_ipath_max_srq_wrs,
 		   uint, S_IWUSR | S_IRUGO);
 MODULE_PARM_DESC(max_srq_wrs, "Maximum number of SRQ WRs support");
-
-MODULE_LICENSE("GPL");
-MODULE_AUTHOR("QLogic <support@pathscale.com>");
-MODULE_DESCRIPTION("QLogic InfiniPath driver");
 
 const int ib_ipath_state_ops[IB_QPS_ERR + 1] = {
 	[IB_QPS_RESET] = 0,
@@ -124,6 +118,16 @@ const int ib_ipath_state_ops[IB_QPS_ERR 
 	[IB_QPS_SQE] = IPATH_POST_RECV_OK | IPATH_PROCESS_RECV_OK,
 	[IB_QPS_ERR] = 0,
 };
+
+struct ipath_ucontext {
+	struct ib_ucontext ibucontext;
+};
+
+static inline struct ipath_ucontext *to_iucontext(struct ib_ucontext
+						  *ibucontext)
+{
+	return container_of(ibucontext, struct ipath_ucontext, ibucontext);
+}
 
 /*
  * Translate ib_wr_opcode into ib_wc_opcode.
@@ -400,7 +404,7 @@ void ipath_ib_rcv(struct ipath_ibdev *de
 	lid = be16_to_cpu(hdr->lrh[1]);
 	if (lid < IPATH_MULTICAST_LID_BASE) {
 		lid &= ~((1 << (dev->mkeyprot_resv_lmc & 7)) - 1);
-		if (unlikely(lid != ipath_layer_get_lid(dev->dd))) {
+		if (unlikely(lid != dev->dd->ipath_lid)) {
 			dev->rcv_errors++;
 			goto bail;
 		}
@@ -511,19 +515,19 @@ void ipath_ib_timer(struct ipath_ibdev *
 	if (dev->pma_sample_status == IB_PMA_SAMPLE_STATUS_STARTED &&
 	    --dev->pma_sample_start == 0) {
 		dev->pma_sample_status = IB_PMA_SAMPLE_STATUS_RUNNING;
-		ipath_layer_snapshot_counters(dev->dd, &dev->ipath_sword,
-					      &dev->ipath_rword,
-					      &dev->ipath_spkts,
-					      &dev->ipath_rpkts,
-					      &dev->ipath_xmit_wait);
+		ipath_snapshot_counters(dev->dd, &dev->ipath_sword,
+					&dev->ipath_rword,
+					&dev->ipath_spkts,
+					&dev->ipath_rpkts,
+					&dev->ipath_xmit_wait);
 	}
 	if (dev->pma_sample_status == IB_PMA_SAMPLE_STATUS_RUNNING) {
 		if (dev->pma_sample_interval == 0) {
 			u64 ta, tb, tc, td, te;
 
 			dev->pma_sample_status = IB_PMA_SAMPLE_STATUS_DONE;
-			ipath_layer_snapshot_counters(dev->dd, &ta, &tb,
-						      &tc, &td, &te);
+			ipath_snapshot_counters(dev->dd, &ta, &tb,
+						&tc, &td, &te);
 
 			dev->ipath_sword = ta - dev->ipath_sword;
 			dev->ipath_rword = tb - dev->ipath_rword;
@@ -551,6 +555,362 @@ void ipath_ib_timer(struct ipath_ibdev *
 		if (atomic_dec_and_test(&qp->refcount))
 			wake_up(&qp->wait);
 	}
+}
+
+static void update_sge(struct ipath_sge_state *ss, u32 length)
+{
+	struct ipath_sge *sge = &ss->sge;
+
+	sge->vaddr += length;
+	sge->length -= length;
+	sge->sge_length -= length;
+	if (sge->sge_length == 0) {
+		if (--ss->num_sge)
+			*sge = *ss->sg_list++;
+	} else if (sge->length == 0 && sge->mr != NULL) {
+		if (++sge->n >= IPATH_SEGSZ) {
+			if (++sge->m >= sge->mr->mapsz)
+				return;
+			sge->n = 0;
+		}
+		sge->vaddr = sge->mr->map[sge->m]->segs[sge->n].vaddr;
+		sge->length = sge->mr->map[sge->m]->segs[sge->n].length;
+	}
+}
+
+#ifdef __LITTLE_ENDIAN
+static inline u32 get_upper_bits(u32 data, u32 shift)
+{
+	return data >> shift;
+}
+
+static inline u32 set_upper_bits(u32 data, u32 shift)
+{
+	return data << shift;
+}
+
+static inline u32 clear_upper_bytes(u32 data, u32 n, u32 off)
+{
+	data <<= ((sizeof(u32) - n) * BITS_PER_BYTE);
+	data >>= ((sizeof(u32) - n - off) * BITS_PER_BYTE);
+	return data;
+}
+#else
+static inline u32 get_upper_bits(u32 data, u32 shift)
+{
+	return data << shift;
+}
+
+static inline u32 set_upper_bits(u32 data, u32 shift)
+{
+	return data >> shift;
+}
+
+static inline u32 clear_upper_bytes(u32 data, u32 n, u32 off)
+{
+	data >>= ((sizeof(u32) - n) * BITS_PER_BYTE);
+	data <<= ((sizeof(u32) - n - off) * BITS_PER_BYTE);
+	return data;
+}
+#endif
+
+static void copy_io(u32 __iomem *piobuf, struct ipath_sge_state *ss,
+		    u32 length)
+{
+	u32 extra = 0;
+	u32 data = 0;
+	u32 last;
+
+	while (1) {
+		u32 len = ss->sge.length;
+		u32 off;
+
+		BUG_ON(len == 0);
+		if (len > length)
+			len = length;
+		if (len > ss->sge.sge_length)
+			len = ss->sge.sge_length;
+		/* If the source address is not aligned, try to align it. */
+		off = (unsigned long)ss->sge.vaddr & (sizeof(u32) - 1);
+		if (off) {
+			u32 *addr = (u32 *)((unsigned long)ss->sge.vaddr &
+					    ~(sizeof(u32) - 1));
+			u32 v = get_upper_bits(*addr, off * BITS_PER_BYTE);
+			u32 y;
+
+			y = sizeof(u32) - off;
+			if (len > y)
+				len = y;
+			if (len + extra >= sizeof(u32)) {
+				data |= set_upper_bits(v, extra *
+						       BITS_PER_BYTE);
+				len = sizeof(u32) - extra;
+				if (len == length) {
+					last = data;
+					break;
+				}
+				__raw_writel(data, piobuf);
+				piobuf++;
+				extra = 0;
+				data = 0;
+			} else {
+				/* Clear unused upper bytes */
+				data |= clear_upper_bytes(v, len, extra);
+				if (len == length) {
+					last = data;
+					break;
+				}
+				extra += len;
+			}
+		} else if (extra) {
+			/* Source address is aligned. */
+			u32 *addr = (u32 *) ss->sge.vaddr;
+			int shift = extra * BITS_PER_BYTE;
+			int ushift = 32 - shift;
+			u32 l = len;
+
+			while (l >= sizeof(u32)) {
+				u32 v = *addr;
+
+				data |= set_upper_bits(v, shift);
+				__raw_writel(data, piobuf);
+				data = get_upper_bits(v, ushift);
+				piobuf++;
+				addr++;
+				l -= sizeof(u32);
+			}
+			/*
+			 * We still have 'extra' number of bytes leftover.
+			 */
+			if (l) {
+				u32 v = *addr;
+
+				if (l + extra >= sizeof(u32)) {
+					data |= set_upper_bits(v, shift);
+					len -= l + extra - sizeof(u32);
+					if (len == length) {
+						last = data;
+						break;
+					}
+					__raw_writel(data, piobuf);
+					piobuf++;
+					extra = 0;
+					data = 0;
+				} else {
+					/* Clear unused upper bytes */
+					data |= clear_upper_bytes(v, l,
+								  extra);
+					if (len == length) {
+						last = data;
+						break;
+					}
+					extra += l;
+				}
+			} else if (len == length) {
+				last = data;
+				break;
+			}
+		} else if (len == length) {
+			u32 w;
+
+			/*
+			 * Need to round up for the last dword in the
+			 * packet.
+			 */
+			w = (len + 3) >> 2;
+			__iowrite32_copy(piobuf, ss->sge.vaddr, w - 1);
+			piobuf += w - 1;
+			last = ((u32 *) ss->sge.vaddr)[w - 1];
+			break;
+		} else {
+			u32 w = len >> 2;
+
+			__iowrite32_copy(piobuf, ss->sge.vaddr, w);
+			piobuf += w;
+
+			extra = len & (sizeof(u32) - 1);
+			if (extra) {
+				u32 v = ((u32 *) ss->sge.vaddr)[w];
+
+				/* Clear unused upper bytes */
+				data = clear_upper_bytes(v, extra, 0);
+			}
+		}
+		update_sge(ss, len);
+		length -= len;
+	}
+	/* Update address before sending packet. */
+	update_sge(ss, length);
+	/* must flush early everything before trigger word */
+	ipath_flush_wc();
+	__raw_writel(last, piobuf);
+	/* be sure trigger word is written */
+	ipath_flush_wc();
+}
+
+/**
+ * ipath_verbs_send - send a packet
+ * @dd: the infinipath device
+ * @hdrwords: the number of words in the header
+ * @hdr: the packet header
+ * @len: the length of the packet in bytes
+ * @ss: the SGE to send
+ */
+int ipath_verbs_send(struct ipath_devdata *dd, u32 hdrwords,
+		     u32 *hdr, u32 len, struct ipath_sge_state *ss)
+{
+	u32 __iomem *piobuf;
+	u32 plen;
+	int ret;
+
+	/* +1 is for the qword padding of pbc */
+	plen = hdrwords + ((len + 3) >> 2) + 1;
+	if (unlikely((plen << 2) > dd->ipath_ibmaxlen)) {
+		ipath_dbg("packet len 0x%x too long, failing\n", plen);
+		ret = -EINVAL;
+		goto bail;
+	}
+
+	/* Get a PIO buffer to use. */
+	piobuf = ipath_getpiobuf(dd, NULL);
+	if (unlikely(piobuf == NULL)) {
+		ret = -EBUSY;
+		goto bail;
+	}
+
+	/*
+	 * Write len to control qword, no flags.
+	 * We have to flush after the PBC for correctness on some cpus
+	 * or WC buffer can be written out of order.
+	 */
+	writeq(plen, piobuf);
+	ipath_flush_wc();
+	piobuf += 2;
+	if (len == 0) {
+		/*
+		 * If there is just the header portion, must flush before
+		 * writing last word of header for correctness, and after
+		 * the last header word (trigger word).
+		 */
+		__iowrite32_copy(piobuf, hdr, hdrwords - 1);
+		ipath_flush_wc();
+		__raw_writel(hdr[hdrwords - 1], piobuf + hdrwords - 1);
+		ipath_flush_wc();
+		ret = 0;
+		goto bail;
+	}
+
+	__iowrite32_copy(piobuf, hdr, hdrwords);
+	piobuf += hdrwords;
+
+	/* The common case is aligned and contained in one segment. */
+	if (likely(ss->num_sge == 1 && len <= ss->sge.length &&
+		   !((unsigned long)ss->sge.vaddr & (sizeof(u32) - 1)))) {
+		u32 w;
+		u32 *addr = (u32 *) ss->sge.vaddr;
+
+		/* Update address before sending packet. */
+		update_sge(ss, len);
+		/* Need to round up for the last dword in the packet. */
+		w = (len + 3) >> 2;
+		__iowrite32_copy(piobuf, addr, w - 1);
+		/* must flush early everything before trigger word */
+		ipath_flush_wc();
+		__raw_writel(addr[w - 1], piobuf + w - 1);
+		/* be sure trigger word is written */
+		ipath_flush_wc();
+		ret = 0;
+		goto bail;
+	}
+	copy_io(piobuf, ss, len);
+	ret = 0;
+
+bail:
+	return ret;
+}
+
+int ipath_snapshot_counters(struct ipath_devdata *dd, u64 *swords,
+			    u64 *rwords, u64 *spkts, u64 *rpkts,
+			    u64 *xmit_wait)
+{
+	int ret;
+
+	if (!(dd->ipath_flags & IPATH_INITTED)) {
+		/* no hardware, freeze, etc. */
+		ipath_dbg("unit %u not usable\n", dd->ipath_unit);
+		ret = -EINVAL;
+		goto bail;
+	}
+	*swords = ipath_snap_cntr(dd, dd->ipath_cregs->cr_wordsendcnt);
+	*rwords = ipath_snap_cntr(dd, dd->ipath_cregs->cr_wordrcvcnt);
+	*spkts = ipath_snap_cntr(dd, dd->ipath_cregs->cr_pktsendcnt);
+	*rpkts = ipath_snap_cntr(dd, dd->ipath_cregs->cr_pktrcvcnt);
+	*xmit_wait = ipath_snap_cntr(dd, dd->ipath_cregs->cr_sendstallcnt);
+
+	ret = 0;
+
+bail:
+	return ret;
+}
+
+/**
+ * ipath_get_counters - get various chip counters
+ * @dd: the infinipath device
+ * @cntrs: counters are placed here
+ *
+ * Return the counters needed by recv_pma_get_portcounters().
+ */
+int ipath_get_counters(struct ipath_devdata *dd,
+		       struct ipath_verbs_counters *cntrs)
+{
+	int ret;
+
+	if (!(dd->ipath_flags & IPATH_INITTED)) {
+		/* no hardware, freeze, etc. */
+		ipath_dbg("unit %u not usable\n", dd->ipath_unit);
+		ret = -EINVAL;
+		goto bail;
+	}
+	cntrs->symbol_error_counter =
+		ipath_snap_cntr(dd, dd->ipath_cregs->cr_ibsymbolerrcnt);
+	cntrs->link_error_recovery_counter =
+		ipath_snap_cntr(dd, dd->ipath_cregs->cr_iblinkerrrecovcnt);
+	/*
+	 * The link downed counter counts when the other side downs the
+	 * connection.  We add in the number of times we downed the link
+	 * due to local link integrity errors to compensate.
+	 */
+	cntrs->link_downed_counter =
+		ipath_snap_cntr(dd, dd->ipath_cregs->cr_iblinkdowncnt);
+	cntrs->port_rcv_errors =
+		ipath_snap_cntr(dd, dd->ipath_cregs->cr_rxdroppktcnt) +
+		ipath_snap_cntr(dd, dd->ipath_cregs->cr_rcvovflcnt) +
+		ipath_snap_cntr(dd, dd->ipath_cregs->cr_portovflcnt) +
+		ipath_snap_cntr(dd, dd->ipath_cregs->cr_err_rlencnt) +
+		ipath_snap_cntr(dd, dd->ipath_cregs->cr_invalidrlencnt) +
+		ipath_snap_cntr(dd, dd->ipath_cregs->cr_erricrccnt) +
+		ipath_snap_cntr(dd, dd->ipath_cregs->cr_errvcrccnt) +
+		ipath_snap_cntr(dd, dd->ipath_cregs->cr_errlpcrccnt) +
+		ipath_snap_cntr(dd, dd->ipath_cregs->cr_badformatcnt);
+	cntrs->port_rcv_remphys_errors =
+		ipath_snap_cntr(dd, dd->ipath_cregs->cr_rcvebpcnt);
+	cntrs->port_xmit_discards =
+		ipath_snap_cntr(dd, dd->ipath_cregs->cr_unsupvlcnt);
+	cntrs->port_xmit_data =
+		ipath_snap_cntr(dd, dd->ipath_cregs->cr_wordsendcnt);
+	cntrs->port_rcv_data =
+		ipath_snap_cntr(dd, dd->ipath_cregs->cr_wordrcvcnt);
+	cntrs->port_xmit_packets =
+		ipath_snap_cntr(dd, dd->ipath_cregs->cr_pktsendcnt);
+	cntrs->port_rcv_packets =
+		ipath_snap_cntr(dd, dd->ipath_cregs->cr_pktrcvcnt);
+	cntrs->local_link_integrity_errors = dd->ipath_lli_errors;
+	cntrs->excessive_buffer_overrun_errors = 0; /* XXX */
+
+	ret = 0;
+
+bail:
+	return ret;
 }
 
 /**
@@ -595,9 +955,9 @@ static int ipath_query_device(struct ib_
 		IB_DEVICE_BAD_QKEY_CNTR | IB_DEVICE_SHUTDOWN_PORT |
 		IB_DEVICE_SYS_IMAGE_GUID;
 	props->page_size_cap = PAGE_SIZE;
-	props->vendor_id = ipath_layer_get_vendorid(dev->dd);
-	props->vendor_part_id = ipath_layer_get_deviceid(dev->dd);
-	props->hw_ver = ipath_layer_get_pcirev(dev->dd);
+	props->vendor_id = dev->dd->ipath_vendorid;
+	props->vendor_part_id = dev->dd->ipath_deviceid;
+	props->hw_ver = dev->dd->ipath_pcirev;
 
 	props->sys_image_guid = dev->sys_image_guid;
 
@@ -618,7 +978,7 @@ static int ipath_query_device(struct ib_
 	props->max_srq_sge = ib_ipath_max_srq_sges;
 	/* props->local_ca_ack_delay */
 	props->atomic_cap = IB_ATOMIC_HCA;
-	props->max_pkeys = ipath_layer_get_npkeys(dev->dd);
+	props->max_pkeys = ipath_get_npkeys(dev->dd);
 	props->max_mcast_grp = ib_ipath_max_mcast_grps;
 	props->max_mcast_qp_attach = ib_ipath_max_mcast_qp_attached;
 	props->max_total_mcast_qp_attach = props->max_mcast_qp_attach *
@@ -643,12 +1003,17 @@ const u8 ipath_cvt_physportstate[16] = {
 	[INFINIPATH_IBCS_LT_STATE_RECOVERIDLE] = 6,
 };
 
+u32 ipath_get_cr_errpkey(struct ipath_devdata *dd)
+{
+	return ipath_read_creg32(dd, dd->ipath_cregs->cr_errpkey);
+}
+
 static int ipath_query_port(struct ib_device *ibdev,
 			    u8 port, struct ib_port_attr *props)
 {
 	struct ipath_ibdev *dev = to_idev(ibdev);
 	enum ib_mtu mtu;
-	u16 lid = ipath_layer_get_lid(dev->dd);
+	u16 lid = dev->dd->ipath_lid;
 	u64 ibcstat;
 
 	memset(props, 0, sizeof(*props));
@@ -656,16 +1021,16 @@ static int ipath_query_port(struct ib_de
 	props->lmc = dev->mkeyprot_resv_lmc & 7;
 	props->sm_lid = dev->sm_lid;
 	props->sm_sl = dev->sm_sl;
-	ibcstat = ipath_layer_get_lastibcstat(dev->dd);
+	ibcstat = dev->dd->ipath_lastibcstat;
 	props->state = ((ibcstat >> 4) & 0x3) + 1;
 	/* See phys_state_show() */
 	props->phys_state = ipath_cvt_physportstate[
-		ipath_layer_get_lastibcstat(dev->dd) & 0xf];
+		dev->dd->ipath_lastibcstat & 0xf];
 	props->port_cap_flags = dev->port_cap_flags;
 	props->gid_tbl_len = 1;
 	props->max_msg_sz = 0x80000000;
-	props->pkey_tbl_len = ipath_layer_get_npkeys(dev->dd);
-	props->bad_pkey_cntr = ipath_layer_get_cr_errpkey(dev->dd) -
+	props->pkey_tbl_len = ipath_get_npkeys(dev->dd);
+	props->bad_pkey_cntr = ipath_get_cr_errpkey(dev->dd) -
 		dev->z_pkey_violations;
 	props->qkey_viol_cntr = dev->qkey_violations;
 	props->active_width = IB_WIDTH_4X;
@@ -675,7 +1040,7 @@ static int ipath_query_port(struct ib_de
 	props->init_type_reply = 0;
 
 	props->max_mtu = IB_MTU_4096;
-	switch (ipath_layer_get_ibmtu(dev->dd)) {
+	switch (dev->dd->ipath_ibmtu) {
 	case 4096:
 		mtu = IB_MTU_4096;
 		break;
@@ -734,7 +1099,7 @@ static int ipath_modify_port(struct ib_d
 	dev->port_cap_flags |= props->set_port_cap_mask;
 	dev->port_cap_flags &= ~props->clr_port_cap_mask;
 	if (port_modify_mask & IB_PORT_SHUTDOWN)
-		ipath_layer_set_linkstate(dev->dd, IPATH_IB_LINKDOWN);
+		ipath_set_linkstate(dev->dd, IPATH_IB_LINKDOWN);
 	if (port_modify_mask & IB_PORT_RESET_QKEY_CNTR)
 		dev->qkey_violations = 0;
 	return 0;
@@ -751,7 +1116,7 @@ static int ipath_query_gid(struct ib_dev
 		goto bail;
 	}
 	gid->global.subnet_prefix = dev->gid_prefix;
-	gid->global.interface_id = ipath_layer_get_guid(dev->dd);
+	gid->global.interface_id = dev->dd->ipath_guid;
 
 	ret = 0;
 
@@ -902,24 +1267,49 @@ static int ipath_query_ah(struct ib_ah *
 	return 0;
 }
 
+/**
+ * ipath_get_npkeys - return the size of the PKEY table for port 0
+ * @dd: the infinipath device
+ */
+unsigned ipath_get_npkeys(struct ipath_devdata *dd)
+{
+	return ARRAY_SIZE(dd->ipath_pd[0]->port_pkeys);
+}
+
+/**
+ * ipath_get_pkey - return the indexed PKEY from the port 0 PKEY table
+ * @dd: the infinipath device
+ * @index: the PKEY index
+ */
+unsigned ipath_get_pkey(struct ipath_devdata *dd, unsigned index)
+{
+	unsigned ret;
+
+	if (index >= ARRAY_SIZE(dd->ipath_pd[0]->port_pkeys))
+		ret = 0;
+	else
+		ret = dd->ipath_pd[0]->port_pkeys[index];
+
+	return ret;
+}
+
 static int ipath_query_pkey(struct ib_device *ibdev, u8 port, u16 index,
 			    u16 *pkey)
 {
 	struct ipath_ibdev *dev = to_idev(ibdev);
 	int ret;
 
-	if (index >= ipath_layer_get_npkeys(dev->dd)) {
+	if (index >= ipath_get_npkeys(dev->dd)) {
 		ret = -EINVAL;
 		goto bail;
 	}
 
-	*pkey = ipath_layer_get_pkey(dev->dd, index);
+	*pkey = ipath_get_pkey(dev->dd, index);
 	ret = 0;
 
 bail:
 	return ret;
 }
-
 
 /**
  * ipath_alloc_ucontext - allocate a ucontest
@@ -953,6 +1343,63 @@ static int ipath_dealloc_ucontext(struct
 
 static int ipath_verbs_register_sysfs(struct ib_device *dev);
 
+static void __verbs_timer(unsigned long arg)
+{
+	struct ipath_devdata *dd = (struct ipath_devdata *) arg;
+
+	/*
+	 * If port 0 receive packet interrupts are not available, or
+	 * can be missed, poll the receive queue
+	 */
+	if (dd->ipath_flags & IPATH_POLL_RX_INTR)
+		ipath_kreceive(dd);
+
+	/* Handle verbs layer timeouts. */
+	ipath_ib_timer(dd->verbs_dev);
+
+	mod_timer(&dd->verbs_timer, jiffies + 1);
+}
+
+static int enable_timer(struct ipath_devdata *dd)
+{
+	/*
+	 * Early chips had a design flaw where the chip and kernel idea
+	 * of the tail register don't always agree, and therefore we won't
+	 * get an interrupt on the next packet received.
+	 * If the board supports per packet receive interrupts, use it.
+	 * Otherwise, the timer function periodically checks for packets
+	 * to cover this case.
+	 * Either way, the timer is needed for verbs layer related
+	 * processing.
+	 */
+	if (dd->ipath_flags & IPATH_GPIO_INTR) {
+		ipath_write_kreg(dd, dd->ipath_kregs->kr_debugportselect,
+				 0x2074076542310ULL);
+		/* Enable GPIO bit 2 interrupt */
+		ipath_write_kreg(dd, dd->ipath_kregs->kr_gpio_mask,
+				 (u64) (1 << 2));
+	}
+
+	init_timer(&dd->verbs_timer);
+	dd->verbs_timer.function = __verbs_timer;
+	dd->verbs_timer.data = (unsigned long)dd;
+	dd->verbs_timer.expires = jiffies + 1;
+	add_timer(&dd->verbs_timer);
+
+	return 0;
+}
+
+static int disable_timer(struct ipath_devdata *dd)
+{
+	/* Disable GPIO bit 2 interrupt */
+	if (dd->ipath_flags & IPATH_GPIO_INTR)
+		ipath_write_kreg(dd, dd->ipath_kregs->kr_gpio_mask, 0);
+
+	del_timer_sync(&dd->verbs_timer);
+
+	return 0;
+}
+
 /**
  * ipath_register_ib_device - register our device with the infiniband core
  * @dd: the device data structure
@@ -960,7 +1407,7 @@ static int ipath_verbs_register_sysfs(st
  */
 int ipath_register_ib_device(struct ipath_devdata *dd)
 {
-	struct ipath_layer_counters cntrs;
+	struct ipath_verbs_counters cntrs;
 	struct ipath_ibdev *idev;
 	struct ib_device *dev;
 	int ret;
@@ -1020,7 +1467,7 @@ int ipath_register_ib_device(struct ipat
 	idev->link_width_enabled = 3;	/* 1x or 4x */
 
 	/* Snapshot current HW counters to "clear" them. */
-	ipath_layer_get_counters(dd, &cntrs);
+	ipath_get_counters(dd, &cntrs);
 	idev->z_symbol_error_counter = cntrs.symbol_error_counter;
 	idev->z_link_error_recovery_counter =
 		cntrs.link_error_recovery_counter;
@@ -1044,14 +1491,14 @@ int ipath_register_ib_device(struct ipat
 	 * device types in the system, we can't be sure this is unique.
 	 */
 	if (!sys_image_guid)
-		sys_image_guid = ipath_layer_get_guid(dd);
+		sys_image_guid = dd->ipath_guid;
 	idev->sys_image_guid = sys_image_guid;
 	idev->ib_unit = dd->ipath_unit;
 	idev->dd = dd;
 
 	strlcpy(dev->name, "ipath%d", IB_DEVICE_NAME_MAX);
 	dev->owner = THIS_MODULE;
-	dev->node_guid = ipath_layer_get_guid(dd);
+	dev->node_guid = dd->ipath_guid;
 	dev->uverbs_abi_ver = IPATH_UVERBS_ABI_VERSION;
 	dev->uverbs_cmd_mask =
 		(1ull << IB_USER_VERBS_CMD_GET_CONTEXT)		|
@@ -1085,7 +1532,7 @@ int ipath_register_ib_device(struct ipat
 		(1ull << IB_USER_VERBS_CMD_POST_SRQ_RECV);
 	dev->node_type = IB_NODE_CA;
 	dev->phys_port_cnt = 1;
-	dev->dma_device = ipath_layer_get_device(dd);
+	dev->dma_device = &dd->pcidev->dev;
 	dev->class_dev.dev = dev->dma_device;
 	dev->query_device = ipath_query_device;
 	dev->modify_device = ipath_modify_device;
@@ -1139,7 +1586,7 @@ int ipath_register_ib_device(struct ipat
 	if (ipath_verbs_register_sysfs(dev))
 		goto err_class;
 
-	ipath_layer_enable_timer(dd);
+	enable_timer(dd);
 
 	goto bail;
 
@@ -1164,7 +1611,7 @@ void ipath_unregister_ib_device(struct i
 {
 	struct ib_device *ibdev = &dev->ibdev;
 
-	ipath_layer_disable_timer(dev->dd);
+	disable_timer(dev->dd);
 
 	ib_unregister_device(ibdev);
 
@@ -1197,7 +1644,7 @@ static ssize_t show_rev(struct class_dev
 	struct ipath_ibdev *dev =
 		container_of(cdev, struct ipath_ibdev, ibdev.class_dev);
 
-	return sprintf(buf, "%x\n", ipath_layer_get_pcirev(dev->dd));
+	return sprintf(buf, "%x\n", dev->dd->ipath_pcirev);
 }
 
 static ssize_t show_hca(struct class_device *cdev, char *buf)
@@ -1206,7 +1653,7 @@ static ssize_t show_hca(struct class_dev
 		container_of(cdev, struct ipath_ibdev, ibdev.class_dev);
 	int ret;
 
-	ret = ipath_layer_get_boardname(dev->dd, buf, 128);
+	ret = dev->dd->ipath_f_get_boardname(dev->dd, buf, 128);
 	if (ret < 0)
 		goto bail;
 	strcat(buf, "\n");
diff --git a/drivers/infiniband/hw/ipath/ipath_verbs.h b/drivers/infiniband/hw/ipath/ipath_verbs.h
--- a/drivers/infiniband/hw/ipath/ipath_verbs.h	Fri Aug 25 11:19:45 2006 -0700
+++ b/drivers/infiniband/hw/ipath/ipath_verbs.h	Fri Aug 25 11:19:45 2006 -0700
@@ -153,19 +153,6 @@ struct ipath_mcast {
 	int n_attached;
 };
 
-/* Memory region */
-struct ipath_mr {
-	struct ib_mr ibmr;
-	struct ipath_mregion mr;	/* must be last */
-};
-
-/* Fast memory region */
-struct ipath_fmr {
-	struct ib_fmr ibfmr;
-	u8 page_shift;
-	struct ipath_mregion mr;	/* must be last */
-};
-
 /* Protection domain */
 struct ipath_pd {
 	struct ib_pd ibpd;
@@ -217,6 +204,54 @@ struct ipath_cq {
 };
 
 /*
+ * A segment is a linear region of low physical memory.
+ * XXX Maybe we should use phys addr here and kmap()/kunmap().
+ * Used by the verbs layer.
+ */
+struct ipath_seg {
+	void *vaddr;
+	size_t length;
+};
+
+/* The number of ipath_segs that fit in a page. */
+#define IPATH_SEGSZ     (PAGE_SIZE / sizeof (struct ipath_seg))
+
+struct ipath_segarray {
+	struct ipath_seg segs[IPATH_SEGSZ];
+};
+
+struct ipath_mregion {
+	u64 user_base;		/* User's address for this region */
+	u64 iova;		/* IB start address of this region */
+	size_t length;
+	u32 lkey;
+	u32 offset;		/* offset (bytes) to start of region */
+	int access_flags;
+	u32 max_segs;		/* number of ipath_segs in all the arrays */
+	u32 mapsz;		/* size of the map array */
+	struct ipath_segarray *map[0];	/* the segments */
+};
+
+/*
+ * These keep track of the copy progress within a memory region.
+ * Used by the verbs layer.
+ */
+struct ipath_sge {
+	struct ipath_mregion *mr;
+	void *vaddr;		/* current pointer into the segment */
+	u32 sge_length;		/* length of the SGE */
+	u32 length;		/* remaining length of the segment */
+	u16 m;			/* current index: mr->map[m] */
+	u16 n;			/* current index: mr->map[m]->segs[n] */
+};
+
+/* Memory region */
+struct ipath_mr {
+	struct ib_mr ibmr;
+	struct ipath_mregion mr;	/* must be last */
+};
+
+/*
  * Send work request queue entry.
  * The size of the sg_list is determined when the QP is created and stored
  * in qp->s_max_sge.
@@ -268,6 +303,12 @@ struct ipath_srq {
 	struct ipath_mmap_info *ip;
 	/* send signal when number of RWQEs < limit */
 	u32 limit;
+};
+
+struct ipath_sge_state {
+	struct ipath_sge *sg_list;      /* next SGE to be used if any */
+	struct ipath_sge sge;   /* progress state for the current SGE */
+	u8 num_sge;
 };
 
 /*
@@ -500,18 +541,24 @@ struct ipath_ibdev {
 	struct ipath_opcode_stats opstats[128];
 };
 
-struct ipath_ucontext {
-	struct ib_ucontext ibucontext;
+struct ipath_verbs_counters {
+	u64 symbol_error_counter;
+	u64 link_error_recovery_counter;
+	u64 link_downed_counter;
+	u64 port_rcv_errors;
+	u64 port_rcv_remphys_errors;
+	u64 port_xmit_discards;
+	u64 port_xmit_data;
+	u64 port_rcv_data;
+	u64 port_xmit_packets;
+	u64 port_rcv_packets;
+	u32 local_link_integrity_errors;
+	u32 excessive_buffer_overrun_errors;
 };
 
 static inline struct ipath_mr *to_imr(struct ib_mr *ibmr)
 {
 	return container_of(ibmr, struct ipath_mr, ibmr);
-}
-
-static inline struct ipath_fmr *to_ifmr(struct ib_fmr *ibfmr)
-{
-	return container_of(ibfmr, struct ipath_fmr, ibfmr);
 }
 
 static inline struct ipath_pd *to_ipd(struct ib_pd *ibpd)
@@ -551,12 +598,6 @@ int ipath_process_mad(struct ib_device *
 		      struct ib_grh *in_grh,
 		      struct ib_mad *in_mad, struct ib_mad *out_mad);
 
-static inline struct ipath_ucontext *to_iucontext(struct ib_ucontext
-						  *ibucontext)
-{
-	return container_of(ibucontext, struct ipath_ucontext, ibucontext);
-}
-
 /*
  * Compare the lower 24 bits of the two values.
  * Returns an integer <, ==, or > than zero.
@@ -567,6 +608,13 @@ static inline int ipath_cmp24(u32 a, u32
 }
 
 struct ipath_mcast *ipath_mcast_find(union ib_gid *mgid);
+
+int ipath_snapshot_counters(struct ipath_devdata *dd, u64 *swords,
+			    u64 *rwords, u64 *spkts, u64 *rpkts,
+			    u64 *xmit_wait);
+
+int ipath_get_counters(struct ipath_devdata *dd,
+		       struct ipath_verbs_counters *cntrs);
 
 int ipath_multicast_attach(struct ib_qp *ibqp, union ib_gid *gid, u16 lid);
 
@@ -598,6 +646,9 @@ void ipath_sqerror_qp(struct ipath_qp *q
 
 void ipath_get_credit(struct ipath_qp *qp, u32 aeth);
 
+int ipath_verbs_send(struct ipath_devdata *dd, u32 hdrwords,
+		     u32 *hdr, u32 len, struct ipath_sge_state *ss);
+
 void ipath_cq_enter(struct ipath_cq *cq, struct ib_wc *entry, int sig);
 
 int ipath_rkey_ok(struct ipath_ibdev *dev, struct ipath_sge_state *ss,
@@ -721,6 +772,12 @@ int ipath_ib_piobufavail(struct ipath_ib
 
 void ipath_ib_timer(struct ipath_ibdev *);
 
+unsigned ipath_get_npkeys(struct ipath_devdata *);
+
+u32 ipath_get_cr_errpkey(struct ipath_devdata *);
+
+unsigned ipath_get_pkey(struct ipath_devdata *, unsigned);
+
 extern const enum ib_wc_opcode ib_ipath_wc_opcode[];
 
 extern const u8 ipath_cvt_physportstate[];

[PATCH 8 of 23] IB/ipath - simplify debugging code after ipath_core and ib_ipath merger

[Bryan O'Sullivan]

Signed-off-by: Bryan O'Sullivan <bryan.osullivan@qlogic.com>

diff --git a/drivers/infiniband/hw/ipath/ipath_driver.c b/drivers/infiniband/hw/ipath/ipath_driver.c
--- a/drivers/infiniband/hw/ipath/ipath_driver.c	Fri Aug 25 11:19:45 2006 -0700
+++ b/drivers/infiniband/hw/ipath/ipath_driver.c	Fri Aug 25 11:19:45 2006 -0700
@@ -58,7 +58,7 @@ const char *ipath_get_unit_name(int unit
  * The size has to be longer than this string, so we can append
  * board/chip information to it in the init code.
  */
-const char ipath_core_version[] = IPATH_IDSTR "\n";
+const char ib_ipath_version[] = IPATH_IDSTR "\n";
 
 static struct idr unit_table;
 DEFINE_SPINLOCK(ipath_devs_lock);
@@ -1847,7 +1847,7 @@ static int __init infinipath_init(void)
 {
 	int ret;
 
-	ipath_dbg(KERN_INFO DRIVER_LOAD_MSG "%s", ipath_core_version);
+	ipath_dbg(KERN_INFO DRIVER_LOAD_MSG "%s", ib_ipath_version);
 
 	/*
 	 * These must be called before the driver is registered with
diff --git a/drivers/infiniband/hw/ipath/ipath_kernel.h b/drivers/infiniband/hw/ipath/ipath_kernel.h
--- a/drivers/infiniband/hw/ipath/ipath_kernel.h	Fri Aug 25 11:19:45 2006 -0700
+++ b/drivers/infiniband/hw/ipath/ipath_kernel.h	Fri Aug 25 11:19:45 2006 -0700
@@ -785,7 +785,7 @@ static inline u32 ipath_read_creg32(cons
 
 struct device_driver;
 
-extern const char ipath_core_version[];
+extern const char ib_ipath_version[];
 
 int ipath_driver_create_group(struct device_driver *);
 void ipath_driver_remove_group(struct device_driver *);
@@ -815,7 +815,7 @@ const char *ipath_get_unit_name(int unit
 
 extern struct mutex ipath_mutex;
 
-#define IPATH_DRV_NAME		"ipath_core"
+#define IPATH_DRV_NAME		"ib_ipath"
 #define IPATH_MAJOR		233
 #define IPATH_USER_MINOR_BASE	0
 #define IPATH_SMA_MINOR		128
diff --git a/drivers/infiniband/hw/ipath/ipath_keys.c b/drivers/infiniband/hw/ipath/ipath_keys.c
--- a/drivers/infiniband/hw/ipath/ipath_keys.c	Fri Aug 25 11:19:45 2006 -0700
+++ b/drivers/infiniband/hw/ipath/ipath_keys.c	Fri Aug 25 11:19:45 2006 -0700
@@ -34,6 +34,7 @@
 #include <asm/io.h>
 
 #include "ipath_verbs.h"
+#include "ipath_kernel.h"
 
 /**
  * ipath_alloc_lkey - allocate an lkey
@@ -60,7 +61,7 @@ int ipath_alloc_lkey(struct ipath_lkey_t
 		r = (r + 1) & (rkt->max - 1);
 		if (r == n) {
 			spin_unlock_irqrestore(&rkt->lock, flags);
-			_VERBS_INFO("LKEY table full\n");
+			ipath_dbg(KERN_INFO "LKEY table full\n");
 			ret = 0;
 			goto bail;
 		}
diff --git a/drivers/infiniband/hw/ipath/ipath_qp.c b/drivers/infiniband/hw/ipath/ipath_qp.c
--- a/drivers/infiniband/hw/ipath/ipath_qp.c	Fri Aug 25 11:19:45 2006 -0700
+++ b/drivers/infiniband/hw/ipath/ipath_qp.c	Fri Aug 25 11:19:45 2006 -0700
@@ -274,7 +274,7 @@ void ipath_free_all_qps(struct ipath_qp_
 				free_qpn(qpt, qp->ibqp.qp_num);
 			if (!atomic_dec_and_test(&qp->refcount) ||
 			    !ipath_destroy_qp(&qp->ibqp))
-				_VERBS_INFO("QP memory leak!\n");
+				ipath_dbg(KERN_INFO "QP memory leak!\n");
 			qp = nqp;
 		}
 	}
@@ -362,8 +362,8 @@ void ipath_error_qp(struct ipath_qp *qp)
 	struct ipath_ibdev *dev = to_idev(qp->ibqp.device);
 	struct ib_wc wc;
 
-	_VERBS_INFO("QP%d/%d in error state\n",
-		    qp->ibqp.qp_num, qp->remote_qpn);
+	ipath_dbg(KERN_INFO "QP%d/%d in error state\n",
+		  qp->ibqp.qp_num, qp->remote_qpn);
 
 	spin_lock(&dev->pending_lock);
 	/* XXX What if its already removed by the timeout code? */
@@ -945,8 +945,8 @@ void ipath_sqerror_qp(struct ipath_qp *q
 	struct ipath_ibdev *dev = to_idev(qp->ibqp.device);
 	struct ipath_swqe *wqe = get_swqe_ptr(qp, qp->s_last);
 
-	_VERBS_INFO("Send queue error on QP%d/%d: err: %d\n",
-		    qp->ibqp.qp_num, qp->remote_qpn, wc->status);
+	ipath_dbg(KERN_INFO "Send queue error on QP%d/%d: err: %d\n",
+		  qp->ibqp.qp_num, qp->remote_qpn, wc->status);
 
 	spin_lock(&dev->pending_lock);
 	/* XXX What if its already removed by the timeout code? */
diff --git a/drivers/infiniband/hw/ipath/ipath_sysfs.c b/drivers/infiniband/hw/ipath/ipath_sysfs.c
--- a/drivers/infiniband/hw/ipath/ipath_sysfs.c	Fri Aug 25 11:19:45 2006 -0700
+++ b/drivers/infiniband/hw/ipath/ipath_sysfs.c	Fri Aug 25 11:19:45 2006 -0700
@@ -75,7 +75,7 @@ static ssize_t show_version(struct devic
 static ssize_t show_version(struct device_driver *dev, char *buf)
 {
 	/* The string printed here is already newline-terminated. */
-	return scnprintf(buf, PAGE_SIZE, "%s", ipath_core_version);
+	return scnprintf(buf, PAGE_SIZE, "%s", ib_ipath_version);
 }
 
 static ssize_t show_num_units(struct device_driver *dev, char *buf)
diff --git a/drivers/infiniband/hw/ipath/ipath_verbs.c b/drivers/infiniband/hw/ipath/ipath_verbs.c
--- a/drivers/infiniband/hw/ipath/ipath_verbs.c	Fri Aug 25 11:19:45 2006 -0700
+++ b/drivers/infiniband/hw/ipath/ipath_verbs.c	Fri Aug 25 11:19:45 2006 -0700
@@ -49,10 +49,6 @@ module_param_named(lkey_table_size, ib_i
 		   S_IRUGO);
 MODULE_PARM_DESC(lkey_table_size,
 		 "LKEY table size in bits (2^n, 1 <= n <= 23)");
-
-unsigned int ib_ipath_debug;	/* debug mask */
-module_param_named(debug, ib_ipath_debug, uint, S_IWUSR | S_IRUGO);
-MODULE_PARM_DESC(debug, "Verbs debug mask");
 
 static unsigned int ib_ipath_max_pds = 0xFFFF;
 module_param_named(max_pds, ib_ipath_max_pds, uint, S_IWUSR | S_IRUGO);
@@ -1598,8 +1594,7 @@ err_lk:
 	kfree(idev->qp_table.table);
 err_qp:
 	ib_dealloc_device(dev);
-	_VERBS_ERROR("ib_ipath%d cannot register verbs (%d)!\n",
-		     dd->ipath_unit, -ret);
+	ipath_dev_err(dd, "cannot register verbs: %d!\n", -ret);
 	idev = NULL;
 
 bail:
@@ -1618,17 +1613,13 @@ void ipath_unregister_ib_device(struct i
 	if (!list_empty(&dev->pending[0]) ||
 	    !list_empty(&dev->pending[1]) ||
 	    !list_empty(&dev->pending[2]))
-		_VERBS_ERROR("ipath%d pending list not empty!\n",
-			     dev->ib_unit);
+		ipath_dev_err(dev->dd, "pending list not empty!\n");
 	if (!list_empty(&dev->piowait))
-		_VERBS_ERROR("ipath%d piowait list not empty!\n",
-			     dev->ib_unit);
+		ipath_dev_err(dev->dd, "piowait list not empty!\n");
 	if (!list_empty(&dev->rnrwait))
-		_VERBS_ERROR("ipath%d rnrwait list not empty!\n",
-			     dev->ib_unit);
+		ipath_dev_err(dev->dd, "rnrwait list not empty!\n");
 	if (!ipath_mcast_tree_empty())
-		_VERBS_ERROR("ipath%d multicast table memory leak!\n",
-			     dev->ib_unit);
+		ipath_dev_err(dev->dd, "multicast table memory leak!\n");
 	/*
 	 * Note that ipath_unregister_ib_device() can be called before all
 	 * the QPs are destroyed!
diff --git a/drivers/infiniband/hw/ipath/ipath_verbs.h b/drivers/infiniband/hw/ipath/ipath_verbs.h
--- a/drivers/infiniband/hw/ipath/ipath_verbs.h	Fri Aug 25 11:19:45 2006 -0700
+++ b/drivers/infiniband/hw/ipath/ipath_verbs.h	Fri Aug 25 11:19:45 2006 -0700
@@ -42,7 +42,6 @@
 #include <rdma/ib_pack.h>
 
 #include "ipath_layer.h"
-#include "verbs_debug.h"
 
 #define QPN_MAX                 (1 << 24)
 #define QPNMAP_ENTRIES          (QPN_MAX / PAGE_SIZE / BITS_PER_BYTE)
diff --git a/drivers/infiniband/hw/ipath/verbs_debug.h b/drivers/infiniband/hw/ipath/verbs_debug.h
deleted file mode 100644
--- a/drivers/infiniband/hw/ipath/verbs_debug.h	Fri Aug 25 11:19:45 2006 -0700
+++ /dev/null	Thu Jan 01 00:00:00 1970 +0000
@@ -1,108 +0,0 @@
-/*
- * Copyright (c) 2006 QLogic, Inc. All rights reserved.
- * Copyright (c) 2003, 2004, 2005, 2006 PathScale, Inc. All rights reserved.
- *
- * This software is available to you under a choice of one of two
- * licenses.  You may choose to be licensed under the terms of the GNU
- * General Public License (GPL) Version 2, available from the file
- * COPYING in the main directory of this source tree, or the
- * OpenIB.org BSD license below:
- *
- *     Redistribution and use in source and binary forms, with or
- *     without modification, are permitted provided that the following
- *     conditions are met:
- *
- *      - Redistributions of source code must retain the above
- *        copyright notice, this list of conditions and the following
- *        disclaimer.
- *
- *      - Redistributions in binary form must reproduce the above
- *        copyright notice, this list of conditions and the following
- *        disclaimer in the documentation and/or other materials
- *        provided with the distribution.
- *
- * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
- * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
- * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
- * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
- * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
- * 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 _VERBS_DEBUG_H
-#define _VERBS_DEBUG_H
-
-/*
- * This file contains tracing code for the ib_ipath kernel module.
- */
-#ifndef _VERBS_DEBUGGING	/* tracing enabled or not */
-#define _VERBS_DEBUGGING 1
-#endif
-
-extern unsigned ib_ipath_debug;
-
-#define _VERBS_ERROR(fmt,...) \
-	do { \
-		printk(KERN_ERR "%s: " fmt, "ib_ipath", ##__VA_ARGS__); \
-	} while(0)
-
-#define _VERBS_UNIT_ERROR(unit,fmt,...) \
-	do { \
-		printk(KERN_ERR "%s: " fmt, "ib_ipath", ##__VA_ARGS__); \
-	} while(0)
-
-#if _VERBS_DEBUGGING
-
-/*
- * Mask values for debugging.  The scheme allows us to compile out any
- * of the debug tracing stuff, and if compiled in, to enable or
- * disable dynamically.
- * This can be set at modprobe time also:
- *      modprobe ib_path ib_ipath_debug=3
- */
-
-#define __VERBS_INFO        0x1	/* generic low verbosity stuff */
-#define __VERBS_DBG         0x2	/* generic debug */
-#define __VERBS_VDBG        0x4	/* verbose debug */
-#define __VERBS_SMADBG      0x8000	/* sma packet debug */
-
-#define _VERBS_INFO(fmt,...) \
-	do { \
-		if (unlikely(ib_ipath_debug&__VERBS_INFO)) \
-			printk(KERN_INFO "%s: " fmt,"ib_ipath", \
-			       ##__VA_ARGS__); \
-	} while(0)
-
-#define _VERBS_DBG(fmt,...) \
-	do { \
-		if (unlikely(ib_ipath_debug&__VERBS_DBG)) \
-			printk(KERN_DEBUG "%s: " fmt, __func__, \
-			       ##__VA_ARGS__); \
-	} while(0)
-
-#define _VERBS_VDBG(fmt,...) \
-	do { \
-		if (unlikely(ib_ipath_debug&__VERBS_VDBG)) \
-			printk(KERN_DEBUG "%s: " fmt, __func__, \
-			       ##__VA_ARGS__); \
-	} while(0)
-
-#define _VERBS_SMADBG(fmt,...) \
-	do { \
-		if (unlikely(ib_ipath_debug&__VERBS_SMADBG)) \
-			printk(KERN_DEBUG "%s: " fmt, __func__, \
-			       ##__VA_ARGS__); \
-	} while(0)
-
-#else /* ! _VERBS_DEBUGGING */
-
-#define _VERBS_INFO(fmt,...)
-#define _VERBS_DBG(fmt,...)
-#define _VERBS_VDBG(fmt,...)
-#define _VERBS_SMADBG(fmt,...)
-
-#endif /* _VERBS_DEBUGGING */
-
-#endif /* _VERBS_DEBUG_H */

[PATCH 9 of 23] IB/ipath - remove stale references to userspace SMA

[Bryan O'Sullivan]

When we first submitted a userspace subnet management agent, it was
rejected, so we left it out of the final driver submission.  This patch
removes a number of vestigial references to it.

Signed-off-by: Bryan O'Sullivan <bryan.osullivan@qlogic.com>

diff --git a/drivers/infiniband/hw/ipath/ipath_common.h b/drivers/infiniband/hw/ipath/ipath_common.h
--- a/drivers/infiniband/hw/ipath/ipath_common.h	Fri Aug 25 11:19:45 2006 -0700
+++ b/drivers/infiniband/hw/ipath/ipath_common.h	Fri Aug 25 11:19:45 2006 -0700
@@ -106,9 +106,9 @@ struct infinipath_stats {
 	__u64 sps_ether_spkts;
 	/* number of "ethernet" packets received by driver */
 	__u64 sps_ether_rpkts;
-	/* number of SMA packets sent by driver */
+	/* number of SMA packets sent by driver. Obsolete. */
 	__u64 sps_sma_spkts;
-	/* number of SMA packets received by driver */
+	/* number of SMA packets received by driver. Obsolete. */
 	__u64 sps_sma_rpkts;
 	/* number of times all ports rcvhdrq was full and packet dropped */
 	__u64 sps_hdrqfull;
@@ -138,7 +138,7 @@ struct infinipath_stats {
 	__u64 sps_pageunlocks;
 	/*
 	 * Number of packets dropped in kernel other than errors (ether
-	 * packets if ipath not configured, sma/mad, etc.)
+	 * packets if ipath not configured, etc.)
 	 */
 	__u64 sps_krdrops;
 	/* pad for future growth */
@@ -153,8 +153,6 @@ struct infinipath_stats {
 #define IPATH_STATUS_DISABLED      0x2	/* hardware disabled */
 /* Device has been disabled via admin request */
 #define IPATH_STATUS_ADMIN_DISABLED    0x4
-#define IPATH_STATUS_OIB_SMA       0x8	/* ipath_mad kernel SMA running */
-#define IPATH_STATUS_SMA          0x10	/* user SMA running */
 /* Chip has been found and initted */
 #define IPATH_STATUS_CHIP_PRESENT 0x20
 /* IB link is at ACTIVE, usable for data traffic */
@@ -463,14 +461,6 @@ struct __ipath_sendpkt {
 	__u32 sps_cnt;		/* number of entries to use in sps_iov */
 	/* array of iov's describing packet. TEMPORARY */
 	struct ipath_iovec sps_iov[4];
-};
-
-/* Passed into SMA special file's ->read and ->write methods. */
-struct ipath_sma_pkt
-{
-	__u32 unit;	/* unit on which to send packet */
-	__u64 data;	/* address of payload in userspace */
-	__u32 len;	/* length of payload */
 };
 
 /*
diff --git a/drivers/infiniband/hw/ipath/ipath_debug.h b/drivers/infiniband/hw/ipath/ipath_debug.h
--- a/drivers/infiniband/hw/ipath/ipath_debug.h	Fri Aug 25 11:19:45 2006 -0700
+++ b/drivers/infiniband/hw/ipath/ipath_debug.h	Fri Aug 25 11:19:45 2006 -0700
@@ -60,7 +60,6 @@
 #define __IPATH_USER_SEND   0x1000	/* use user mode send */
 #define __IPATH_KERNEL_SEND 0x2000	/* use kernel mode send */
 #define __IPATH_EPKTDBG     0x4000	/* print ethernet packet data */
-#define __IPATH_SMADBG      0x8000	/* sma packet debug */
 #define __IPATH_IPATHDBG    0x10000	/* Ethernet (IPATH) gen debug */
 #define __IPATH_IPATHWARN   0x20000	/* Ethernet (IPATH) warnings */
 #define __IPATH_IPATHERR    0x40000	/* Ethernet (IPATH) errors */
@@ -84,7 +83,6 @@
 /* print mmap/nopage stuff, not using VDBG any more */
 #define __IPATH_MMDBG     0x0
 #define __IPATH_EPKTDBG   0x0	/* print ethernet packet data */
-#define __IPATH_SMADBG    0x0   /* process startup (init)/exit messages */
 #define __IPATH_IPATHDBG  0x0	/* Ethernet (IPATH) table dump on */
 #define __IPATH_IPATHWARN 0x0	/* Ethernet (IPATH) warnings on   */
 #define __IPATH_IPATHERR  0x0	/* Ethernet (IPATH) errors on   */
diff --git a/drivers/infiniband/hw/ipath/ipath_driver.c b/drivers/infiniband/hw/ipath/ipath_driver.c
--- a/drivers/infiniband/hw/ipath/ipath_driver.c	Fri Aug 25 11:19:45 2006 -0700
+++ b/drivers/infiniband/hw/ipath/ipath_driver.c	Fri Aug 25 11:19:45 2006 -0700
@@ -64,7 +64,7 @@ DEFINE_SPINLOCK(ipath_devs_lock);
 DEFINE_SPINLOCK(ipath_devs_lock);
 LIST_HEAD(ipath_dev_list);
 
-wait_queue_head_t ipath_sma_state_wait;
+wait_queue_head_t ipath_state_wait;
 
 unsigned ipath_debug = __IPATH_INFO;
 
@@ -618,15 +618,16 @@ static int ipath_wait_linkstate(struct i
 static int ipath_wait_linkstate(struct ipath_devdata *dd, u32 state,
 				int msecs)
 {
-	dd->ipath_sma_state_wanted = state;
-	wait_event_interruptible_timeout(ipath_sma_state_wait,
+	dd->ipath_state_wanted = state;
+	wait_event_interruptible_timeout(ipath_state_wait,
 					 (dd->ipath_flags & state),
 					 msecs_to_jiffies(msecs));
-	dd->ipath_sma_state_wanted = 0;
+	dd->ipath_state_wanted = 0;
 
 	if (!(dd->ipath_flags & state)) {
 		u64 val;
-		ipath_cdbg(SMA, "Didn't reach linkstate %s within %u ms\n",
+		ipath_cdbg(VERBOSE, "Didn't reach linkstate %s within %u"
+			   " ms\n",
 			   /* test INIT ahead of DOWN, both can be set */
 			   (state & IPATH_LINKINIT) ? "INIT" :
 			   ((state & IPATH_LINKDOWN) ? "DOWN" :
@@ -1155,7 +1156,7 @@ int ipath_setrcvhdrsize(struct ipath_dev
  *
  * do appropriate marking as busy, etc.
  * returns buffer number if one found (>=0), negative number is error.
- * Used by ipath_sma_send_pkt and ipath_layer_send
+ * Used by ipath_layer_send
  */
 u32 __iomem *ipath_getpiobuf(struct ipath_devdata *dd, u32 * pbufnum)
 {
@@ -1448,7 +1449,7 @@ static void ipath_set_ib_lstate(struct i
 	int linkcmd = (which >> INFINIPATH_IBCC_LINKCMD_SHIFT) &
 			INFINIPATH_IBCC_LINKCMD_MASK;
 
-	ipath_cdbg(SMA, "Trying to move unit %u to %s, current ltstate "
+	ipath_cdbg(VERBOSE, "Trying to move unit %u to %s, current ltstate "
 		   "is %s\n", dd->ipath_unit,
 		   what[linkcmd],
 		   ipath_ibcstatus_str[
@@ -1457,7 +1458,7 @@ static void ipath_set_ib_lstate(struct i
 			    INFINIPATH_IBCS_LINKTRAININGSTATE_SHIFT) &
 			   INFINIPATH_IBCS_LINKTRAININGSTATE_MASK]);
 	/* flush all queued sends when going to DOWN or INIT, to be sure that
-	 * they don't block SMA and other MAD packets */
+	 * they don't block MAD packets */
 	if (!linkcmd || linkcmd == INFINIPATH_IBCC_LINKCMD_INIT) {
 		ipath_write_kreg(dd, dd->ipath_kregs->kr_sendctrl,
 				 INFINIPATH_S_ABORT);
diff --git a/drivers/infiniband/hw/ipath/ipath_file_ops.c b/drivers/infiniband/hw/ipath/ipath_file_ops.c
--- a/drivers/infiniband/hw/ipath/ipath_file_ops.c	Fri Aug 25 11:19:45 2006 -0700
+++ b/drivers/infiniband/hw/ipath/ipath_file_ops.c	Fri Aug 25 11:19:45 2006 -0700
@@ -1816,7 +1816,7 @@ int ipath_user_add(struct ipath_devdata 
 		if (ret < 0) {
 			ipath_dev_err(dd, "Could not create wildcard "
 				      "minor: error %d\n", -ret);
-			goto bail_sma;
+			goto bail_user;
 		}
 
 		atomic_set(&user_setup, 1);
@@ -1832,7 +1832,7 @@ int ipath_user_add(struct ipath_devdata 
 
 	goto bail;
 
-bail_sma:
+bail_user:
 	user_cleanup();
 bail:
 	return ret;
diff --git a/drivers/infiniband/hw/ipath/ipath_fs.c b/drivers/infiniband/hw/ipath/ipath_fs.c
--- a/drivers/infiniband/hw/ipath/ipath_fs.c	Fri Aug 25 11:19:45 2006 -0700
+++ b/drivers/infiniband/hw/ipath/ipath_fs.c	Fri Aug 25 11:19:45 2006 -0700
@@ -191,8 +191,8 @@ static ssize_t atomic_port_info_read(str
 	portinfo[4] = (dd->ipath_lid << 16);
 
 	/*
-	 * Notimpl yet SMLID (should we store this in the driver, in case
-	 * SMA dies?)  CapabilityMask is 0, we don't support any of these
+	 * Notimpl yet SMLID.
+	 * CapabilityMask is 0, we don't support any of these
 	 * DiagCode is 0; we don't store any diag info for now Notimpl yet
 	 * M_KeyLeasePeriod (we don't support M_Key)
 	 */
diff --git a/drivers/infiniband/hw/ipath/ipath_init_chip.c b/drivers/infiniband/hw/ipath/ipath_init_chip.c
--- a/drivers/infiniband/hw/ipath/ipath_init_chip.c	Fri Aug 25 11:19:45 2006 -0700
+++ b/drivers/infiniband/hw/ipath/ipath_init_chip.c	Fri Aug 25 11:19:45 2006 -0700
@@ -53,8 +53,8 @@ MODULE_PARM_DESC(cfgports, "Set max numb
 MODULE_PARM_DESC(cfgports, "Set max number of ports to use");
 
 /*
- * Number of buffers reserved for driver (layered drivers and SMA
- * send).  Reserved at end of buffer list.   Initialized based on
+ * Number of buffers reserved for driver (verbs and layered drivers.)
+ * Reserved at end of buffer list.   Initialized based on
  * number of PIO buffers if not set via module interface.
  * The problem with this is that it's global, but we'll use different
  * numbers for different chip types.  So the default value is not
@@ -80,7 +80,7 @@ MODULE_PARM_DESC(kpiobufs, "Set number o
  *
  * Allocate the eager TID buffers and program them into infinipath.
  * We use the network layer alloc_skb() allocator to allocate the
- * memory, and either use the buffers as is for things like SMA
+ * memory, and either use the buffers as is for things like verbs
  * packets, or pass the buffers up to the ipath layered driver and
  * thence the network layer, replacing them as we do so (see
  * ipath_rcv_layer()).
@@ -450,9 +450,9 @@ static void enable_chip(struct ipath_dev
 	u32 val;
 	int i;
 
-	if (!reinit) {
-		init_waitqueue_head(&ipath_sma_state_wait);
-	}
+	if (!reinit)
+		init_waitqueue_head(&ipath_state_wait);
+
 	ipath_write_kreg(dd, dd->ipath_kregs->kr_rcvctrl,
 			 dd->ipath_rcvctrl);
 
diff --git a/drivers/infiniband/hw/ipath/ipath_intr.c b/drivers/infiniband/hw/ipath/ipath_intr.c
--- a/drivers/infiniband/hw/ipath/ipath_intr.c	Fri Aug 25 11:19:45 2006 -0700
+++ b/drivers/infiniband/hw/ipath/ipath_intr.c	Fri Aug 25 11:19:45 2006 -0700
@@ -201,7 +201,7 @@ static void handle_e_ibstatuschanged(str
 				  ib_linkstate(lstate));
 		}
 		else
-			ipath_cdbg(SMA, "Unit %u link state %s, last "
+			ipath_cdbg(VERBOSE, "Unit %u link state %s, last "
 				   "was %s\n", dd->ipath_unit,
 				   ib_linkstate(lstate),
 				   ib_linkstate((unsigned)
@@ -213,7 +213,7 @@ static void handle_e_ibstatuschanged(str
 		if (lstate == IPATH_IBSTATE_INIT ||
 		    lstate == IPATH_IBSTATE_ARM ||
 		    lstate == IPATH_IBSTATE_ACTIVE)
-			ipath_cdbg(SMA, "Unit %u link state down"
+			ipath_cdbg(VERBOSE, "Unit %u link state down"
 				   " (state 0x%x), from %s\n",
 				   dd->ipath_unit,
 				   (u32)val & IPATH_IBSTATE_MASK,
@@ -269,7 +269,7 @@ static void handle_e_ibstatuschanged(str
 			     INFINIPATH_IBCS_LINKSTATE_MASK)
 			    == INFINIPATH_IBCS_L_STATE_ACTIVE)
 				/* if from up to down be more vocal */
-				ipath_cdbg(SMA,
+				ipath_cdbg(VERBOSE,
 					   "Unit %u link now down (%s)\n",
 					   dd->ipath_unit,
 					   ipath_ibcstatus_str[ltstate]);
@@ -596,11 +596,11 @@ static int handle_errors(struct ipath_de
 
 	if (!noprint && *msg)
 		ipath_dev_err(dd, "%s error\n", msg);
-	if (dd->ipath_sma_state_wanted & dd->ipath_flags) {
-		ipath_cdbg(VERBOSE, "sma wanted state %x, iflags now %x, "
-			   "waking\n", dd->ipath_sma_state_wanted,
+	if (dd->ipath_state_wanted & dd->ipath_flags) {
+		ipath_cdbg(VERBOSE, "driver wanted state %x, iflags now %x, "
+			   "waking\n", dd->ipath_state_wanted,
 			   dd->ipath_flags);
-		wake_up_interruptible(&ipath_sma_state_wait);
+		wake_up_interruptible(&ipath_state_wait);
 	}
 
 	return chkerrpkts;
diff --git a/drivers/infiniband/hw/ipath/ipath_kernel.h b/drivers/infiniband/hw/ipath/ipath_kernel.h
--- a/drivers/infiniband/hw/ipath/ipath_kernel.h	Fri Aug 25 11:19:45 2006 -0700
+++ b/drivers/infiniband/hw/ipath/ipath_kernel.h	Fri Aug 25 11:19:45 2006 -0700
@@ -245,8 +245,8 @@ struct ipath_devdata {
 	u32 ipath_pioavregs;
 	/* IPATH_POLL, etc. */
 	u32 ipath_flags;
-	/* ipath_flags sma is waiting for */
-	u32 ipath_sma_state_wanted;
+	/* ipath_flags driver is waiting for */
+	u32 ipath_state_wanted;
 	/* last buffer for user use, first buf for kernel use is this
 	 * index. */
 	u32 ipath_lastport_piobuf;
@@ -306,10 +306,6 @@ struct ipath_devdata {
 	u32 ipath_pcibar0;
 	/* so we can rewrite it after a chip reset */
 	u32 ipath_pcibar1;
-	/* sequential tries for SMA send and no bufs */
-	u32 ipath_nosma_bufs;
-	/* duration (seconds) ipath_nosma_bufs set */
-	u32 ipath_nosma_secs;
 
 	/* HT/PCI Vendor ID (here for NodeInfo) */
 	u16 ipath_vendorid;
@@ -534,7 +530,7 @@ void ipath_diag_remove(struct ipath_devd
 void ipath_diag_remove(struct ipath_devdata *);
 void ipath_diag_bringup_link(struct ipath_devdata *);
 
-extern wait_queue_head_t ipath_sma_state_wait;
+extern wait_queue_head_t ipath_state_wait;
 
 int ipath_user_add(struct ipath_devdata *dd);
 void ipath_user_remove(struct ipath_devdata *dd);
@@ -818,7 +814,6 @@ extern struct mutex ipath_mutex;
 #define IPATH_DRV_NAME		"ib_ipath"
 #define IPATH_MAJOR		233
 #define IPATH_USER_MINOR_BASE	0
-#define IPATH_SMA_MINOR		128
 #define IPATH_DIAG_MINOR_BASE	129
 #define IPATH_NMINORS		255
 
diff --git a/drivers/infiniband/hw/ipath/ipath_layer.c b/drivers/infiniband/hw/ipath/ipath_layer.c
--- a/drivers/infiniband/hw/ipath/ipath_layer.c	Fri Aug 25 11:19:45 2006 -0700
+++ b/drivers/infiniband/hw/ipath/ipath_layer.c	Fri Aug 25 11:19:45 2006 -0700
@@ -161,9 +161,6 @@ int ipath_layer_register(void *(*l_add)(
 
 		if (dd->ipath_layer.l_arg)
 			continue;
-
-		if (!(*dd->ipath_statusp & IPATH_STATUS_SMA))
-			*dd->ipath_statusp |= IPATH_STATUS_OIB_SMA;
 
 		spin_unlock_irqrestore(&ipath_devs_lock, flags);
 		dd->ipath_layer.l_arg = l_add(dd->ipath_unit, dd);
diff --git a/drivers/infiniband/hw/ipath/ipath_layer.h b/drivers/infiniband/hw/ipath/ipath_layer.h
--- a/drivers/infiniband/hw/ipath/ipath_layer.h	Fri Aug 25 11:19:45 2006 -0700
+++ b/drivers/infiniband/hw/ipath/ipath_layer.h	Fri Aug 25 11:19:45 2006 -0700
@@ -66,9 +66,6 @@ int ipath_layer_set_piointbufavail_int(s
 #define IPATH_LAYER_INT_SEND_CONTINUE 0x10
 #define IPATH_LAYER_INT_BCAST 0x40
 
-/* _verbs_layer.l_flags */
-#define IPATH_VERBS_KERNEL_SMA 0x1
-
 extern unsigned ipath_debug; /* debugging bit mask */
 
 #endif				/* _IPATH_LAYER_H */
diff --git a/drivers/infiniband/hw/ipath/ipath_qp.c b/drivers/infiniband/hw/ipath/ipath_qp.c
--- a/drivers/infiniband/hw/ipath/ipath_qp.c	Fri Aug 25 11:19:45 2006 -0700
+++ b/drivers/infiniband/hw/ipath/ipath_qp.c	Fri Aug 25 11:19:45 2006 -0700
@@ -645,33 +645,6 @@ __be32 ipath_compute_aeth(struct ipath_q
 }
 
 /**
- * set_verbs_flags - set the verbs layer flags
- * @dd: the infinipath device
- * @flags: the flags to set
- */
-static int set_verbs_flags(struct ipath_devdata *dd, unsigned flags)
-{
-	struct ipath_devdata *ss;
-	unsigned long lflags;
-
-	spin_lock_irqsave(&ipath_devs_lock, lflags);
-
-	list_for_each_entry(ss, &ipath_dev_list, ipath_list) {
-		if (!(ss->ipath_flags & IPATH_INITTED))
-			continue;
-		if ((flags & IPATH_VERBS_KERNEL_SMA) &&
-		    !(*ss->ipath_statusp & IPATH_STATUS_SMA))
-			*ss->ipath_statusp |= IPATH_STATUS_OIB_SMA;
-		else
-			*ss->ipath_statusp &= ~IPATH_STATUS_OIB_SMA;
-	}
-
-	spin_unlock_irqrestore(&ipath_devs_lock, lflags);
-
-	return 0;
-}
-
-/**
  * ipath_create_qp - create a queue pair for a device
  * @ibpd: the protection domain who's device we create the queue pair for
  * @init_attr: the attributes of the queue pair
@@ -784,10 +757,6 @@ struct ib_qp *ipath_create_qp(struct ib_
 		}
 		qp->ip = NULL;
 		ipath_reset_qp(qp);
-
-		/* Tell the core driver that the kernel SMA is present. */
-		if (init_attr->qp_type == IB_QPT_SMI)
-			set_verbs_flags(dev->dd, IPATH_VERBS_KERNEL_SMA);
 		break;
 
 	default:
@@ -862,10 +831,6 @@ int ipath_destroy_qp(struct ib_qp *ibqp)
 	struct ipath_ibdev *dev = to_idev(ibqp->device);
 	unsigned long flags;
 
-	/* Tell the core driver that the kernel SMA is gone. */
-	if (qp->ibqp.qp_type == IB_QPT_SMI)
-		set_verbs_flags(dev->dd, 0);
-
 	spin_lock_irqsave(&qp->s_lock, flags);
 	qp->state = IB_QPS_ERR;
 	spin_unlock_irqrestore(&qp->s_lock, flags);
diff --git a/drivers/infiniband/hw/ipath/ipath_stats.c b/drivers/infiniband/hw/ipath/ipath_stats.c
--- a/drivers/infiniband/hw/ipath/ipath_stats.c	Fri Aug 25 11:19:45 2006 -0700
+++ b/drivers/infiniband/hw/ipath/ipath_stats.c	Fri Aug 25 11:19:45 2006 -0700
@@ -271,33 +271,6 @@ void ipath_get_faststats(unsigned long o
 		}
 	}
 
-	if (dd->ipath_nosma_bufs) {
-		dd->ipath_nosma_secs += 5;
-		if (dd->ipath_nosma_secs >= 30) {
-			ipath_cdbg(SMA, "No SMA bufs avail %u seconds; "
-				   "cancelling pending sends\n",
-				   dd->ipath_nosma_secs);
-			/*
-			 * issue an abort as well, in case we have a packet
-			 * stuck in launch fifo.  This could corrupt an
-			 * outgoing user packet in the worst case,
-			 * but this is a pretty catastrophic, anyway.
-			 */
-			ipath_write_kreg(dd, dd->ipath_kregs->kr_sendctrl,
-					 INFINIPATH_S_ABORT);
-			ipath_disarm_piobufs(dd, dd->ipath_lastport_piobuf,
-					     dd->ipath_piobcnt2k +
-					     dd->ipath_piobcnt4k -
-					     dd->ipath_lastport_piobuf);
-			/* start again, if necessary */
-			dd->ipath_nosma_secs = 0;
-		} else
-			ipath_cdbg(SMA, "No SMA bufs avail %u tries, "
-				   "after %u seconds\n",
-				   dd->ipath_nosma_bufs,
-				   dd->ipath_nosma_secs);
-	}
-
 done:
 	mod_timer(&dd->ipath_stats_timer, jiffies + HZ * 5);
 }
diff --git a/drivers/infiniband/hw/ipath/ipath_sysfs.c b/drivers/infiniband/hw/ipath/ipath_sysfs.c
--- a/drivers/infiniband/hw/ipath/ipath_sysfs.c	Fri Aug 25 11:19:45 2006 -0700
+++ b/drivers/infiniband/hw/ipath/ipath_sysfs.c	Fri Aug 25 11:19:45 2006 -0700
@@ -107,8 +107,8 @@ static const char *ipath_status_str[] = 
 	"Initted",
 	"Disabled",
 	"Admin_Disabled",
-	"OIB_SMA",
-	"SMA",
+	"", /* This used to be the old "OIB_SMA" status. */
+	"", /* This used to be the old "SMA" status. */
 	"Present",
 	"IB_link_up",
 	"IB_configured",
diff --git a/drivers/infiniband/hw/ipath/ipath_verbs.c b/drivers/infiniband/hw/ipath/ipath_verbs.c
--- a/drivers/infiniband/hw/ipath/ipath_verbs.c	Fri Aug 25 11:19:45 2006 -0700
+++ b/drivers/infiniband/hw/ipath/ipath_verbs.c	Fri Aug 25 11:19:45 2006 -0700
@@ -1573,7 +1573,7 @@ int ipath_register_ib_device(struct ipat
 	dev->mmap = ipath_mmap;
 
 	snprintf(dev->node_desc, sizeof(dev->node_desc),
-		 IPATH_IDSTR " %s kernel_SMA", system_utsname.nodename);
+		 IPATH_IDSTR " %s", system_utsname.nodename);
 
 	ret = ib_register_device(dev);
 	if (ret)

[PATCH 10 of 23] IB/ipath - trivial cleanups

[Bryan O'Sullivan]

Signed-off-by: Bryan O'Sullivan <bryan.osullivan@qlogic.com>

diff --git a/drivers/infiniband/hw/ipath/ipath_kernel.h b/drivers/infiniband/hw/ipath/ipath_kernel.h
--- a/drivers/infiniband/hw/ipath/ipath_kernel.h	Fri Aug 25 11:19:45 2006 -0700
+++ b/drivers/infiniband/hw/ipath/ipath_kernel.h	Fri Aug 25 11:19:45 2006 -0700
@@ -528,7 +528,6 @@ void ipath_cdev_cleanup(struct cdev **cd
 
 int ipath_diag_add(struct ipath_devdata *);
 void ipath_diag_remove(struct ipath_devdata *);
-void ipath_diag_bringup_link(struct ipath_devdata *);
 
 extern wait_queue_head_t ipath_state_wait;
 

[PATCH 11 of 23] IB/ipath - add new minor device to allow sending of diag packets

[Bryan O'Sullivan]

Signed-off-by: Bryan O'Sullivan <bryan.osullivan@qlogic.com>

diff --git a/drivers/infiniband/hw/ipath/ipath_common.h b/drivers/infiniband/hw/ipath/ipath_common.h
--- a/drivers/infiniband/hw/ipath/ipath_common.h	Fri Aug 25 11:19:45 2006 -0700
+++ b/drivers/infiniband/hw/ipath/ipath_common.h	Fri Aug 25 11:19:45 2006 -0700
@@ -461,6 +461,13 @@ struct __ipath_sendpkt {
 	__u32 sps_cnt;		/* number of entries to use in sps_iov */
 	/* array of iov's describing packet. TEMPORARY */
 	struct ipath_iovec sps_iov[4];
+};
+
+/* Passed into diag data special file's ->write method. */
+struct ipath_diag_pkt {
+	__u32 unit;
+	__u64 data;
+	__u32 len;
 };
 
 /*
diff --git a/drivers/infiniband/hw/ipath/ipath_diag.c b/drivers/infiniband/hw/ipath/ipath_diag.c
--- a/drivers/infiniband/hw/ipath/ipath_diag.c	Fri Aug 25 11:19:45 2006 -0700
+++ b/drivers/infiniband/hw/ipath/ipath_diag.c	Fri Aug 25 11:19:45 2006 -0700
@@ -41,6 +41,7 @@
  * through the /sys/bus/pci resource mmap interface.
  */
 
+#include <linux/io.h>
 #include <linux/pci.h>
 #include <asm/uaccess.h>
 
@@ -273,6 +274,158 @@ bail:
 	return ret;
 }
 
+static ssize_t ipath_diagpkt_write(struct file *fp,
+				   const char __user *data,
+				   size_t count, loff_t *off);
+
+static struct file_operations diagpkt_file_ops = {
+	.owner = THIS_MODULE,
+	.write = ipath_diagpkt_write,
+};
+
+static struct cdev *diagpkt_cdev;
+static struct class_device *diagpkt_class_dev;
+
+int __init ipath_diagpkt_add(void)
+{
+	return ipath_cdev_init(IPATH_DIAGPKT_MINOR,
+			       "ipath_diagpkt", &diagpkt_file_ops,
+			       &diagpkt_cdev, &diagpkt_class_dev);
+}
+
+void __exit ipath_diagpkt_remove(void)
+{
+	ipath_cdev_cleanup(&diagpkt_cdev, &diagpkt_class_dev);
+}
+
+/**
+ * ipath_diagpkt_write - write an IB packet
+ * @fp: the diag data device file pointer
+ * @data: ipath_diag_pkt structure saying where to get the packet
+ * @count: size of data to write
+ * @off: unused by this code
+ */
+static ssize_t ipath_diagpkt_write(struct file *fp,
+				   const char __user *data,
+				   size_t count, loff_t *off)
+{
+	u32 __iomem *piobuf;
+	u32 plen, clen, pbufn;
+	struct ipath_diag_pkt dp;
+	u32 *tmpbuf = NULL;
+	struct ipath_devdata *dd;
+	ssize_t ret = 0;
+	u64 val;
+
+	if (count < sizeof(dp)) {
+		ret = -EINVAL;
+		goto bail;
+	}
+
+	if (copy_from_user(&dp, data, sizeof(dp))) {
+		ret = -EFAULT;
+		goto bail;
+	}
+
+	/* send count must be an exact number of dwords */
+	if (dp.len & 3) {
+		ret = -EINVAL;
+		goto bail;
+	}
+
+	clen = dp.len >> 2;
+
+	dd = ipath_lookup(dp.unit);
+	if (!dd || !(dd->ipath_flags & IPATH_PRESENT) ||
+	    !dd->ipath_kregbase) {
+		ipath_cdbg(VERBOSE, "illegal unit %u for diag data send\n",
+			   dp.unit);
+		ret = -ENODEV;
+		goto bail;
+	}
+
+	if (ipath_diag_inuse && !diag_set_link &&
+	    !(dd->ipath_flags & IPATH_LINKACTIVE)) {
+		diag_set_link = 1;
+		ipath_cdbg(VERBOSE, "Trying to set to set link active for "
+			   "diag pkt\n");
+		ipath_set_linkstate(dd, IPATH_IB_LINKARM);
+		ipath_set_linkstate(dd, IPATH_IB_LINKACTIVE);
+	}
+
+	if (!(dd->ipath_flags & IPATH_INITTED)) {
+		/* no hardware, freeze, etc. */
+		ipath_cdbg(VERBOSE, "unit %u not usable\n", dd->ipath_unit);
+		ret = -ENODEV;
+		goto bail;
+	}
+	val = dd->ipath_lastibcstat & IPATH_IBSTATE_MASK;
+	if (val != IPATH_IBSTATE_INIT && val != IPATH_IBSTATE_ARM &&
+	    val != IPATH_IBSTATE_ACTIVE) {
+		ipath_cdbg(VERBOSE, "unit %u not ready (state %llx)\n",
+			   dd->ipath_unit, (unsigned long long) val);
+		ret = -EINVAL;
+		goto bail;
+	}
+
+	/* need total length before first word written */
+	/* +1 word is for the qword padding */
+	plen = sizeof(u32) + dp.len;
+
+	if ((plen + 4) > dd->ipath_ibmaxlen) {
+		ipath_dbg("Pkt len 0x%x > ibmaxlen %x\n",
+			  plen - 4, dd->ipath_ibmaxlen);
+		ret = -EINVAL;
+		goto bail;	/* before writing pbc */
+	}
+	tmpbuf = vmalloc(plen);
+	if (!tmpbuf) {
+		dev_info(&dd->pcidev->dev, "Unable to allocate tmp buffer, "
+			 "failing\n");
+		ret = -ENOMEM;
+		goto bail;
+	}
+
+	if (copy_from_user(tmpbuf,
+			   (const void __user *) (unsigned long) dp.data,
+			   dp.len)) {
+		ret = -EFAULT;
+		goto bail;
+	}
+
+	piobuf = ipath_getpiobuf(dd, &pbufn);
+	if (!piobuf) {
+		ipath_cdbg(VERBOSE, "No PIO buffers avail unit for %u\n",
+			   dd->ipath_unit);
+		ret = -EBUSY;
+		goto bail;
+	}
+
+	plen >>= 2;		/* in dwords */
+
+	if (ipath_debug & __IPATH_PKTDBG)
+		ipath_cdbg(VERBOSE, "unit %u 0x%x+1w pio%d\n",
+			   dd->ipath_unit, plen - 1, pbufn);
+
+	/* we have to flush after the PBC for correctness on some cpus
+	 * or WC buffer can be written out of order */
+	writeq(plen, piobuf);
+	ipath_flush_wc();
+	/* copy all by the trigger word, then flush, so it's written
+	 * to chip before trigger word, then write trigger word, then
+	 * flush again, so packet is sent. */
+	__iowrite32_copy(piobuf + 2, tmpbuf, clen - 1);
+	ipath_flush_wc();
+	__raw_writel(tmpbuf[clen - 1], piobuf + clen + 1);
+	ipath_flush_wc();
+
+	ret = sizeof(dp);
+
+bail:
+	vfree(tmpbuf);
+	return ret;
+}
+
 static int ipath_diag_release(struct inode *in, struct file *fp)
 {
 	mutex_lock(&ipath_mutex);
diff --git a/drivers/infiniband/hw/ipath/ipath_driver.c b/drivers/infiniband/hw/ipath/ipath_driver.c
--- a/drivers/infiniband/hw/ipath/ipath_driver.c	Fri Aug 25 11:19:45 2006 -0700
+++ b/drivers/infiniband/hw/ipath/ipath_driver.c	Fri Aug 25 11:19:45 2006 -0700
@@ -1881,7 +1881,17 @@ static int __init infinipath_init(void)
 		goto bail_group;
 	}
 
+	ret = ipath_diagpkt_add();
+	if (ret < 0) {
+		printk(KERN_ERR IPATH_DRV_NAME ": Unable to create "
+		       "diag data device: error %d\n", -ret);
+		goto bail_ipathfs;
+	}
+ 
 	goto bail;
+
+bail_ipathfs:
+	ipath_exit_ipathfs();
 
 bail_group:
 	ipath_driver_remove_group(&ipath_driver.driver);
@@ -1993,6 +2003,8 @@ static void __exit infinipath_cleanup(vo
 	struct ipath_devdata *dd, *tmp;
 	unsigned long flags;
 
+	ipath_diagpkt_remove();
+
 	ipath_exit_ipathfs();
 
 	ipath_driver_remove_group(&ipath_driver.driver);
diff --git a/drivers/infiniband/hw/ipath/ipath_kernel.h b/drivers/infiniband/hw/ipath/ipath_kernel.h
--- a/drivers/infiniband/hw/ipath/ipath_kernel.h	Fri Aug 25 11:19:45 2006 -0700
+++ b/drivers/infiniband/hw/ipath/ipath_kernel.h	Fri Aug 25 11:19:45 2006 -0700
@@ -789,6 +789,9 @@ void ipath_device_remove_group(struct de
 void ipath_device_remove_group(struct device *, struct ipath_devdata *);
 int ipath_expose_reset(struct device *);
 
+int ipath_diagpkt_add(void);
+void ipath_diagpkt_remove(void);
+
 int ipath_init_ipathfs(void);
 void ipath_exit_ipathfs(void);
 int ipathfs_add_device(struct ipath_devdata *);
@@ -813,6 +816,7 @@ extern struct mutex ipath_mutex;
 #define IPATH_DRV_NAME		"ib_ipath"
 #define IPATH_MAJOR		233
 #define IPATH_USER_MINOR_BASE	0
+#define IPATH_DIAGPKT_MINOR	127
 #define IPATH_DIAG_MINOR_BASE	129
 #define IPATH_NMINORS		255
 

Re: [PATCH 11 of 23] IB/ipath - add new minor device to allow sending of diag packets

[Roland Dreier]

 > +	if (ret < 0) {
 > +		printk(KERN_ERR IPATH_DRV_NAME ": Unable to create "
 > +		       "diag data device: error %d\n", -ret);
 > +		goto bail_ipathfs;
 > +	}
 > + 

The last line adds trailing whitespace, which git complains about.
When patchbombing, can you run your patches through "git apply --check
--whitespace=error-all" or the equivalent?

Thanks,
  Roland

Re: [PATCH 11 of 23] IB/ipath - add new minor device to allow sending of diag packets

[Bryan O'Sullivan]

On Fri, 2006-08-25 at 12:50 -0700, Roland Dreier wrote:

> The last line adds trailing whitespace, which git complains about.
> When patchbombing, can you run your patches through "git apply --check
> --whitespace=error-all" or the equivalent?

Sure.  Thanks for spotting that.

	<b

[PATCH 12 of 23] IB/ipath - do not allow use of CQ entries with invalid counts

[Bryan O'Sullivan]

Signed-off-by: Bryan O'Sullivan <bryan.osullivan@qlogic.com>

diff --git a/drivers/infiniband/hw/ipath/ipath_cq.c b/drivers/infiniband/hw/ipath/ipath_cq.c
--- a/drivers/infiniband/hw/ipath/ipath_cq.c	Fri Aug 25 11:19:45 2006 -0700
+++ b/drivers/infiniband/hw/ipath/ipath_cq.c	Fri Aug 25 11:19:45 2006 -0700
@@ -172,7 +172,7 @@ struct ib_cq *ipath_create_cq(struct ib_
 	struct ipath_cq_wc *wc;
 	struct ib_cq *ret;
 
-	if (entries > ib_ipath_max_cqes) {
+	if (entries < 1 || entries > ib_ipath_max_cqes) {
 		ret = ERR_PTR(-EINVAL);
 		goto done;
 	}
@@ -324,6 +324,11 @@ int ipath_resize_cq(struct ib_cq *ibcq, 
 	u32 head, tail, n;
 	int ret;
 
+	if (cqe < 1 || cqe > ib_ipath_max_cqes) {
+		ret = -EINVAL;
+		goto bail;
+	}
+
 	/*
 	 * Need to use vmalloc() if we want to support large #s of entries.
 	 */

[PATCH 13 of 23] IB/ipath - account for attached QPs correctly

[Bryan O'Sullivan]

Signed-off-by: Bryan O'Sullivan <bryan.osullivan@qlogic.com>

diff --git a/drivers/infiniband/hw/ipath/ipath_verbs_mcast.c b/drivers/infiniband/hw/ipath/ipath_verbs_mcast.c
--- a/drivers/infiniband/hw/ipath/ipath_verbs_mcast.c	Fri Aug 25 11:19:45 2006 -0700
+++ b/drivers/infiniband/hw/ipath/ipath_verbs_mcast.c	Fri Aug 25 11:19:45 2006 -0700
@@ -217,6 +217,8 @@ static int ipath_mcast_add(struct ipath_
 	dev->n_mcast_grps_allocated++;
 	spin_unlock(&dev->n_mcast_grps_lock);
 
+	mcast->n_attached++;
+
 	list_add_tail_rcu(&mqp->list, &mcast->qp_list);
 
 	atomic_inc(&mcast->refcount);

[PATCH 14 of 23] IB/ipath - support new QLogic product naming scheme

[Bryan O'Sullivan]

This patch only renames files, fixes product names, and updates
comments.

Signed-off-by: Bryan O'Sullivan <bryan.osullivan@qlogic.com>

diff --git a/drivers/infiniband/hw/ipath/Makefile b/drivers/infiniband/hw/ipath/Makefile
--- a/drivers/infiniband/hw/ipath/Makefile	Fri Aug 25 11:19:45 2006 -0700
+++ b/drivers/infiniband/hw/ipath/Makefile	Fri Aug 25 11:19:45 2006 -0700
@@ -10,7 +10,8 @@ ib_ipath-y := \
 	ipath_eeprom.o \
 	ipath_file_ops.o \
 	ipath_fs.o \
-	ipath_ht400.o \
+	ipath_iba6110.o \
+	ipath_iba6120.o \
 	ipath_init_chip.o \
 	ipath_intr.o \
 	ipath_keys.o \
@@ -18,7 +19,6 @@ ib_ipath-y := \
 	ipath_mad.o \
 	ipath_mmap.o \
 	ipath_mr.o \
-	ipath_pe800.o \
 	ipath_qp.o \
 	ipath_rc.o \
 	ipath_ruc.o \
diff --git a/drivers/infiniband/hw/ipath/ipath_driver.c b/drivers/infiniband/hw/ipath/ipath_driver.c
--- a/drivers/infiniband/hw/ipath/ipath_driver.c	Fri Aug 25 11:19:45 2006 -0700
+++ b/drivers/infiniband/hw/ipath/ipath_driver.c	Fri Aug 25 11:19:45 2006 -0700
@@ -401,10 +401,10 @@ static int __devinit ipath_init_one(stru
 	/* setup the chip-specific functions, as early as possible. */
 	switch (ent->device) {
 	case PCI_DEVICE_ID_INFINIPATH_HT:
-		ipath_init_ht400_funcs(dd);
+		ipath_init_iba6110_funcs(dd);
 		break;
 	case PCI_DEVICE_ID_INFINIPATH_PE800:
-		ipath_init_pe800_funcs(dd);
+		ipath_init_iba6120_funcs(dd);
 		break;
 	default:
 		ipath_dev_err(dd, "Found unknown QLogic deviceid 0x%x, "
@@ -969,7 +969,8 @@ reloop:
 		 */
 		if (l == hdrqtail || (i && !(i&0xf))) {
 			u64 lval;
-			if (l == hdrqtail) /* PE-800 interrupt only on last */
+			if (l == hdrqtail)
+				/* request IBA6120 interrupt only on last */
 				lval = dd->ipath_rhdrhead_intr_off | l;
 			else
 				lval = l;
@@ -983,7 +984,7 @@ reloop:
 	}
 
 	if (!dd->ipath_rhdrhead_intr_off && !reloop) {
-		/* HT-400 workaround; we can have a race clearing chip
+		/* IBA6110 workaround; we can have a race clearing chip
 		 * interrupt with another interrupt about to be delivered,
 		 * and can clear it before it is delivered on the GPIO
 		 * workaround.  By doing the extra check here for the
diff --git a/drivers/infiniband/hw/ipath/ipath_file_ops.c b/drivers/infiniband/hw/ipath/ipath_file_ops.c
--- a/drivers/infiniband/hw/ipath/ipath_file_ops.c	Fri Aug 25 11:19:45 2006 -0700
+++ b/drivers/infiniband/hw/ipath/ipath_file_ops.c	Fri Aug 25 11:19:45 2006 -0700
@@ -1110,7 +1110,7 @@ static int ipath_mmap(struct file *fp, s
 		ret = mmap_rcvegrbufs(vma, pd);
 	else if (pgaddr == (u64) pd->port_rcvhdrq_phys) {
 		/*
-		 * The rcvhdrq itself; readonly except on HT-400 (so have
+		 * The rcvhdrq itself; readonly except on HT (so have
 		 * to allow writable mapping), multiple pages, contiguous
 		 * from an i/o perspective.
 		 */
@@ -1298,14 +1298,14 @@ static int find_best_unit(struct file *f
 	 * This code is present to allow a knowledgeable person to
 	 * specify the layout of processes to processors before opening
 	 * this driver, and then we'll assign the process to the "closest"
-	 * HT-400 to that processor (we assume reasonable connectivity,
+	 * InfiniPath chip to that processor (we assume reasonable connectivity,
 	 * for now).  This code assumes that if affinity has been set
 	 * before this point, that at most one cpu is set; for now this
 	 * is reasonable.  I check for both cpus_empty() and cpus_full(),
 	 * in case some kernel variant sets none of the bits when no
 	 * affinity is set.  2.6.11 and 12 kernels have all present
 	 * cpus set.  Some day we'll have to fix it up further to handle
-	 * a cpu subset.  This algorithm fails for two HT-400's connected
+	 * a cpu subset.  This algorithm fails for two HT chips connected
 	 * in tunnel fashion.  Eventually this needs real topology
 	 * information.  There may be some issues with dual core numbering
 	 * as well.  This needs more work prior to release.
diff --git a/drivers/infiniband/hw/ipath/ipath_ht400.c b/drivers/infiniband/hw/ipath/ipath_iba6110.c
rename from drivers/infiniband/hw/ipath/ipath_ht400.c
rename to drivers/infiniband/hw/ipath/ipath_iba6110.c
--- a/drivers/infiniband/hw/ipath/ipath_iba6110.c	Fri Aug 25 11:19:45 2006 -0700
+++ b/drivers/infiniband/hw/ipath/ipath_iba6110.c	Fri Aug 25 11:19:45 2006 -0700
@@ -33,7 +33,7 @@
 
 /*
  * This file contains all of the code that is specific to the InfiniPath
- * HT-400 chip.
+ * HT chip.
  */
 
 #include <linux/pci.h>
@@ -43,7 +43,7 @@
 #include "ipath_registers.h"
 
 /*
- * This lists the InfiniPath HT400 registers, in the actual chip layout.
+ * This lists the InfiniPath registers, in the actual chip layout.
  * This structure should never be directly accessed.
  *
  * The names are in InterCap form because they're taken straight from
@@ -537,7 +537,7 @@ static void ipath_ht_handle_hwerrors(str
 	if (hwerrs & INFINIPATH_HWE_HTCMISCERR7)
 		strlcat(msg, "[HT core Misc7]", msgl);
 	if (hwerrs & INFINIPATH_HWE_MEMBISTFAILED) {
-		strlcat(msg, "[Memory BIST test failed, HT-400 unusable]",
+		strlcat(msg, "[Memory BIST test failed, InfiniPath hardware unusable]",
 			msgl);
 		/* ignore from now on, so disable until driver reloaded */
 		dd->ipath_hwerrmask &= ~INFINIPATH_HWE_MEMBISTFAILED;
@@ -553,7 +553,7 @@ static void ipath_ht_handle_hwerrors(str
 
 	if (hwerrs & _IPATH_PLL_FAIL) {
 		snprintf(bitsmsg, sizeof bitsmsg,
-			 "[PLL failed (%llx), HT-400 unusable]",
+			 "[PLL failed (%llx), InfiniPath hardware unusable]",
 			 (unsigned long long) (hwerrs & _IPATH_PLL_FAIL));
 		strlcat(msg, bitsmsg, msgl);
 		/* ignore from now on, so disable until driver reloaded */
@@ -610,18 +610,18 @@ static int ipath_ht_boardname(struct ipa
 		break;
 	case 5:
 		/*
-		 * HT-460 original production board; two production levels, with
+		 * original production board; two production levels, with
 		 * different serial number ranges.   See ipath_ht_early_init() for
 		 * case where we enable IPATH_GPIO_INTR for later serial # range.
 		 */
-		n = "InfiniPath_HT-460";
+		n = "InfiniPath_QHT7040";
 		break;
 	case 6:
 		n = "OEM_Board_3";
 		break;
 	case 7:
-		/* HT-460 small form factor production board */
-		n = "InfiniPath_HT-465";
+		/* small form factor production board */
+		n = "InfiniPath_QHT7140";
 		break;
 	case 8:
 		n = "LS/X-1";
@@ -633,7 +633,7 @@ static int ipath_ht_boardname(struct ipa
 		n = "OEM_Board_2";
 		break;
 	case 11:
-		n = "InfiniPath_HT-470";
+		n = "InfiniPath_HT-470"; /* obsoleted */
 		break;
 	case 12:
 		n = "OEM_Board_4";
@@ -641,7 +641,7 @@ static int ipath_ht_boardname(struct ipa
 	default:		/* don't know, just print the number */
 		ipath_dev_err(dd, "Don't yet know about board "
 			      "with ID %u\n", boardrev);
-		snprintf(name, namelen, "Unknown_InfiniPath_HT-4xx_%u",
+		snprintf(name, namelen, "Unknown_InfiniPath_QHT7xxx_%u",
 			 boardrev);
 		break;
 	}
@@ -650,11 +650,10 @@ static int ipath_ht_boardname(struct ipa
 
 	if (dd->ipath_majrev != 3 || (dd->ipath_minrev < 2 || dd->ipath_minrev > 3)) {
 		/*
-		 * This version of the driver only supports the HT-400
-		 * Rev 3.2
+		 * This version of the driver only supports Rev 3.2 and 3.3
 		 */
 		ipath_dev_err(dd,
-			      "Unsupported HT-400 revision %u.%u!\n",
+			      "Unsupported InfiniPath hardware revision %u.%u!\n",
 			      dd->ipath_majrev, dd->ipath_minrev);
 		ret = 1;
 		goto bail;
@@ -738,7 +737,7 @@ static void ipath_check_htlink(struct ip
 
 static int ipath_setup_ht_reset(struct ipath_devdata *dd)
 {
-	ipath_dbg("No reset possible for HT-400\n");
+	ipath_dbg("No reset possible for this InfiniPath hardware\n");
 	return 0;
 }
 
@@ -925,7 +924,7 @@ static int set_int_handler(struct ipath_
 
 	/*
 	 * kernels with CONFIG_PCI_MSI set the vector in the irq field of
-	 * struct pci_device, so we use that to program the HT-400 internal
+	 * struct pci_device, so we use that to program the internal
 	 * interrupt register (not config space) with that value. The BIOS
 	 * must still have done the basic MSI setup.
 	 */
@@ -1013,7 +1012,7 @@ bail:
  * @dd: the infinipath device
  *
  * Called during driver unload.
- * This is currently a nop for the HT-400, not for all chips
+ * This is currently a nop for the HT chip, not for all chips
  */
 static void ipath_setup_ht_cleanup(struct ipath_devdata *dd)
 {
@@ -1470,7 +1469,7 @@ static int ipath_ht_early_init(struct ip
 	dd->ipath_rcvhdrsize = IPATH_DFLT_RCVHDRSIZE;
 
 	/*
-	 * For HT-400, we allocate a somewhat overly large eager buffer,
+	 * For HT, we allocate a somewhat overly large eager buffer,
 	 * such that we can guarantee that we can receive the largest
 	 * packet that we can send out.  To truly support a 4KB MTU,
 	 * we need to bump this to a large value.  To date, other than
@@ -1531,7 +1530,7 @@ static int ipath_ht_early_init(struct ip
 	if(dd->ipath_boardrev == 5 && dd->ipath_serial[0] == '1' &&
 		dd->ipath_serial[1] == '2' && dd->ipath_serial[2] == '8') {
 		/*
-		 * Later production HT-460 has same changes as HT-465, so
+		 * Later production QHT7040 has same changes as QHT7140, so
 		 * can use GPIO interrupts.  They have serial #'s starting
 		 * with 128, rather than 112.
 		 */
@@ -1560,13 +1559,13 @@ static int ipath_ht_get_base_info(struct
 }
 
 /**
- * ipath_init_ht400_funcs - set up the chip-specific function pointers
+ * ipath_init_iba6110_funcs - set up the chip-specific function pointers
  * @dd: the infinipath device
  *
  * This is global, and is called directly at init to set up the
  * chip-specific function pointers for later use.
  */
-void ipath_init_ht400_funcs(struct ipath_devdata *dd)
+void ipath_init_iba6110_funcs(struct ipath_devdata *dd)
 {
 	dd->ipath_f_intrsetup = ipath_ht_intconfig;
 	dd->ipath_f_bus = ipath_setup_ht_config;
diff --git a/drivers/infiniband/hw/ipath/ipath_pe800.c b/drivers/infiniband/hw/ipath/ipath_iba6120.c
rename from drivers/infiniband/hw/ipath/ipath_pe800.c
rename to drivers/infiniband/hw/ipath/ipath_iba6120.c
--- a/drivers/infiniband/hw/ipath/ipath_iba6120.c	Fri Aug 25 11:19:45 2006 -0700
+++ b/drivers/infiniband/hw/ipath/ipath_iba6120.c	Fri Aug 25 11:19:45 2006 -0700
@@ -32,7 +32,7 @@
  */
 /*
  * This file contains all of the code that is specific to the
- * InfiniPath PE-800 chip.
+ * InfiniPath PCIe chip.
  */
 
 #include <linux/interrupt.h>
@@ -45,9 +45,9 @@
 
 /*
  * This file contains all the chip-specific register information and
- * access functions for the QLogic InfiniPath PE800, the PCI-Express chip.
- *
- * This lists the InfiniPath PE800 registers, in the actual chip layout.
+ * access functions for the QLogic InfiniPath PCI-Express chip.
+ *
+ * This lists the InfiniPath registers, in the actual chip layout.
  * This structure should never be directly accessed.
  */
 struct _infinipath_do_not_use_kernel_regs {
@@ -213,7 +213,6 @@ static const struct ipath_kregs ipath_pe
 	.kr_rcvhdraddr = IPATH_KREG_OFFSET(RcvHdrAddr0),
 	.kr_rcvhdrtailaddr = IPATH_KREG_OFFSET(RcvHdrTailAddr0),
 
-	/* This group is pe-800-specific; and used only in this file */
 	/* The rcvpktled register controls one of the debug port signals, so
 	 * a packet activity LED can be connected to it. */
 	.kr_rcvpktledcnt = IPATH_KREG_OFFSET(RcvPktLEDCnt),
@@ -388,7 +387,7 @@ static void ipath_pe_handle_hwerrors(str
 	*msg = '\0';
 
 	if (hwerrs & INFINIPATH_HWE_MEMBISTFAILED) {
-		strlcat(msg, "[Memory BIST test failed, PE-800 unusable]",
+		strlcat(msg, "[Memory BIST test failed, InfiniPath hardware unusable]",
 			msgl);
 		/* ignore from now on, so disable until driver reloaded */
 		*dd->ipath_statusp |= IPATH_STATUS_HWERROR;
@@ -433,7 +432,7 @@ static void ipath_pe_handle_hwerrors(str
 
 	if (hwerrs & _IPATH_PLL_FAIL) {
 		snprintf(bitsmsg, sizeof bitsmsg,
-			 "[PLL failed (%llx), PE-800 unusable]",
+			 "[PLL failed (%llx), InfiniPath hardware unusable]",
 			 (unsigned long long) hwerrs & _IPATH_PLL_FAIL);
 		strlcat(msg, bitsmsg, msgl);
 		/* ignore from now on, so disable until driver reloaded */
@@ -511,22 +510,25 @@ static int ipath_pe_boardname(struct ipa
 		n = "InfiniPath_Emulation";
 		break;
 	case 1:
-		n = "InfiniPath_PE-800-Bringup";
+		n = "InfiniPath_QLE7140-Bringup";
 		break;
 	case 2:
-		n = "InfiniPath_PE-880";
+		n = "InfiniPath_QLE7140";
 		break;
 	case 3:
-		n = "InfiniPath_PE-850";
+		n = "InfiniPath_QMI7140";
 		break;
 	case 4:
-		n = "InfiniPath_PE-860";
+		n = "InfiniPath_QEM7140";
+		break;
+	case 5:
+		n = "InfiniPath_QMH7140";
 		break;
 	default:
 		ipath_dev_err(dd,
 			      "Don't yet know about board with ID %u\n",
 			      boardrev);
-		snprintf(name, namelen, "Unknown_InfiniPath_PE-8xx_%u",
+		snprintf(name, namelen, "Unknown_InfiniPath_PCIe_%u",
 			 boardrev);
 		break;
 	}
@@ -534,7 +536,7 @@ static int ipath_pe_boardname(struct ipa
 		snprintf(name, namelen, "%s", n);
 
 	if (dd->ipath_majrev != 4 || !dd->ipath_minrev || dd->ipath_minrev>2) {
-		ipath_dev_err(dd, "Unsupported PE-800 revision %u.%u!\n",
+		ipath_dev_err(dd, "Unsupported InfiniPath hardware revision %u.%u!\n",
 			      dd->ipath_majrev, dd->ipath_minrev);
 		ret = 1;
 	} else
@@ -705,7 +707,7 @@ static void ipath_pe_quiet_serdes(struct
 	ipath_write_kreg(dd, dd->ipath_kregs->kr_serdesconfig0, val);
 }
 
-/* this is not yet needed on the PE800, so just return 0. */
+/* this is not yet needed on this chip, so just return 0. */
 static int ipath_pe_intconfig(struct ipath_devdata *dd)
 {
 	return 0;
@@ -759,8 +761,8 @@ static void ipath_setup_pe_setextled(str
  *
  * This is called during driver unload.
  * We do the pci_disable_msi here, not in generic code, because it
- * isn't used for the HT-400. If we do end up needing pci_enable_msi
- * at some point in the future for HT-400, we'll move the call back
+ * isn't used for the HT chips. If we do end up needing pci_enable_msi
+ * at some point in the future for HT, we'll move the call back
  * into the main init_one code.
  */
 static void ipath_setup_pe_cleanup(struct ipath_devdata *dd)
@@ -780,10 +782,10 @@ static void ipath_setup_pe_cleanup(struc
  * late in 2.6.16).
  * All that can be done is to edit the kernel source to remove the quirk
  * check until that is fixed.
- * We do not need to call enable_msi() for our HyperTransport chip (HT-400),
- * even those it uses MSI, and we want to avoid the quirk warning, so
- * So we call enable_msi only for the PE-800.  If we do end up needing
- * pci_enable_msi at some point in the future for HT-400, we'll move the
+ * We do not need to call enable_msi() for our HyperTransport chip,
+ * even though it uses MSI, and we want to avoid the quirk warning, so
+ * So we call enable_msi only for PCIe.  If we do end up needing
+ * pci_enable_msi at some point in the future for HT, we'll move the
  * call back into the main init_one code.
  * We save the msi lo and hi values, so we can restore them after
  * chip reset (the kernel PCI infrastructure doesn't yet handle that
@@ -971,8 +973,7 @@ static int ipath_setup_pe_reset(struct i
 	int ret;
 
 	/* Use ERROR so it shows up in logs, etc. */
-	ipath_dev_err(dd, "Resetting PE-800 unit %u\n",
-		      dd->ipath_unit);
+	ipath_dev_err(dd, "Resetting InfiniPath unit %u\n", dd->ipath_unit);
 	/* keep chip from being accessed in a few places */
 	dd->ipath_flags &= ~(IPATH_INITTED|IPATH_PRESENT);
 	val = dd->ipath_control | INFINIPATH_C_RESET;
@@ -1078,7 +1079,7 @@ static void ipath_pe_put_tid(struct ipat
  * @port: the port
  *
  * clear all TID entries for a port, expected and eager.
- * Used from ipath_close().  On PE800, TIDs are only 32 bits,
+ * Used from ipath_close().  On this chip, TIDs are only 32 bits,
  * not 64, but they are still on 64 bit boundaries, so tidbase
  * is declared as u64 * for the pointer math, even though we write 32 bits
  */
@@ -1148,9 +1149,9 @@ static int ipath_pe_early_init(struct ip
 	dd->ipath_flags |= IPATH_4BYTE_TID;
 
 	/*
-	 * For openib, we need to be able to handle an IB header of 96 bytes
-	 * or 24 dwords.  HT-400 has arbitrary sized receive buffers, so we
-	 * made them the same size as the PIO buffers.  The PE-800 does not
+	 * For openfabrics, we need to be able to handle an IB header of
+	 * 24 dwords.  HT chip has arbitrary sized receive buffers, so we
+	 * made them the same size as the PIO buffers.  This chip does not
 	 * handle arbitrary size buffers, so we need the header large enough
 	 * to handle largest IB header, but still have room for a 2KB MTU
 	 * standard IB packet.
@@ -1158,11 +1159,10 @@ static int ipath_pe_early_init(struct ip
 	dd->ipath_rcvhdrentsize = 24;
 	dd->ipath_rcvhdrsize = IPATH_DFLT_RCVHDRSIZE;
 
-	/* For HT-400, we allocate a somewhat overly large eager buffer,
-	 * such that we can guarantee that we can receive the largest packet
-	 * that we can send out.  To truly support a 4KB MTU, we need to
-	 * bump this to a larger value.  We'll do this when I get around to
-	 * testing 4KB sends on the PE-800, which I have not yet done.
+	/*
+	 * To truly support a 4KB MTU (for usermode), we need to
+	 * bump this to a larger value.  For now, we use them for
+	 * the kernel only.
 	 */
 	dd->ipath_rcvegrbufsize = 2048;
 	/*
@@ -1175,9 +1175,9 @@ static int ipath_pe_early_init(struct ip
 	dd->ipath_init_ibmaxlen = dd->ipath_ibmaxlen;
 
 	/*
-	 * For PE-800, we can request a receive interrupt for 1 or
+	 * We can request a receive interrupt for 1 or
 	 * more packets from current offset.  For now, we set this
-	 * up for a single packet, to match the HT-400 behavior.
+	 * up for a single packet.
 	 */
 	dd->ipath_rhdrhead_intr_off = 1ULL<<32;
 
@@ -1216,13 +1216,13 @@ static int ipath_pe_get_base_info(struct
 }
 
 /**
- * ipath_init_pe800_funcs - set up the chip-specific function pointers
+ * ipath_init_iba6120_funcs - set up the chip-specific function pointers
  * @dd: the infinipath device
  *
  * This is global, and is called directly at init to set up the
  * chip-specific function pointers for later use.
  */
-void ipath_init_pe800_funcs(struct ipath_devdata *dd)
+void ipath_init_iba6120_funcs(struct ipath_devdata *dd)
 {
 	dd->ipath_f_intrsetup = ipath_pe_intconfig;
 	dd->ipath_f_bus = ipath_setup_pe_config;
diff --git a/drivers/infiniband/hw/ipath/ipath_kernel.h b/drivers/infiniband/hw/ipath/ipath_kernel.h
--- a/drivers/infiniband/hw/ipath/ipath_kernel.h	Fri Aug 25 11:19:45 2006 -0700
+++ b/drivers/infiniband/hw/ipath/ipath_kernel.h	Fri Aug 25 11:19:45 2006 -0700
@@ -236,7 +236,7 @@ struct ipath_devdata {
 	u64 ipath_tidtemplate;
 	/* value to write to free TIDs */
 	u64 ipath_tidinvalid;
-	/* PE-800 rcv interrupt setup */
+	/* IBA6120 rcv interrupt setup */
 	u64 ipath_rhdrhead_intr_off;
 
 	/* size of memory at ipath_kregbase */
@@ -621,10 +621,8 @@ int ipath_waitfor_mdio_cmdready(struct i
 int ipath_waitfor_mdio_cmdready(struct ipath_devdata *);
 int ipath_waitfor_complete(struct ipath_devdata *, ipath_kreg, u64, u64 *);
 u32 __iomem *ipath_getpiobuf(struct ipath_devdata *, u32 *);
-/* init PE-800-specific func */
-void ipath_init_pe800_funcs(struct ipath_devdata *);
-/* init HT-400-specific func */
-void ipath_init_ht400_funcs(struct ipath_devdata *);
+void ipath_init_iba6120_funcs(struct ipath_devdata *);
+void ipath_init_iba6110_funcs(struct ipath_devdata *);
 void ipath_get_eeprom_info(struct ipath_devdata *);
 u64 ipath_snap_cntr(struct ipath_devdata *, ipath_creg);
 
diff --git a/drivers/infiniband/hw/ipath/ipath_registers.h b/drivers/infiniband/hw/ipath/ipath_registers.h
--- a/drivers/infiniband/hw/ipath/ipath_registers.h	Fri Aug 25 11:19:45 2006 -0700
+++ b/drivers/infiniband/hw/ipath/ipath_registers.h	Fri Aug 25 11:19:45 2006 -0700
@@ -36,8 +36,7 @@
 
 /*
  * This file should only be included by kernel source, and by the diags.  It
- * defines the registers, and their contents, for the InfiniPath HT-400
- * chip.
+ * defines the registers, and their contents, for InfiniPath chips.
  */
 
 /*
@@ -286,7 +285,7 @@
 
 #define INFINIPATH_RT_ADDR_MASK 0xFFFFFFFFFFULL	/* 40 bits valid */
 
-/* TID entries (memory), HT400-only */
+/* TID entries (memory), HT-only */
 #define INFINIPATH_RT_VALID 0x8000000000000000ULL
 #define INFINIPATH_RT_ADDR_SHIFT 0
 #define INFINIPATH_RT_BUFSIZE_MASK 0x3FFF

[PATCH 15 of 23] IB/ipath - add serial number to hardware freeze error message

[Bryan O'Sullivan]

Also added the word "Hardware" after "Fatal" to make it more obvious
that it's hardware, not software.

Signed-off-by: Bryan O'Sullivan <bryan.osullivan@qlogic.com>

diff --git a/drivers/infiniband/hw/ipath/ipath_iba6110.c b/drivers/infiniband/hw/ipath/ipath_iba6110.c
--- a/drivers/infiniband/hw/ipath/ipath_iba6110.c	Fri Aug 25 11:19:45 2006 -0700
+++ b/drivers/infiniband/hw/ipath/ipath_iba6110.c	Fri Aug 25 11:19:45 2006 -0700
@@ -461,8 +461,9 @@ static void ipath_ht_handle_hwerrors(str
 			 * times.
 			 */
 			if (dd->ipath_flags & IPATH_INITTED) {
-				ipath_dev_err(dd, "Fatal Error (freeze "
-					      "mode), no longer usable\n");
+				ipath_dev_err(dd, "Fatal Hardware Error (freeze "
+					      "mode), no longer usable, SN %.16s\n",
+						  dd->ipath_serial);
 				isfatal = 1;
 			}
 			*dd->ipath_statusp &= ~IPATH_STATUS_IB_READY;
diff --git a/drivers/infiniband/hw/ipath/ipath_iba6120.c b/drivers/infiniband/hw/ipath/ipath_iba6120.c
--- a/drivers/infiniband/hw/ipath/ipath_iba6120.c	Fri Aug 25 11:19:45 2006 -0700
+++ b/drivers/infiniband/hw/ipath/ipath_iba6120.c	Fri Aug 25 11:19:45 2006 -0700
@@ -363,8 +363,9 @@ static void ipath_pe_handle_hwerrors(str
 			 * and we get here multiple times
 			 */
 			if (dd->ipath_flags & IPATH_INITTED) {
-				ipath_dev_err(dd, "Fatal Error (freeze "
-					      "mode), no longer usable\n");
+				ipath_dev_err(dd, "Fatal Hardware Error (freeze "
+					      "mode), no longer usable, SN %.16s\n",
+						  dd->ipath_serial);
 				isfatal = 1;
 			}
 			/*

[PATCH 16 of 23] IB/ipath - be more strict about testing the modify QP verb

[Bryan O'Sullivan]

Signed-off-by: Bryan O'Sullivan <bryan.osullivan@qlogic.com>

diff --git a/drivers/infiniband/hw/ipath/ipath_qp.c b/drivers/infiniband/hw/ipath/ipath_qp.c
--- a/drivers/infiniband/hw/ipath/ipath_qp.c	Fri Aug 25 11:19:45 2006 -0700
+++ b/drivers/infiniband/hw/ipath/ipath_qp.c	Fri Aug 25 11:19:45 2006 -0700
@@ -455,11 +455,16 @@ int ipath_modify_qp(struct ib_qp *ibqp, 
 				attr_mask))
 		goto inval;
 
-	if (attr_mask & IB_QP_AV)
+	if (attr_mask & IB_QP_AV) {
 		if (attr->ah_attr.dlid == 0 ||
 		    attr->ah_attr.dlid >= IPATH_MULTICAST_LID_BASE)
 			goto inval;
 
+		if ((attr->ah_attr.ah_flags & IB_AH_GRH) &&
+		    (attr->ah_attr.grh.sgid_index > 1))
+			goto inval;
+	}
+
 	if (attr_mask & IB_QP_PKEY_INDEX)
 		if (attr->pkey_index >= ipath_get_npkeys(dev->dd))
 			goto inval;
@@ -468,6 +473,27 @@ int ipath_modify_qp(struct ib_qp *ibqp, 
 		if (attr->min_rnr_timer > 31)
 			goto inval;
 
+	if (attr_mask & IB_QP_PORT)
+		if (attr->port_num == 0 ||
+		    attr->port_num > ibqp->device->phys_port_cnt)
+			goto inval;
+
+	if (attr_mask & IB_QP_PATH_MTU)
+		if (attr->path_mtu > IB_MTU_4096)
+			goto inval;
+
+	if (attr_mask & IB_QP_MAX_DEST_RD_ATOMIC)
+		if (attr->max_dest_rd_atomic > 1)
+			goto inval;
+ 
+	if (attr_mask & IB_QP_MAX_QP_RD_ATOMIC)
+		if (attr->max_rd_atomic > 1)
+			goto inval;
+ 
+	if (attr_mask & IB_QP_PATH_MIG_STATE)
+		if (attr->path_mig_state != IB_MIG_MIGRATED)
+			goto inval;
+ 
 	switch (new_state) {
 	case IB_QPS_RESET:
 		ipath_reset_qp(qp);
@@ -517,6 +543,9 @@ int ipath_modify_qp(struct ib_qp *ibqp, 
 
 	if (attr_mask & IB_QP_MIN_RNR_TIMER)
 		qp->r_min_rnr_timer = attr->min_rnr_timer;
+
+	if (attr_mask & IB_QP_TIMEOUT)
+		qp->timeout = attr->timeout;
 
 	if (attr_mask & IB_QP_QKEY)
 		qp->qkey = attr->qkey;
@@ -564,7 +593,7 @@ int ipath_query_qp(struct ib_qp *ibqp, s
 	attr->max_dest_rd_atomic = 1;
 	attr->min_rnr_timer = qp->r_min_rnr_timer;
 	attr->port_num = 1;
-	attr->timeout = 0;
+	attr->timeout = qp->timeout;
 	attr->retry_cnt = qp->s_retry_cnt;
 	attr->rnr_retry = qp->s_rnr_retry;
 	attr->alt_port_num = 0;
diff --git a/drivers/infiniband/hw/ipath/ipath_verbs.h b/drivers/infiniband/hw/ipath/ipath_verbs.h
--- a/drivers/infiniband/hw/ipath/ipath_verbs.h	Fri Aug 25 11:19:45 2006 -0700
+++ b/drivers/infiniband/hw/ipath/ipath_verbs.h	Fri Aug 25 11:19:45 2006 -0700
@@ -371,6 +371,7 @@ struct ipath_qp {
 	u8 s_retry;		/* requester retry counter */
 	u8 s_rnr_retry;		/* requester RNR retry counter */
 	u8 s_pkey_index;	/* PKEY index to use */
+	u8 timeout;		/* Timeout for this QP */
 	enum ib_mtu path_mtu;
 	u32 remote_qpn;
 	u32 qkey;		/* QKEY for this QP (for UD or RD) */

[PATCH 17 of 23] IB/ipath - validate path_mig_state properly

[Bryan O'Sullivan]

Signed-off-by: Bryan O'Sullivan <bryan.osullivan@qlogic.com>

diff --git a/drivers/infiniband/hw/ipath/ipath_qp.c b/drivers/infiniband/hw/ipath/ipath_qp.c
--- a/drivers/infiniband/hw/ipath/ipath_qp.c	Fri Aug 25 11:19:45 2006 -0700
+++ b/drivers/infiniband/hw/ipath/ipath_qp.c	Fri Aug 25 11:19:45 2006 -0700
@@ -491,7 +491,8 @@ int ipath_modify_qp(struct ib_qp *ibqp, 
 			goto inval;
  
 	if (attr_mask & IB_QP_PATH_MIG_STATE)
-		if (attr->path_mig_state != IB_MIG_MIGRATED)
+		if (attr->path_mig_state != IB_MIG_MIGRATED &&
+		    attr->path_mig_state != IB_MIG_REARM)
 			goto inval;
  
 	switch (new_state) {

[PATCH 18 of 23] IB/ipath - put a limit on the number of QPs that can be created

[Bryan O'Sullivan]

Signed-off-by: Bryan O'Sullivan <bryan.osullivan@qlogic.com>

diff --git a/drivers/infiniband/hw/ipath/ipath_qp.c b/drivers/infiniband/hw/ipath/ipath_qp.c
--- a/drivers/infiniband/hw/ipath/ipath_qp.c	Fri Aug 25 11:19:45 2006 -0700
+++ b/drivers/infiniband/hw/ipath/ipath_qp.c	Fri Aug 25 11:19:45 2006 -0700
@@ -833,9 +833,21 @@ struct ib_qp *ipath_create_qp(struct ib_
 		}
 	}
 
+	spin_lock(&dev->n_qps_lock);
+	if (dev->n_qps_allocated == ib_ipath_max_qps) {
+		spin_unlock(&dev->n_qps_lock);
+		ret = ERR_PTR(-ENOMEM);
+		goto bail_ip;
+	}
+
+	dev->n_qps_allocated++;
+	spin_unlock(&dev->n_qps_lock);
+
 	ret = &qp->ibqp;
 	goto bail;
 
+bail_ip:
+	kfree(qp->ip);
 bail_rwq:
 	vfree(qp->r_rq.wq);
 bail_qp:
@@ -864,6 +876,9 @@ int ipath_destroy_qp(struct ib_qp *ibqp)
 	spin_lock_irqsave(&qp->s_lock, flags);
 	qp->state = IB_QPS_ERR;
 	spin_unlock_irqrestore(&qp->s_lock, flags);
+	spin_lock(&dev->n_qps_lock);
+	dev->n_qps_allocated--;
+	spin_unlock(&dev->n_qps_lock);
 
 	/* Stop the sending tasklet. */
 	tasklet_kill(&qp->s_task);
diff --git a/drivers/infiniband/hw/ipath/ipath_verbs.c b/drivers/infiniband/hw/ipath/ipath_verbs.c
--- a/drivers/infiniband/hw/ipath/ipath_verbs.c	Fri Aug 25 11:19:45 2006 -0700
+++ b/drivers/infiniband/hw/ipath/ipath_verbs.c	Fri Aug 25 11:19:45 2006 -0700
@@ -72,6 +72,10 @@ module_param_named(max_qp_wrs, ib_ipath_
 module_param_named(max_qp_wrs, ib_ipath_max_qp_wrs, uint,
 		   S_IWUSR | S_IRUGO);
 MODULE_PARM_DESC(max_qp_wrs, "Maximum number of QP WRs to support");
+
+unsigned int ib_ipath_max_qps = 16384;
+module_param_named(max_qps, ib_ipath_max_qps, uint, S_IWUSR | S_IRUGO);
+MODULE_PARM_DESC(max_qps, "Maximum number of QPs to support");
 
 unsigned int ib_ipath_max_sges = 0x60;
 module_param_named(max_sges, ib_ipath_max_sges, uint, S_IWUSR | S_IRUGO);
@@ -958,7 +962,7 @@ static int ipath_query_device(struct ib_
 	props->sys_image_guid = dev->sys_image_guid;
 
 	props->max_mr_size = ~0ull;
-	props->max_qp = dev->qp_table.max;
+	props->max_qp = ib_ipath_max_qps;
 	props->max_qp_wr = ib_ipath_max_qp_wrs;
 	props->max_sge = ib_ipath_max_sges;
 	props->max_cq = ib_ipath_max_cqs;
@@ -1420,6 +1424,7 @@ int ipath_register_ib_device(struct ipat
 	spin_lock_init(&idev->n_pds_lock);
 	spin_lock_init(&idev->n_ahs_lock);
 	spin_lock_init(&idev->n_cqs_lock);
+	spin_lock_init(&idev->n_qps_lock);
 	spin_lock_init(&idev->n_srqs_lock);
 	spin_lock_init(&idev->n_mcast_grps_lock);
 
diff --git a/drivers/infiniband/hw/ipath/ipath_verbs.h b/drivers/infiniband/hw/ipath/ipath_verbs.h
--- a/drivers/infiniband/hw/ipath/ipath_verbs.h	Fri Aug 25 11:19:45 2006 -0700
+++ b/drivers/infiniband/hw/ipath/ipath_verbs.h	Fri Aug 25 11:19:45 2006 -0700
@@ -482,6 +482,8 @@ struct ipath_ibdev {
 	spinlock_t n_ahs_lock;
 	u32 n_cqs_allocated;	/* number of CQs allocated for device */
 	spinlock_t n_cqs_lock;
+	u32 n_qps_allocated;	/* number of QPs allocated for device */
+	spinlock_t n_qps_lock;
 	u32 n_srqs_allocated;	/* number of SRQs allocated for device */
 	spinlock_t n_srqs_lock;
 	u32 n_mcast_grps_allocated; /* number of mcast groups allocated */
@@ -792,6 +794,8 @@ extern unsigned int ib_ipath_max_cqs;
 
 extern unsigned int ib_ipath_max_qp_wrs;
 
+extern unsigned int ib_ipath_max_qps;
+
 extern unsigned int ib_ipath_max_sges;
 
 extern unsigned int ib_ipath_max_mcast_grps;

[PATCH 19 of 23] IB/ipath - handle sq_sig_all field correctly

[Bryan O'Sullivan]

Signed-off-by: Bryan O'Sullivan <bryan.osullivan@qlogic.com>

diff --git a/drivers/infiniband/hw/ipath/ipath_qp.c b/drivers/infiniband/hw/ipath/ipath_qp.c
--- a/drivers/infiniband/hw/ipath/ipath_qp.c	Fri Aug 25 11:19:45 2006 -0700
+++ b/drivers/infiniband/hw/ipath/ipath_qp.c	Fri Aug 25 11:19:45 2006 -0700
@@ -606,9 +606,10 @@ int ipath_query_qp(struct ib_qp *ibqp, s
 	init_attr->recv_cq = qp->ibqp.recv_cq;
 	init_attr->srq = qp->ibqp.srq;
 	init_attr->cap = attr->cap;
-	init_attr->sq_sig_type =
-		(qp->s_flags & (1 << IPATH_S_SIGNAL_REQ_WR))
-		? IB_SIGNAL_REQ_WR : 0;
+	if (qp->s_flags & (1 << IPATH_S_SIGNAL_REQ_WR))
+		init_attr->sq_sig_type = IB_SIGNAL_REQ_WR;
+	else
+		init_attr->sq_sig_type = IB_SIGNAL_ALL_WR;
 	init_attr->qp_type = qp->ibqp.qp_type;
 	init_attr->port_num = 1;
 	return 0;
@@ -776,8 +777,10 @@ struct ib_qp *ipath_create_qp(struct ib_
 		qp->s_wq = swq;
 		qp->s_size = init_attr->cap.max_send_wr + 1;
 		qp->s_max_sge = init_attr->cap.max_send_sge;
-		qp->s_flags = init_attr->sq_sig_type == IB_SIGNAL_REQ_WR ?
-			1 << IPATH_S_SIGNAL_REQ_WR : 0;
+		if (init_attr->sq_sig_type == IB_SIGNAL_REQ_WR)
+			qp->s_flags = 1 << IPATH_S_SIGNAL_REQ_WR;
+		else
+			qp->s_flags = 0;
 		dev = to_idev(ibpd->device);
 		err = ipath_alloc_qpn(&dev->qp_table, qp,
 				      init_attr->qp_type);

[PATCH 20 of 23] IB/ipath - allow SMA to be disabled

[Bryan O'Sullivan]

This is useful for testing purposes.

Signed-off-by: Bryan O'Sullivan <bryan.osullivan@qlogic.com>

diff --git a/drivers/infiniband/hw/ipath/ipath_verbs.c b/drivers/infiniband/hw/ipath/ipath_verbs.c
--- a/drivers/infiniband/hw/ipath/ipath_verbs.c	Fri Aug 25 11:19:45 2006 -0700
+++ b/drivers/infiniband/hw/ipath/ipath_verbs.c	Fri Aug 25 11:19:45 2006 -0700
@@ -107,6 +107,10 @@ module_param_named(max_srq_wrs, ib_ipath
 		   uint, S_IWUSR | S_IRUGO);
 MODULE_PARM_DESC(max_srq_wrs, "Maximum number of SRQ WRs support");
 
+static unsigned int ib_ipath_disable_sma;
+module_param_named(disable_sma, ib_ipath_disable_sma, uint, S_IWUSR | S_IRUGO);
+MODULE_PARM_DESC(ib_ipath_disable_sma, "Disable the SMA");
+
 const int ib_ipath_state_ops[IB_QPS_ERR + 1] = {
 	[IB_QPS_RESET] = 0,
 	[IB_QPS_INIT] = IPATH_POST_RECV_OK,
@@ -354,6 +358,9 @@ static void ipath_qp_rcv(struct ipath_ib
 	switch (qp->ibqp.qp_type) {
 	case IB_QPT_SMI:
 	case IB_QPT_GSI:
+		if (ib_ipath_disable_sma)
+			break;
+		/* FALLTHROUGH */
 	case IB_QPT_UD:
 		ipath_ud_rcv(dev, hdr, has_grh, data, tlen, qp);
 		break;

[PATCH 21 of 23] IB/ipath - fix return value from ipath_poll

[Bryan O'Sullivan]

This stops the generic poll code from waiting for a timeout.

Signed-off-by: Bryan O'Sullivan <bryan.osullivan@qlogic.com>

diff --git a/drivers/infiniband/hw/ipath/ipath_file_ops.c b/drivers/infiniband/hw/ipath/ipath_file_ops.c
--- a/drivers/infiniband/hw/ipath/ipath_file_ops.c	Fri Aug 25 11:19:45 2006 -0700
+++ b/drivers/infiniband/hw/ipath/ipath_file_ops.c	Fri Aug 25 11:19:45 2006 -0700
@@ -1150,6 +1150,7 @@ static unsigned int ipath_poll(struct fi
 	struct ipath_portdata *pd;
 	u32 head, tail;
 	int bit;
+	unsigned pollflag = 0;
 	struct ipath_devdata *dd;
 
 	pd = port_fp(fp);
@@ -1186,9 +1187,12 @@ static unsigned int ipath_poll(struct fi
 			clear_bit(IPATH_PORT_WAITING_RCV, &pd->port_flag);
 			pd->port_rcvwait_to++;
 		}
+		else
+			pollflag = POLLIN | POLLRDNORM;
 	}
 	else {
 		/* it's already happened; don't do wait_event overhead */
+		pollflag = POLLIN | POLLRDNORM;
 		pd->port_rcvnowait++;
 	}
 
@@ -1196,7 +1200,7 @@ static unsigned int ipath_poll(struct fi
 	ipath_write_kreg(dd, dd->ipath_kregs->kr_rcvctrl,
 			 dd->ipath_rcvctrl);
 
-	return 0;
+	return pollflag;
 }
 
 static int try_alloc_port(struct ipath_devdata *dd, int port,

[PATCH 22 of 23] IB/ipath - print warning if LID not acquired within one minute

[Bryan O'Sullivan]

Signed-off-by: Bryan O'Sullivan <bryan.osullivan@qlogic.com>

diff --git a/drivers/infiniband/hw/ipath/ipath_driver.c b/drivers/infiniband/hw/ipath/ipath_driver.c
--- a/drivers/infiniband/hw/ipath/ipath_driver.c	Fri Aug 25 11:19:45 2006 -0700
+++ b/drivers/infiniband/hw/ipath/ipath_driver.c	Fri Aug 25 11:19:46 2006 -0700
@@ -114,6 +114,13 @@ static int __devinit ipath_init_one(stru
 #define PCI_DEVICE_ID_INFINIPATH_HT 0xd
 #define PCI_DEVICE_ID_INFINIPATH_PE800 0x10
 
+/*
+ * Number of seconds before we complain about not getting a LID
+ * assignment.
+ */
+
+#define LID_TIMEOUT 60
+
 static const struct pci_device_id ipath_pci_tbl[] = {
 	{ PCI_DEVICE(PCI_VENDOR_ID_PATHSCALE, PCI_DEVICE_ID_INFINIPATH_HT) },
 	{ PCI_DEVICE(PCI_VENDOR_ID_PATHSCALE, PCI_DEVICE_ID_INFINIPATH_PE800) },
@@ -129,6 +136,29 @@ static struct pci_driver ipath_driver = 
 	.id_table = ipath_pci_tbl,
 };
 
+
+static void check_link_status(void *data)
+{
+	struct ipath_devdata *dd = data;
+
+	/*
+	 * If we're in the NOCABLE state, try again in another minute.
+	 */
+
+	if (dd->ipath_flags & IPATH_STATUS_IB_NOCABLE) {
+		schedule_delayed_work(&dd->link_task, HZ * LID_TIMEOUT);
+		return;
+	}
+
+	/*
+	 * If we don't have a LID, let the user know and don't bother
+	 * checking again.
+	 */
+
+	if (dd->ipath_lid == 0)
+		dev_info(&dd->pcidev->dev,
+			 "We don't have a LID yet (no subnet manager?)");
+}
 
 static inline void read_bars(struct ipath_devdata *dd, struct pci_dev *dev,
 			     u32 *bar0, u32 *bar1)
@@ -196,6 +226,8 @@ static struct ipath_devdata *ipath_alloc
 
 	dd->pcidev = pdev;
 	pci_set_drvdata(pdev, dd);
+
+	INIT_WORK(&dd->link_task, check_link_status, dd);
 
 	list_add(&dd->ipath_list, &ipath_dev_list);
 
@@ -509,6 +541,9 @@ static int __devinit ipath_init_one(stru
 	ipath_diag_add(dd);
 	ipath_register_ib_device(dd);
 
+	/* Check that we have a LID in LID_TIMEOUT seconds. */
+	schedule_delayed_work(&dd->link_task, HZ * LID_TIMEOUT);
+
 	goto bail;
 
 bail_iounmap:
@@ -536,6 +571,9 @@ static void __devexit ipath_remove_one(s
 		return;
 
 	dd = pci_get_drvdata(pdev);
+
+	cancel_delayed_work(&dd->link_task);
+
 	ipath_unregister_ib_device(dd->verbs_dev);
 	ipath_diag_remove(dd);
 	ipath_user_remove(dd);
@@ -1644,6 +1682,8 @@ int ipath_set_lid(struct ipath_devdata *
 	dd->ipath_lid = arg;
 	dd->ipath_lmc = lmc;
 
+	dev_info(&dd->pcidev->dev, "We got a lid: %u\n", arg);
+
 	return 0;
 }
 
diff --git a/drivers/infiniband/hw/ipath/ipath_kernel.h b/drivers/infiniband/hw/ipath/ipath_kernel.h
--- a/drivers/infiniband/hw/ipath/ipath_kernel.h	Fri Aug 25 11:19:45 2006 -0700
+++ b/drivers/infiniband/hw/ipath/ipath_kernel.h	Fri Aug 25 11:19:46 2006 -0700
@@ -508,6 +508,9 @@ struct ipath_devdata {
 	u32 ipath_lli_counter;
 	/* local link integrity errors */
 	u32 ipath_lli_errors;
+
+	/* Link status check work */
+	struct work_struct link_task;
 };
 
 extern struct list_head ipath_dev_list;

Re: [PATCH 22 of 23] IB/ipath - print warning if LID not acquired within one minute

[Roland Dreier]

1) What makes ipath special so that we want this warning for ipath
devices but not other IB hardware?  If this warning is actually
useful, then I think it would make more sense to start a timer when
any IB device is added, and warn if ports with a physical link don't
become active after the timeout time.  But I'm having a hard time
seeing why we want this message in the kernel log.

2) You do cancel_delayed_work() but not flush_scheduled_work(), so
it's possible for your timeout function to be running after the module
text is gone.

 - R.

Re: [openib-general] [PATCH 22 of 23] IB/ipath - print warning if LID not acquired within one minute

[Robert Walsh]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Roland Dreier wrote:
> 1) What makes ipath special so that we want this warning for ipath
> devices but not other IB hardware?

There's nothing special about our hardware that requires this.  We just
wanted that in there so we could direct customers to look at dmesg to
see if the warning popped up if they call with a problem.  It is useful
to have for this purpose.

> If this warning is actually
> useful, then I think it would make more sense to start a timer when
> any IB device is added, and warn if ports with a physical link don't
> become active after the timeout time.

I'd be OK with doing that, too.

> But I'm having a hard time
> seeing why we want this message in the kernel log.

It's useful when you're trying to track down problems.

> 2) You do cancel_delayed_work() but not flush_scheduled_work(), so
> it's possible for your timeout function to be running after the module
> text is gone.

OK - I'll fix this up.  Thanks for spotting it.

Regards,
 Robert.

-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.5 (GNU/Linux)
Comment: Using GnuPG with Fedora - http://enigmail.mozdev.org

iQEVAwUBRO9gU/zvnpzTd9fxAQJBQwgAkbgrEA4/UpbcD0gsGC+39r5ZAAz+4d3I
/QAIKn239juMf8TfrlekAzD9MCj5Rna1bk3yu1gu/Z0Jg5vHvQNmBxDtveQ4bDyu
1DAUbvmclNknzM00LtMHI6AZfYbRgsbCIKXJw0reXctAkbJAvMU0U6Ff1imvO0Tw
38C24ktDalaaKpz4DHO261UHlmtD4wlJojKLYI5yH39JSHK449zjJznrP9W8SPIU
RbxGktSsD69gQXmpqgY5KEmbcukZ9AIF4VHTG2uEz1aO7eOQ+1BsUg140EcWXC//
R1Jg56WhCYsMDVik7+u994VgQi34beos9pwbLIUkq+315VHN3QFbQg==
=XhNA
-----END PGP SIGNATURE-----

Re: [PATCH 22 of 23] IB/ipath - print warning if LID not acquired within one minute

[Michael S. Tsirkin]

Quoting r. Robert Walsh <rjwalsh@pathscale.com>:
> Subject: Re: [PATCH 22 of 23] IB/ipath - print warning if LID not acquired within one minute
> 
> -----BEGIN PGP SIGNED MESSAGE-----
> Hash: SHA1
> 
> Roland Dreier wrote:
> > 1) What makes ipath special so that we want this warning for ipath
> > devices but not other IB hardware?
> 
> There's nothing special about our hardware that requires this.  We just
> wanted that in there so we could direct customers to look at dmesg to
> see if the warning popped up if they call with a problem.  It is useful
> to have for this purpose.
> 
> > If this warning is actually
> > useful, then I think it would make more sense to start a timer when
> > any IB device is added, and warn if ports with a physical link don't
> > become active after the timeout time.
> 
> I'd be OK with doing that, too.

Looks like your devices are all single-port. With a multi port
device it is quite common to have one port down.

> > But I'm having a hard time
> > seeing why we want this message in the kernel log.
> 
> It's useful when you're trying to track down problems.

How about doing this in userspace by looking at port state in sysfs?
You can diagnose a much wider class of problems this way.

-- 
MST

Re: [PATCH 22 of 23] IB/ipath - print warning if LID not acquired within one minute

[Roland Dreier]

    Michael> Looks like your devices are all single-port. With a multi
    Michael> port device it is quite common to have one port down.

My reading of the patch is that it warns if the link is up physically
but does not come up logically.  Which would still be reasonable for a
multi-port device.

But I am still wondering about when this is really useful.

 - R.

Re: [PATCH 22 of 23] IB/ipath - print warning if LID not acquired within one minute

[Robert Walsh]

Roland Dreier wrote:
>     Michael> Looks like your devices are all single-port. With a multi
>     Michael> port device it is quite common to have one port down.
> 
> My reading of the patch is that it warns if the link is up physically
> but does not come up logically.  Which would still be reasonable for a
> multi-port device.
> 
> But I am still wondering about when this is really useful.

Well, either you think it is or it isn't.  We like it: it's easier than 
pointing customers at something in /sys.

Regards,
  Robert.

[PATCH 23 of 23] IB/ipath - control receive polarity inversion

[Bryan O'Sullivan]

Signed-off-by: Bryan O'Sullivan <bryan.osullivan@qlogic.com>

diff --git a/drivers/infiniband/hw/ipath/ipath_driver.c b/drivers/infiniband/hw/ipath/ipath_driver.c
--- a/drivers/infiniband/hw/ipath/ipath_driver.c	Fri Aug 25 11:19:46 2006 -0700
+++ b/drivers/infiniband/hw/ipath/ipath_driver.c	Fri Aug 25 11:19:46 2006 -0700
@@ -2156,5 +2156,22 @@ bail:
 	return ret;
 }
 
+int ipath_set_rx_pol_inv(struct ipath_devdata *dd, u8 new_pol_inv)
+{
+	u64 val;
+	if ( new_pol_inv > INFINIPATH_XGXS_RX_POL_MASK ) {
+		return -1;
+	}
+	if ( dd->ipath_rx_pol_inv != new_pol_inv ) {
+		dd->ipath_rx_pol_inv = new_pol_inv;
+		val = ipath_read_kreg64(dd, dd->ipath_kregs->kr_xgxsconfig);
+		val &= ~(INFINIPATH_XGXS_RX_POL_MASK <<
+                         INFINIPATH_XGXS_RX_POL_SHIFT);
+                val |= ((u64)dd->ipath_rx_pol_inv) <<
+                        INFINIPATH_XGXS_RX_POL_SHIFT;
+		ipath_write_kreg(dd, dd->ipath_kregs->kr_xgxsconfig, val);
+	}
+	return 0;
+}
 module_init(infinipath_init);
 module_exit(infinipath_cleanup);
diff --git a/drivers/infiniband/hw/ipath/ipath_iba6110.c b/drivers/infiniband/hw/ipath/ipath_iba6110.c
--- a/drivers/infiniband/hw/ipath/ipath_iba6110.c	Fri Aug 25 11:19:46 2006 -0700
+++ b/drivers/infiniband/hw/ipath/ipath_iba6110.c	Fri Aug 25 11:19:46 2006 -0700
@@ -1290,6 +1290,15 @@ static int ipath_ht_bringup_serdes(struc
 		val &= ~INFINIPATH_XGXS_RESET;
 		change = 1;
 	}
+	if (((val >> INFINIPATH_XGXS_RX_POL_SHIFT) &
+	     INFINIPATH_XGXS_RX_POL_MASK) != dd->ipath_rx_pol_inv ) {
+		/* need to compensate for Tx inversion in partner */
+		val &= ~(INFINIPATH_XGXS_RX_POL_MASK <<
+		         INFINIPATH_XGXS_RX_POL_SHIFT);
+		val |= dd->ipath_rx_pol_inv <<
+			INFINIPATH_XGXS_RX_POL_SHIFT;
+		change = 1;
+	}
 	if (change)
 		ipath_write_kreg(dd, dd->ipath_kregs->kr_xgxsconfig, val);
 
diff --git a/drivers/infiniband/hw/ipath/ipath_iba6120.c b/drivers/infiniband/hw/ipath/ipath_iba6120.c
--- a/drivers/infiniband/hw/ipath/ipath_iba6120.c	Fri Aug 25 11:19:46 2006 -0700
+++ b/drivers/infiniband/hw/ipath/ipath_iba6120.c	Fri Aug 25 11:19:46 2006 -0700
@@ -654,6 +654,15 @@ static int ipath_pe_bringup_serdes(struc
 		val &= ~INFINIPATH_XGXS_RESET;
 		change = 1;
 	}
+	if (((val >> INFINIPATH_XGXS_RX_POL_SHIFT) &
+	     INFINIPATH_XGXS_RX_POL_MASK) != dd->ipath_rx_pol_inv ) {
+		/* need to compensate for Tx inversion in partner */
+		val &= ~(INFINIPATH_XGXS_RX_POL_MASK <<
+		         INFINIPATH_XGXS_RX_POL_SHIFT);
+		val |= dd->ipath_rx_pol_inv <<
+			INFINIPATH_XGXS_RX_POL_SHIFT;
+		change = 1;
+	}
 	if (change)
 		ipath_write_kreg(dd, dd->ipath_kregs->kr_xgxsconfig, val);
 
diff --git a/drivers/infiniband/hw/ipath/ipath_kernel.h b/drivers/infiniband/hw/ipath/ipath_kernel.h
--- a/drivers/infiniband/hw/ipath/ipath_kernel.h	Fri Aug 25 11:19:46 2006 -0700
+++ b/drivers/infiniband/hw/ipath/ipath_kernel.h	Fri Aug 25 11:19:46 2006 -0700
@@ -503,6 +503,8 @@ struct ipath_devdata {
 	u8 ipath_pci_cacheline;
 	/* LID mask control */
 	u8 ipath_lmc;
+	/* Rx Polarity inversion (compensate for ~tx on partner) */
+	u8 ipath_rx_pol_inv;
 
 	/* local link integrity counter */
 	u32 ipath_lli_counter;
@@ -570,6 +572,7 @@ int ipath_set_linkstate(struct ipath_dev
 int ipath_set_linkstate(struct ipath_devdata *, u8);
 int ipath_set_mtu(struct ipath_devdata *, u16);
 int ipath_set_lid(struct ipath_devdata *, u32, u8);
+int ipath_set_rx_pol_inv(struct ipath_devdata *dd, u8 new_pol_inv);
 
 /* for use in system calls, where we want to know device type, etc. */
 #define port_fp(fp) ((struct ipath_portdata *) (fp)->private_data)
diff --git a/drivers/infiniband/hw/ipath/ipath_registers.h b/drivers/infiniband/hw/ipath/ipath_registers.h
--- a/drivers/infiniband/hw/ipath/ipath_registers.h	Fri Aug 25 11:19:46 2006 -0700
+++ b/drivers/infiniband/hw/ipath/ipath_registers.h	Fri Aug 25 11:19:46 2006 -0700
@@ -282,6 +282,8 @@
 #define INFINIPATH_XGXS_RESET          0x7ULL
 #define INFINIPATH_XGXS_MDIOADDR_MASK  0xfULL
 #define INFINIPATH_XGXS_MDIOADDR_SHIFT 4
+#define INFINIPATH_XGXS_RX_POL_SHIFT 19
+#define INFINIPATH_XGXS_RX_POL_MASK 0xfULL
 
 #define INFINIPATH_RT_ADDR_MASK 0xFFFFFFFFFFULL	/* 40 bits valid */
 
diff --git a/drivers/infiniband/hw/ipath/ipath_sysfs.c b/drivers/infiniband/hw/ipath/ipath_sysfs.c
--- a/drivers/infiniband/hw/ipath/ipath_sysfs.c	Fri Aug 25 11:19:46 2006 -0700
+++ b/drivers/infiniband/hw/ipath/ipath_sysfs.c	Fri Aug 25 11:19:46 2006 -0700
@@ -561,6 +561,33 @@ bail:
 	return ret;
 }
 
+static ssize_t store_rx_pol_inv(struct device *dev,
+			  struct device_attribute *attr,
+			  const char *buf,
+			  size_t count)
+{
+	struct ipath_devdata *dd = dev_get_drvdata(dev);
+	int ret, r;
+	u16 val;
+
+	ret = ipath_parse_ushort(buf, &val);
+	if (ret < 0)
+		goto invalid;
+
+	r = ipath_set_rx_pol_inv(dd, val);
+	if (r < 0) {
+		ret = r;
+		goto bail;
+	}
+
+	goto bail;
+invalid:
+	ipath_dev_err(dd, "attempt to set invalid Rx Polarity invert\n");
+bail:
+	return ret;
+}
+
+
 static DRIVER_ATTR(num_units, S_IRUGO, show_num_units, NULL);
 static DRIVER_ATTR(version, S_IRUGO, show_version, NULL);
 
@@ -587,6 +614,7 @@ static DEVICE_ATTR(status_str, S_IRUGO, 
 static DEVICE_ATTR(status_str, S_IRUGO, show_status_str, NULL);
 static DEVICE_ATTR(boardversion, S_IRUGO, show_boardversion, NULL);
 static DEVICE_ATTR(unit, S_IRUGO, show_unit, NULL);
+static DEVICE_ATTR(rx_pol_inv, S_IWUSR, NULL, store_rx_pol_inv);
 
 static struct attribute *dev_attributes[] = {
 	&dev_attr_guid.attr,
@@ -601,6 +629,7 @@ static struct attribute *dev_attributes[
 	&dev_attr_boardversion.attr,
 	&dev_attr_unit.attr,
 	&dev_attr_enabled.attr,
+	&dev_attr_rx_pol_inv.attr,
 	NULL
 };
 

Re: [PATCH 23 of 23] IB/ipath - control receive polarity inversion

[Roland Dreier]

Applied 1-21 and 23 to my for-2.6.19 branch, and skipped 22 for now.

 - R.