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

* 'for-linus' of master.kernel.org:/pub/scm/linux/kernel/git/roland/infiniband: (33 commits)
  IB/ipath: Fix lockdep error upon "ifconfig ibN down"
  IB/ipath: Fix races with ib_resize_cq()
  IB/ipath: Support new PCIE device, QLE7142
  IB/ipath: Set CPU affinity early
  IB/ipath: Fix EEPROM read when driver is compiled with -Os
  IB/ipath: Fix and recover TXE piobuf and PBC parity errors
  IB/ipath: Change HT CRC message to indicate how to resolve problem
  IB/ipath: Clean up module exit code
  IB/ipath: Call mtrr_del with correct arguments
  IB/ipath: Flush RWQEs if access error or invalid error seen
  IB/ipath: Improved support for PowerPC
  IB/ipath: Drop unnecessary "(void *)" casts
  IB/ipath: Support multiple simultaneous devices of different types
  IB/ipath: Fix mismatch in shifts and masks for printing debug info
  IB/ipath: Fix compiler warnings and errors on non-x86_64 systems
  IB/ipath: Print more informative parity error messages
  IB/ipath: Ensure that PD of MR matches PD of QP checking the Rkey
  IB/ipath: RC and UC should validate SLID and DLID
  IB/ipath: Only allow complete writes to flash
  IB/ipath: Count SRQs properly
  ...

38 files changed, 1973 insertions, 884 deletions

diff --git a/drivers/infiniband/hw/amso1100/c2_ae.c b/drivers/infiniband/hw/amso1100/c2_ae.c
index 08f46c83a3a..3aae4978e1c 100644
--- a/drivers/infiniband/hw/amso1100/c2_ae.c
+++ b/drivers/infiniband/hw/amso1100/c2_ae.c
@@ -197,7 +197,7 @@ void c2_ae_event(struct c2_dev *c2dev, u32 mq_index)
 			"resource=%x, qp_state=%s\n",
 			__FUNCTION__,
 			to_event_str(event_id),
-			be64_to_cpu(wr->ae.ae_generic.user_context),
+			(unsigned long long) be64_to_cpu(wr->ae.ae_generic.user_context),
 			be32_to_cpu(wr->ae.ae_generic.resource_type),
 			be32_to_cpu(wr->ae.ae_generic.resource),
 			to_qp_state_str(be32_to_cpu(wr->ae.ae_generic.qp_state)));
diff --git a/drivers/infiniband/hw/amso1100/c2_alloc.c b/drivers/infiniband/hw/amso1100/c2_alloc.c
index 1d2529992c0..028a60bbfca 100644
--- a/drivers/infiniband/hw/amso1100/c2_alloc.c
+++ b/drivers/infiniband/hw/amso1100/c2_alloc.c
@@ -115,7 +115,7 @@ u16 *c2_alloc_mqsp(struct c2_dev *c2dev, struct sp_chunk *head,
 			    ((unsigned long) &(head->shared_ptr[mqsp]) -
 			     (unsigned long) head);
 		pr_debug("%s addr %p dma_addr %llx\n", __FUNCTION__,
-			 &(head->shared_ptr[mqsp]), (u64)*dma_addr);
+			 &(head->shared_ptr[mqsp]), (unsigned long long) *dma_addr);
 		return &(head->shared_ptr[mqsp]);
 	}
 	return NULL;
diff --git a/drivers/infiniband/hw/amso1100/c2_cm.c b/drivers/infiniband/hw/amso1100/c2_cm.c
index 485254efdd1..75b93e9b881 100644
--- a/drivers/infiniband/hw/amso1100/c2_cm.c
+++ b/drivers/infiniband/hw/amso1100/c2_cm.c
@@ -302,7 +302,7 @@ int c2_llp_accept(struct iw_cm_id *cm_id, struct iw_cm_conn_param *iw_param)
 	vq_req = vq_req_alloc(c2dev);
 	if (!vq_req) {
 		err = -ENOMEM;
-		goto bail1;
+		goto bail0;
 	}
 	vq_req->qp = qp;
 	vq_req->cm_id = cm_id;
@@ -311,7 +311,7 @@ int c2_llp_accept(struct iw_cm_id *cm_id, struct iw_cm_conn_param *iw_param)
 	wr = kmalloc(c2dev->req_vq.msg_size, GFP_KERNEL);
 	if (!wr) {
 		err = -ENOMEM;
-		goto bail2;
+		goto bail1;
 	}
 
 	/* Build the WR */
@@ -331,7 +331,7 @@ int c2_llp_accept(struct iw_cm_id *cm_id, struct iw_cm_conn_param *iw_param)
 	/* Validate private_data length */
 	if (iw_param->private_data_len > C2_MAX_PRIVATE_DATA_SIZE) {
 		err = -EINVAL;
-		goto bail2;
+		goto bail1;
 	}
 
 	if (iw_param->private_data) {
@@ -348,19 +348,19 @@ int c2_llp_accept(struct iw_cm_id *cm_id, struct iw_cm_conn_param *iw_param)
 	err = vq_send_wr(c2dev, (union c2wr *) wr);
 	if (err) {
 		vq_req_put(c2dev, vq_req);
-		goto bail2;
+		goto bail1;
 	}
 
 	/* Wait for reply from adapter */
 	err = vq_wait_for_reply(c2dev, vq_req);
 	if (err)
-		goto bail2;
+		goto bail1;
 
 	/* Check that reply is present */
 	reply = (struct c2wr_cr_accept_rep *) (unsigned long) vq_req->reply_msg;
 	if (!reply) {
 		err = -ENOMEM;
-		goto bail2;
+		goto bail1;
 	}
 
 	err = c2_errno(reply);
@@ -368,9 +368,8 @@ int c2_llp_accept(struct iw_cm_id *cm_id, struct iw_cm_conn_param *iw_param)
 
 	if (!err)
 		c2_set_qp_state(qp, C2_QP_STATE_RTS);
- bail2:
-	kfree(wr);
  bail1:
+	kfree(wr);
 	vq_req_free(c2dev, vq_req);
  bail0:
 	if (err) {
diff --git a/drivers/infiniband/hw/amso1100/c2_provider.c b/drivers/infiniband/hw/amso1100/c2_provider.c
index dd6af551108..da98d9f7142 100644
--- a/drivers/infiniband/hw/amso1100/c2_provider.c
+++ b/drivers/infiniband/hw/amso1100/c2_provider.c
@@ -390,14 +390,18 @@ static struct ib_mr *c2_reg_phys_mr(struct ib_pd *ib_pd,
 	}
 
 	mr = kmalloc(sizeof(*mr), GFP_KERNEL);
-	if (!mr)
+	if (!mr) {
+		vfree(page_list);
 		return ERR_PTR(-ENOMEM);
+	}
 
 	mr->pd = to_c2pd(ib_pd);
 	pr_debug("%s - page shift %d, pbl_depth %d, total_len %u, "
 		"*iova_start %llx, first pa %llx, last pa %llx\n",
 		__FUNCTION__, page_shift, pbl_depth, total_len,
-		*iova_start, page_list[0], page_list[pbl_depth-1]);
+		(unsigned long long) *iova_start,
+	       	(unsigned long long) page_list[0],
+	       	(unsigned long long) page_list[pbl_depth-1]);
   	err = c2_nsmr_register_phys_kern(to_c2dev(ib_pd->device), page_list,
  					 (1 << page_shift), pbl_depth,
 					 total_len, 0, iova_start,
diff --git a/drivers/infiniband/hw/amso1100/c2_rnic.c b/drivers/infiniband/hw/amso1100/c2_rnic.c
index f49a32b7a8f..e37c5688c21 100644
--- a/drivers/infiniband/hw/amso1100/c2_rnic.c
+++ b/drivers/infiniband/hw/amso1100/c2_rnic.c
@@ -527,7 +527,7 @@ int c2_rnic_init(struct c2_dev *c2dev)
 				      		DMA_FROM_DEVICE);
 	pci_unmap_addr_set(&c2dev->rep_vq, mapping, c2dev->rep_vq.host_dma);
 	pr_debug("%s rep_vq va %p dma %llx\n", __FUNCTION__, q1_pages,
-		 (u64)c2dev->rep_vq.host_dma);
+		 (unsigned long long) c2dev->rep_vq.host_dma);
 	c2_mq_rep_init(&c2dev->rep_vq,
 		   1,
 		   qsize,
@@ -550,7 +550,7 @@ int c2_rnic_init(struct c2_dev *c2dev)
 				      		DMA_FROM_DEVICE);
 	pci_unmap_addr_set(&c2dev->aeq, mapping, c2dev->aeq.host_dma);
 	pr_debug("%s aeq va %p dma %llx\n", __FUNCTION__, q1_pages,
-		 (u64)c2dev->rep_vq.host_dma);
+		 (unsigned long long) c2dev->rep_vq.host_dma);
 	c2_mq_rep_init(&c2dev->aeq,
 		       2,
 		       qsize,
diff --git a/drivers/infiniband/hw/ipath/ipath_common.h b/drivers/infiniband/hw/ipath/ipath_common.h
index f577905e3ac..54139d39818 100644
--- a/drivers/infiniband/hw/ipath/ipath_common.h
+++ b/drivers/infiniband/hw/ipath/ipath_common.h
@@ -141,8 +141,9 @@ struct infinipath_stats {
 	 * packets if ipath not configured, etc.)
 	 */
 	__u64 sps_krdrops;
+	__u64 sps_txeparity; /* PIO buffer parity error, recovered */
 	/* pad for future growth */
-	__u64 __sps_pad[46];
+	__u64 __sps_pad[45];
 };
 
 /*
@@ -185,6 +186,9 @@ typedef enum _ipath_ureg {
 #define IPATH_RUNTIME_PCIE	0x2
 #define IPATH_RUNTIME_FORCE_WC_ORDER	0x4
 #define IPATH_RUNTIME_RCVHDR_COPY	0x8
+#define IPATH_RUNTIME_MASTER	0x10
+#define IPATH_RUNTIME_PBC_REWRITE 0x20
+#define IPATH_RUNTIME_LOOSE_DMA_ALIGN 0x40
 
 /*
  * This structure is returned by ipath_userinit() immediately after
@@ -202,7 +206,8 @@ struct ipath_base_info {
 	/* version of software, for feature checking. */
 	__u32 spi_sw_version;
 	/* InfiniPath port assigned, goes into sent packets */
-	__u32 spi_port;
+	__u16 spi_port;
+	__u16 spi_subport;
 	/*
 	 * IB MTU, packets IB data must be less than this.
 	 * The MTU is in bytes, and will be a multiple of 4 bytes.
@@ -218,7 +223,7 @@ struct ipath_base_info {
 	__u32 spi_tidcnt;
 	/* size of the TID Eager list in infinipath, in entries */
 	__u32 spi_tidegrcnt;
-	/* size of a single receive header queue entry. */
+	/* size of a single receive header queue entry in words. */
 	__u32 spi_rcvhdrent_size;
 	/*
 	 * Count of receive header queue entries allocated.
@@ -310,6 +315,12 @@ struct ipath_base_info {
 	__u32 spi_filler_for_align;
 	/* address of readonly memory copy of the rcvhdrq tail register. */
 	__u64 spi_rcvhdr_tailaddr;
+
+	/* shared memory pages for subports if IPATH_RUNTIME_MASTER is set */
+	__u64 spi_subport_uregbase;
+	__u64 spi_subport_rcvegrbuf;
+	__u64 spi_subport_rcvhdr_base;
+
 } __attribute__ ((aligned(8)));
 
 
@@ -328,12 +339,12 @@ struct ipath_base_info {
 
 /*
  * Minor version differences are always compatible
- * a within a major version, however if if user software is larger
+ * a within a major version, however if user software is larger
  * than driver software, some new features and/or structure fields
  * may not be implemented; the user code must deal with this if it
- * cares, or it must abort after initialization reports the difference
+ * cares, or it must abort after initialization reports the difference.
  */
-#define IPATH_USER_SWMINOR 2
+#define IPATH_USER_SWMINOR 3
 
 #define IPATH_USER_SWVERSION ((IPATH_USER_SWMAJOR<<16) | IPATH_USER_SWMINOR)
 
@@ -379,7 +390,16 @@ struct ipath_user_info {
 	 */
 	__u32 spu_rcvhdrsize;
 
-	__u64 spu_unused; /* kept for compatible layout */
+	/*
+	 * If two or more processes wish to share a port, each process
+	 * must set the spu_subport_cnt and spu_subport_id to the same
+	 * values.  The only restriction on the spu_subport_id is that
+	 * it be unique for a given node.
+	 */
+	__u16 spu_subport_cnt;
+	__u16 spu_subport_id;
+
+	__u32 spu_unused; /* kept for compatible layout */
 
 	/*
 	 * address of struct base_info to write to
@@ -392,19 +412,25 @@ struct ipath_user_info {
 
 #define IPATH_CMD_MIN		16
 
-#define IPATH_CMD_USER_INIT	16	/* set up userspace */
+#define __IPATH_CMD_USER_INIT	16	/* old set up userspace (for old user code) */
 #define IPATH_CMD_PORT_INFO	17	/* find out what resources we got */
 #define IPATH_CMD_RECV_CTRL	18	/* control receipt of packets */
 #define IPATH_CMD_TID_UPDATE	19	/* update expected TID entries */
 #define IPATH_CMD_TID_FREE	20	/* free expected TID entries */
 #define IPATH_CMD_SET_PART_KEY	21	/* add partition key */
+#define IPATH_CMD_SLAVE_INFO	22	/* return info on slave processes */
+#define IPATH_CMD_ASSIGN_PORT	23	/* allocate HCA and port */
+#define IPATH_CMD_USER_INIT 	24	/* set up userspace */
 
-#define IPATH_CMD_MAX		21
+#define IPATH_CMD_MAX		24
 
 struct ipath_port_info {
 	__u32 num_active;	/* number of active units */
 	__u32 unit;		/* unit (chip) assigned to caller */
-	__u32 port;		/* port on unit assigned to caller */
+	__u16 port;		/* port on unit assigned to caller */
+	__u16 subport;		/* subport on unit assigned to caller */
+	__u16 num_ports;	/* number of ports available on unit */
+	__u16 num_subports;	/* number of subport slaves opened on port */
 };
 
 struct ipath_tid_info {
@@ -435,6 +461,8 @@ struct ipath_cmd {
 		__u32 recv_ctrl;
 		/* partition key to set */
 		__u16 part_key;
+		/* user address of __u32 bitmask of active slaves */
+		__u64 slave_mask_addr;
 	} cmd;
 };
 
@@ -596,6 +624,10 @@ struct infinipath_counters {
 
 /* K_PktFlags bits */
 #define INFINIPATH_KPF_INTR 0x1
+#define INFINIPATH_KPF_SUBPORT_MASK 0x3
+#define INFINIPATH_KPF_SUBPORT_SHIFT 1
+
+#define INFINIPATH_MAX_SUBPORT	4
 
 /* SendPIO per-buffer control */
 #define INFINIPATH_SP_TEST    0x40
@@ -610,7 +642,7 @@ struct ipath_header {
 	/*
 	 * Version - 4 bits, Port - 4 bits, TID - 10 bits and Offset -
 	 * 14 bits before ECO change ~28 Dec 03.  After that, Vers 4,
-	 * Port 3, TID 11, offset 14.
+	 * Port 4, TID 11, offset 13.
 	 */
 	__le32 ver_port_tid_offset;
 	__le16 chksum;
diff --git a/drivers/infiniband/hw/ipath/ipath_cq.c b/drivers/infiniband/hw/ipath/ipath_cq.c
index 049221bc590..87462e0cb4d 100644
--- a/drivers/infiniband/hw/ipath/ipath_cq.c
+++ b/drivers/infiniband/hw/ipath/ipath_cq.c
@@ -46,7 +46,7 @@
  */
 void ipath_cq_enter(struct ipath_cq *cq, struct ib_wc *entry, int solicited)
 {
-	struct ipath_cq_wc *wc = cq->queue;
+	struct ipath_cq_wc *wc;
 	unsigned long flags;
 	u32 head;
 	u32 next;
@@ -57,6 +57,7 @@ void ipath_cq_enter(struct ipath_cq *cq, struct ib_wc *entry, int solicited)
 	 * Note that the head pointer might be writable by user processes.
 	 * Take care to verify it is a sane value.
 	 */
+	wc = cq->queue;
 	head = wc->head;
 	if (head >= (unsigned) cq->ibcq.cqe) {
 		head = cq->ibcq.cqe;
@@ -109,21 +110,27 @@ void ipath_cq_enter(struct ipath_cq *cq, struct ib_wc *entry, int solicited)
 int ipath_poll_cq(struct ib_cq *ibcq, int num_entries, struct ib_wc *entry)
 {
 	struct ipath_cq *cq = to_icq(ibcq);
-	struct ipath_cq_wc *wc = cq->queue;
+	struct ipath_cq_wc *wc;
 	unsigned long flags;
 	int npolled;
+	u32 tail;
 
 	spin_lock_irqsave(&cq->lock, flags);
 
+	wc = cq->queue;
+	tail = wc->tail;
+	if (tail > (u32) cq->ibcq.cqe)
+		tail = (u32) cq->ibcq.cqe;
 	for (npolled = 0; npolled < num_entries; ++npolled, ++entry) {
-		if (wc->tail == wc->head)
+		if (tail == wc->head)
 			break;
-		*entry = wc->queue[wc->tail];
-		if (wc->tail >= cq->ibcq.cqe)
-			wc->tail = 0;
+		*entry = wc->queue[tail];
+		if (tail >= cq->ibcq.cqe)
+			tail = 0;
 		else
-			wc->tail++;
+			tail++;
 	}
+	wc->tail = tail;
 
 	spin_unlock_irqrestore(&cq->lock, flags);
 
@@ -177,11 +184,6 @@ struct ib_cq *ipath_create_cq(struct ib_device *ibdev, int entries,
 		goto done;
 	}
 
-	if (dev->n_cqs_allocated == ib_ipath_max_cqs) {
-		ret = ERR_PTR(-ENOMEM);
-		goto done;
-	}
-
 	/* Allocate the completion queue structure. */
 	cq = kmalloc(sizeof(*cq), GFP_KERNEL);
 	if (!cq) {
@@ -237,6 +239,16 @@ struct ib_cq *ipath_create_cq(struct ib_device *ibdev, int entries,
 	} else
 		cq->ip = NULL;
 
+	spin_lock(&dev->n_cqs_lock);
+	if (dev->n_cqs_allocated == ib_ipath_max_cqs) {
+		spin_unlock(&dev->n_cqs_lock);
+		ret = ERR_PTR(-ENOMEM);
+		goto bail_wc;
+	}
+
+	dev->n_cqs_allocated++;
+	spin_unlock(&dev->n_cqs_lock);
+
 	/*
 	 * ib_create_cq() will initialize cq->ibcq except for cq->ibcq.cqe.
 	 * The number of entries should be >= the number requested or return
@@ -253,7 +265,6 @@ struct ib_cq *ipath_create_cq(struct ib_device *ibdev, int entries,
 
 	ret = &cq->ibcq;
 
-	dev->n_cqs_allocated++;
 	goto done;
 
 bail_wc:
@@ -280,7 +291,9 @@ int ipath_destroy_cq(struct ib_cq *ibcq)
 	struct ipath_cq *cq = to_icq(ibcq);
 
 	tasklet_kill(&cq->comptask);
+	spin_lock(&dev->n_cqs_lock);
 	dev->n_cqs_allocated--;
+	spin_unlock(&dev->n_cqs_lock);
 	if (cq->ip)
 		kref_put(&cq->ip->ref, ipath_release_mmap_info);
 	else
@@ -316,10 +329,16 @@ int ipath_req_notify_cq(struct ib_cq *ibcq, enum ib_cq_notify notify)
 	return 0;
 }
 
+/**
+ * ipath_resize_cq - change the size of the CQ
+ * @ibcq: the completion queue
+ *
+ * Returns 0 for success.
+ */
 int ipath_resize_cq(struct ib_cq *ibcq, int cqe, struct ib_udata *udata)
 {
 	struct ipath_cq *cq = to_icq(ibcq);
-	struct ipath_cq_wc *old_wc = cq->queue;
+	struct ipath_cq_wc *old_wc;
 	struct ipath_cq_wc *wc;
 	u32 head, tail, n;
 	int ret;
@@ -355,6 +374,7 @@ int ipath_resize_cq(struct ib_cq *ibcq, int cqe, struct ib_udata *udata)
 	 * Make sure head and tail are sane since they
 	 * might be user writable.
 	 */
+	old_wc = cq->queue;
 	head = old_wc->head;
 	if (head > (u32) cq->ibcq.cqe)
 		head = (u32) cq->ibcq.cqe;
diff --git a/drivers/infiniband/hw/ipath/ipath_driver.c b/drivers/infiniband/hw/ipath/ipath_driver.c
index 2108466c7e3..12cefa658f3 100644
--- a/drivers/infiniband/hw/ipath/ipath_driver.c
+++ b/drivers/infiniband/hw/ipath/ipath_driver.c
@@ -95,16 +95,6 @@ const char *ipath_ibcstatus_str[] = {
 	"RecovIdle",
 };
 
-/*
- * These variables are initialized in the chip-specific files
- * but are defined here.
- */
-u16 ipath_gpio_sda_num, ipath_gpio_scl_num;
-u64 ipath_gpio_sda, ipath_gpio_scl;
-u64 infinipath_i_bitsextant;
-ipath_err_t infinipath_e_bitsextant, infinipath_hwe_bitsextant;
-u32 infinipath_i_rcvavail_mask, infinipath_i_rcvurg_mask;
-
 static void __devexit ipath_remove_one(struct pci_dev *);
 static int __devinit ipath_init_one(struct pci_dev *,
 				    const struct pci_device_id *);
@@ -527,28 +517,146 @@ bail:
 	return ret;
 }
 
+static void __devexit cleanup_device(struct ipath_devdata *dd)
+{
+	int port;
+
+	ipath_shutdown_device(dd);
+
+	if (*dd->ipath_statusp & IPATH_STATUS_CHIP_PRESENT) {
+		/* can't do anything more with chip; needs re-init */
+		*dd->ipath_statusp &= ~IPATH_STATUS_CHIP_PRESENT;
+		if (dd->ipath_kregbase) {
+			/*
+			 * if we haven't already cleaned up before these are
+			 * to ensure any register reads/writes "fail" until
+			 * re-init
+			 */
+			dd->ipath_kregbase = NULL;
+			dd->ipath_uregbase = 0;
+			dd->ipath_sregbase = 0;
+			dd->ipath_cregbase = 0;
+			dd->ipath_kregsize = 0;
+		}
+		ipath_disable_wc(dd);
+	}
+
+	if (dd->ipath_pioavailregs_dma) {
+		dma_free_coherent(&dd->pcidev->dev, PAGE_SIZE,
+				  (void *) dd->ipath_pioavailregs_dma,
+				  dd->ipath_pioavailregs_phys);
+		dd->ipath_pioavailregs_dma = NULL;
+	}
+	if (dd->ipath_dummy_hdrq) {
+		dma_free_coherent(&dd->pcidev->dev,
+			dd->ipath_pd[0]->port_rcvhdrq_size,
+			dd->ipath_dummy_hdrq, dd->ipath_dummy_hdrq_phys);
+		dd->ipath_dummy_hdrq = NULL;
+	}
+
+	if (dd->ipath_pageshadow) {
+		struct page **tmpp = dd->ipath_pageshadow;
+		dma_addr_t *tmpd = dd->ipath_physshadow;
+		int i, cnt = 0;
+
+		ipath_cdbg(VERBOSE, "Unlocking any expTID pages still "
+			   "locked\n");
+		for (port = 0; port < dd->ipath_cfgports; port++) {
+			int port_tidbase = port * dd->ipath_rcvtidcnt;
+			int maxtid = port_tidbase + dd->ipath_rcvtidcnt;
+			for (i = port_tidbase; i < maxtid; i++) {
+				if (!tmpp[i])
+					continue;
+				pci_unmap_page(dd->pcidev, tmpd[i],
+					PAGE_SIZE, PCI_DMA_FROMDEVICE);
+				ipath_release_user_pages(&tmpp[i], 1);
+				tmpp[i] = NULL;
+				cnt++;
+			}
+		}
+		if (cnt) {
+			ipath_stats.sps_pageunlocks += cnt;
+			ipath_cdbg(VERBOSE, "There were still %u expTID "
+				   "entries locked\n", cnt);
+		}
+		if (ipath_stats.sps_pagelocks ||
+		    ipath_stats.sps_pageunlocks)
+			ipath_cdbg(VERBOSE, "%llu pages locked, %llu "
+				   "unlocked via ipath_m{un}lock\n",
+				   (unsigned long long)
+				   ipath_stats.sps_pagelocks,
+				   (unsigned long long)
+				   ipath_stats.sps_pageunlocks);
+
+		ipath_cdbg(VERBOSE, "Free shadow page tid array at %p\n",
+			   dd->ipath_pageshadow);
+		vfree(dd->ipath_pageshadow);
+		dd->ipath_pageshadow = NULL;
+	}
+
+	/*
+	 * free any resources still in use (usually just kernel ports)
+	 * at unload; we do for portcnt, not cfgports, because cfgports
+	 * could have changed while we were loaded.
+	 */
+	for (port = 0; port < dd->ipath_portcnt; port++) {
+		struct ipath_portdata *pd = dd->ipath_pd[port];
+		dd->ipath_pd[port] = NULL;
+		ipath_free_pddata(dd, pd);
+	}
+	kfree(dd->ipath_pd);
+	/*
+	 * debuggability, in case some cleanup path tries to use it
+	 * after this
+	 */
+	dd->ipath_pd = NULL;
+}
+
 static void __devexit ipath_remove_one(struct pci_dev *pdev)
 {
-	struct ipath_devdata *dd;
+	struct ipath_devdata *dd = pci_get_drvdata(pdev);
 
-	ipath_cdbg(VERBOSE, "removing, pdev=%p\n", pdev);
-	if (!pdev)
-		return;
+	ipath_cdbg(VERBOSE, "removing, pdev=%p, dd=%p\n", pdev, dd);
+
+	if (dd->verbs_dev)
+		ipath_unregister_ib_device(dd->verbs_dev);
 
-	dd = pci_get_drvdata(pdev);
-	ipath_unregister_ib_device(dd->verbs_dev);
 	ipath_diag_remove(dd);
 	ipath_user_remove(dd);
 	ipathfs_remove_device(dd);
 	ipath_device_remove_group(&pdev->dev, dd);
+
 	ipath_cdbg(VERBOSE, "Releasing pci memory regions, dd %p, "
 		   "unit %u\n", dd, (u32) dd->ipath_unit);
-	if (dd->ipath_kregbase) {
-		ipath_cdbg(VERBOSE, "Unmapping kregbase %p\n",
-			   dd->ipath_kregbase);
-		iounmap((volatile void __iomem *) dd->ipath_kregbase);
-		dd->ipath_kregbase = NULL;
-	}
+
+	cleanup_device(dd);
+
+	/*
+	 * turn off rcv, send, and interrupts for all ports, all drivers
+	 * should also hard reset the chip here?
+	 * free up port 0 (kernel) rcvhdr, egr bufs, and eventually tid bufs
+	 * for all versions of the driver, if they were allocated
+	 */
+	if (pdev->irq) {
+		ipath_cdbg(VERBOSE,
+			   "unit %u free_irq of irq %x\n",
+			   dd->ipath_unit, pdev->irq);
+		free_irq(pdev->irq, dd);
+	} else
+		ipath_dbg("irq is 0, not doing free_irq "
+			  "for unit %u\n", dd->ipath_unit);
+	/*
+	 * we check for NULL here, because it's outside
+	 * the kregbase check, and we need to call it
+	 * after the free_irq.	Thus it's possible that
+	 * the function pointers were never initialized.
+	 */
+	if (dd->ipath_f_cleanup)
+		/* clean up chip-specific stuff */
+		dd->ipath_f_cleanup(dd);
+
+	ipath_cdbg(VERBOSE, "Unmapping kregbase %p\n", dd->ipath_kregbase);
+	iounmap((volatile void __iomem *) dd->ipath_kregbase);
 	pci_release_regions(pdev);
 	ipath_cdbg(VERBOSE, "calling pci_disable_device\n");
 	pci_disable_device(pdev);
@@ -760,8 +868,8 @@ static void get_rhf_errstring(u32 err, char *msg, size_t len)
 static inline void *ipath_get_egrbuf(struct ipath_devdata *dd, u32 bufnum,
 				     int err)
 {
-	return dd->ipath_port0_skbs ?
-		(void *)dd->ipath_port0_skbs[bufnum]->data : NULL;
+	return dd->ipath_port0_skbinfo ?
+		(void *) dd->ipath_port0_skbinfo[bufnum].skb->data : NULL;
 }
 
 /**
@@ -783,31 +891,34 @@ struct sk_buff *ipath_alloc_skb(struct ipath_devdata *dd,
 	 */
 
 	/*
-	 * We need 4 extra bytes for unaligned transfer copying
+	 * We need 2 extra bytes for ipath_ether data sent in the
+	 * key header.  In order to keep everything dword aligned,
+	 * we'll reserve 4 bytes.
 	 */
+	len = dd->ipath_ibmaxlen + 4;
+
 	if (dd->ipath_flags & IPATH_4BYTE_TID) {
-		/* we need a 4KB multiple alignment, and there is no way
+		/* We need a 2KB multiple alignment, and there is no way
 		 * to do it except to allocate extra and then skb_reserve
 		 * enough to bring it up to the right alignment.
 		 */
-		len = dd->ipath_ibmaxlen + 4 + (1 << 11) - 1;
+		len += 2047;
 	}
-	else
-		len = dd->ipath_ibmaxlen + 4;
+
 	skb = __dev_alloc_skb(len, gfp_mask);
 	if (!skb) {
 		ipath_dev_err(dd, "Failed to allocate skbuff, length %u\n",
 			      len);
 		goto bail;
 	}
+
+	skb_reserve(skb, 4);
+
 	if (dd->ipath_flags & IPATH_4BYTE_TID) {
-		u32 una = ((1 << 11) - 1) & (unsigned long)(skb->data + 4);
+		u32 una = (unsigned long)skb->data & 2047;
 		if (una)
-			skb_reserve(skb, 4 + (1 << 11) - una);
-		else
-			skb_reserve(skb, 4);
-	} else
-		skb_reserve(skb, 4);
+			skb_reserve(skb, 2048 - una);
+	}
 
 bail:
 	return skb;
@@ -1326,6 +1437,9 @@ int ipath_create_rcvhdrq(struct ipath_devdata *dd,
 				      "for port %u rcvhdrqtailaddr failed\n",
 				      pd->port_port);
 			ret = -ENOMEM;
+			dma_free_coherent(&dd->pcidev->dev, amt,
+					  pd->port_rcvhdrq, pd->port_rcvhdrq_phys);
+			pd->port_rcvhdrq = NULL;
 			goto bail;
 		}
 		pd->port_rcvhdrqtailaddr_phys = phys_hdrqtail;
@@ -1347,12 +1461,13 @@ int ipath_create_rcvhdrq(struct ipath_devdata *dd,
 		ipath_cdbg(VERBOSE, "reuse port %d rcvhdrq @%p %llx phys; "
 			   "hdrtailaddr@%p %llx physical\n",
 			   pd->port_port, pd->port_rcvhdrq,
-			   pd->port_rcvhdrq_phys, pd->port_rcvhdrtail_kvaddr,
-			   (unsigned long long)pd->port_rcvhdrqtailaddr_phys);
+			   (unsigned long long) pd->port_rcvhdrq_phys,
+			   pd->port_rcvhdrtail_kvaddr, (unsigned long long)
+			   pd->port_rcvhdrqtailaddr_phys);
 
 	/* clear for security and sanity on each use */
 	memset(pd->port_rcvhdrq, 0, pd->port_rcvhdrq_size);
-	memset((void *)pd->port_rcvhdrtail_kvaddr, 0, PAGE_SIZE);
+	memset(pd->port_rcvhdrtail_kvaddr, 0, PAGE_SIZE);
 
 	/*
 	 * tell chip each time we init it, even if we are re-using previous
@@ -1805,7 +1920,7 @@ void ipath_free_pddata(struct ipath_devdata *dd, struct ipath_portdata *pd)
 		pd->port_rcvhdrq = NULL;
 		if (pd->port_rcvhdrtail_kvaddr) {
 			dma_free_coherent(&dd->pcidev->dev, PAGE_SIZE,
-					 (void *)pd->port_rcvhdrtail_kvaddr,
+					 pd->port_rcvhdrtail_kvaddr,
 					 pd->port_rcvhdrqtailaddr_phys);
 			pd->port_rcvhdrtail_kvaddr = NULL;
 		}
@@ -1824,24 +1939,32 @@ void ipath_free_pddata(struct ipath_devdata *dd, struct ipath_portdata *pd)
 			dma_free_coherent(&dd->pcidev->dev, size,
 				base, pd->port_rcvegrbuf_phys[e]);
 		}
-		vfree(pd->port_rcvegrbuf);
+		kfree(pd->port_rcvegrbuf);
 		pd->port_rcvegrbuf = NULL;
-		vfree(pd->port_rcvegrbuf_phys);
+		kfree(pd->port_rcvegrbuf_phys);
 		pd-&g