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
  ...

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)));

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;

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) {

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,

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,

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;

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;

drivers/infiniband/hw/ipath/ipath_eeprom.c

@@ -100,9 +100,9 @@ static int i2c_gpio_set(struct ipath_devdata *dd,
 	gpioval = &dd->ipath_gpio_out;
 	read_val = ipath_read_kreg64(dd, dd->ipath_kregs->kr_extctrl);
 	if (line == i2c_line_scl)
-		mask = ipath_gpio_scl;
+		mask = dd->ipath_gpio_scl;
 	else
-		mask = ipath_gpio_sda;
+		mask = dd->ipath_gpio_sda;
 
 	if (new_line_state == i2c_line_high)
 		/* tri-state the output rather than force high */
@@ -119,12 +119,12 @@ static int i2c_gpio_set(struct ipath_devdata *dd,
 		write_val = 0x0UL;
 
 	if (line == i2c_line_scl) {
-		write_val <<= ipath_gpio_scl_num;
-		*gpioval = *gpioval & ~(1UL << ipath_gpio_scl_num);
+		write_val <<= dd->ipath_gpio_scl_num;
+		*gpioval = *gpioval & ~(1UL << dd->ipath_gpio_scl_num);
 		*gpioval |= write_val;
 	} else {
-		write_val <<= ipath_gpio_sda_num;
-		*gpioval = *gpioval & ~(1UL << ipath_gpio_sda_num);
+		write_val <<= dd->ipath_gpio_sda_num;
+		*gpioval = *gpioval & ~(1UL << dd->ipath_gpio_sda_num);
 		*gpioval |= write_val;
 	}
 	ipath_write_kreg(dd, dd->ipath_kregs->kr_gpio_out, *gpioval);
@@ -157,9 +157,9 @@ static int i2c_gpio_get(struct ipath_devdata *dd,
 	read_val = ipath_read_kreg64(dd, dd->ipath_kregs->kr_extctrl);
 	/* config line to be an input */
 	if (line == i2c_line_scl)
-		mask = ipath_gpio_scl;
+		mask = dd->ipath_gpio_scl;
 	else
-		mask = ipath_gpio_sda;
+		mask = dd->ipath_gpio_sda;
 	write_val = read_val & ~mask;
 	ipath_write_kreg(dd, dd->ipath_kregs->kr_extctrl, write_val);
 	read_val = ipath_read_kreg64(dd, dd->ipath_kregs->kr_extstatus);
@@ -187,6 +187,7 @@ static int i2c_gpio_get(struct ipath_devdata *dd,
 static void i2c_wait_for_writes(struct ipath_devdata *dd)
 {
 	(void)ipath_read_kreg32(dd, dd->ipath_kregs->kr_scratch);
+	rmb();
 }
 
 static void scl_out(struct ipath_devdata *dd, u8 bit)

drivers/infiniband/hw/ipath/ipath_fs.c

@@ -356,19 +356,16 @@ static ssize_t flash_write(struct file *file, const char __user *buf,
 
 	pos = *ppos;
 
-	if ( pos < 0) {
+	if (pos != 0) {
 		ret = -EINVAL;
 		goto bail;
 	}
 
-	if (pos >= sizeof(struct ipath_flash)) {
-		ret = 0;
+	if (count != sizeof(struct ipath_flash)) {
+		ret = -EINVAL;
 		goto bail;
 	}
 
-	if (count > sizeof(struct ipath_flash) - pos)
-		count = sizeof(struct ipath_flash) - pos;
-
 	tmp = kmalloc(count, GFP_KERNEL);
 	if (!tmp) {
 		ret = -ENOMEM;

drivers/infiniband/hw/ipath/ipath_iba6110.c

@@ -252,8 +252,8 @@ static const struct ipath_cregs ipath_ht_cregs = {
 };
 
 /* kr_intstatus, kr_intclear, kr_intmask bits */
-#define INFINIPATH_I_RCVURG_MASK 0x1FF
-#define INFINIPATH_I_RCVAVAIL_MASK 0x1FF
+#define INFINIPATH_I_RCVURG_MASK ((1U<<9)-1)
+#define INFINIPATH_I_RCVAVAIL_MASK ((1U<<9)-1)
 
 /* kr_hwerrclear, kr_hwerrmask, kr_hwerrstatus, bits */
 #define INFINIPATH_HWE_HTCMEMPARITYERR_SHIFT 0
@@ -338,7 +338,7 @@ static void hwerr_crcbits(struct ipath_devdata *dd, ipath_err_t hwerrs,
 	if (crcbits) {
 		u16 ctrl0, ctrl1;
 		snprintf(bitsmsg, sizeof bitsmsg,
-			 "[HT%s lane %s CRC (%llx); ignore till reload]",
+			 "[HT%s lane %s CRC (%llx); powercycle to completely clear]",
 			 !(crcbits & _IPATH_HTLINK1_CRCBITS) ?
 			 "0 (A)" : (!(crcbits & _IPATH_HTLINK0_CRCBITS)
 				    ? "1 (B)" : "0+1 (A+B)"),
@@ -389,17 +389,28 @@ static void hwerr_crcbits(struct ipath_devdata *dd, ipath_err_t hwerrs,
 				     _IPATH_HTLINK1_CRCBITS)));
 }
 
+/* 6110 specific hardware errors... */
+static const struct ipath_hwerror_msgs ipath_6110_hwerror_msgs[] = {
+	INFINIPATH_HWE_MSG(HTCBUSIREQPARITYERR, "HTC Ireq Parity"),
+	INFINIPATH_HWE_MSG(HTCBUSTREQPARITYERR, "HTC Treq Parity"),
+	INFINIPATH_HWE_MSG(HTCBUSTRESPPARITYERR, "HTC Tresp Parity"),
+	INFINIPATH_HWE_MSG(HTCMISCERR5, "HT core Misc5"),
+	INFINIPATH_HWE_MSG(HTCMISCERR6, "HT core Misc6"),
+	INFINIPATH_HWE_MSG(HTCMISCERR7, "HT core Misc7"),
+	INFINIPATH_HWE_MSG(RXDSYNCMEMPARITYERR, "Rx Dsync"),
+	INFINIPATH_HWE_MSG(SERDESPLLFAILED, "SerDes PLL"),
+};
+
 /**
- * ipath_ht_handle_hwerrors - display hardware errors
+ * ipath_ht_handle_hwerrors - display hardware errors.
  * @dd: the infinipath device
  * @msg: the output buffer
  * @msgl: the size of the output buffer
  *
- * Use same msg buffer as regular errors to avoid
- * excessive stack use.  Most hardware errors are catastrophic, but for
- * right now, we'll print them and continue.
- * We reuse the same message buffer as ipath_handle_errors() to avoid
- * excessive stack usage.
+ * Use same msg buffer as regular errors to avoid excessive stack
+ * use.  Most hardware errors are catastrophic, but for right now,
+ * we'll print them and continue.  We reuse the same message buffer as
+ * ipath_handle_errors() to avoid excessive stack usage.
  */
 static void ipath_ht_handle_hwerrors(struct ipath_devdata *dd, char *msg,
 				     size_t msgl)
@@ -440,19 +451,49 @@ static void ipath_ht_handle_hwerrors(struct ipath_devdata *dd, char *msg,
 	 * make sure we get this much out, unless told to be quiet,
 	 * or it's occurred within the last 5 seconds
 	 */
-	if ((hwerrs & ~dd->ipath_lasthwerror) ||
+	if ((hwerrs & ~(dd->ipath_lasthwerror |
+			((INFINIPATH_HWE_TXEMEMPARITYERR_PIOBUF |
+			  INFINIPATH_HWE_TXEMEMPARITYERR_PIOPBC)
+			<< INFINIPATH_HWE_TXEMEMPARITYERR_SHIFT))) ||
 	    (ipath_debug & __IPATH_VERBDBG))
 		dev_info(&dd->pcidev->dev, "Hardware error: hwerr=0x%llx "
 			 "(cleared)\n", (unsigned long long) hwerrs);
 	dd->ipath_lasthwerror |= hwerrs;
 
-	if (hwerrs & ~infinipath_hwe_bitsextant)
+	if (hwerrs & ~dd->ipath_hwe_bitsextant)
 		ipath_dev_err(dd, "hwerror interrupt with unknown errors "
 			      "%llx set\n", (unsigned long long)
-			      (hwerrs & ~infinipath_hwe_bitsextant));
+			      (hwerrs & ~dd->ipath_hwe_bitsextant));
 
 	ctrl = ipath_read_kreg32(dd, dd->ipath_kregs->kr_control);
 	if (ctrl & INFINIPATH_C_FREEZEMODE) {
+		/*
+		 * parity errors in send memory are recoverable,
+		 * just cancel the send (if indicated in * sendbuffererror),
+		 * count the occurrence, unfreeze (if no other handled
+		 * hardware error bits are set), and continue. They can
+		 * occur if a processor speculative read is done to the PIO
+		 * buffer while we are sending a packet, for example.
+		 */
+		if (hwerrs & ((INFINIPATH_HWE_TXEMEMPARITYERR_PIOBUF |
+			       INFINIPATH_HWE_TXEMEMPARITYERR_PIOPBC)
+			      << INFINIPATH_HWE_TXEMEMPARITYERR_SHIFT)) {
+			ipath_stats.sps_txeparity++;
+			ipath_dbg("Recovering from TXE parity error (%llu), "
+			    	  "hwerrstatus=%llx\n",
+				  (unsigned long long) ipath_stats.sps_txeparity,
+				  (unsigned long long) hwerrs);
+			ipath_disarm_senderrbufs(dd);
+			hwerrs &= ~((INFINIPATH_HWE_TXEMEMPARITYERR_PIOBUF |
+				     INFINIPATH_HWE_TXEMEMPARITYERR_PIOPBC)
+				    << INFINIPATH_HWE_TXEMEMPARITYERR_SHIFT);
+			if (!hwerrs) { /* else leave in freeze mode */
+				ipath_write_kreg(dd,
+						 dd->ipath_kregs->kr_control,
+						 dd->ipath_control);
+				return;
+			}
+		}
 		if (hwerrs) {
 			/*
 			 * if any set that we aren't ignoring; only
@@ -499,44 +540,16 @@ static void ipath_ht_handle_hwerrors(struct ipath_devdata *dd, char *msg,
 			 bits);
 		strlcat(msg, bitsmsg, msgl);
 	}
-	if (hwerrs & (INFINIPATH_HWE_RXEMEMPARITYERR_MASK
-		      << INFINIPATH_HWE_RXEMEMPARITYERR_SHIFT)) {
-		bits = (u32) ((hwerrs >>
-			       INFINIPATH_HWE_RXEMEMPARITYERR_SHIFT) &
-			      INFINIPATH_HWE_RXEMEMPARITYERR_MASK);
-		snprintf(bitsmsg, sizeof bitsmsg, "[RXE Parity Errs %x] ",
-			 bits);
-		strlcat(msg, bitsmsg, msgl);
-	}
-	if (hwerrs & (INFINIPATH_HWE_TXEMEMPARITYERR_MASK
-		      << INFINIPATH_HWE_TXEMEMPARITYERR_SHIFT)) {
-		bits = (u32) ((hwerrs >>
-			       INFINIPATH_HWE_TXEMEMPARITYERR_SHIFT) &
-			      INFINIPATH_HWE_TXEMEMPARITYERR_MASK);
-		snprintf(bitsmsg, sizeof bitsmsg, "[TXE Parity Errs %x] ",
-			 bits);
-		strlcat(msg, bitsmsg, msgl);
-	}
-	if (hwerrs & INFINIPATH_HWE_IBCBUSTOSPCPARITYERR)
-		strlcat(msg, "[IB2IPATH Parity]", msgl);
-	if (hwerrs & INFINIPATH_HWE_IBCBUSFRSPCPARITYERR)
-		strlcat(msg, "[IPATH2IB Parity]", msgl);
-	if (hwerrs & INFINIPATH_HWE_HTCBUSIREQPARITYERR)
-		strlcat(msg, "[HTC Ireq Parity]", msgl);
-	if (hwerrs & INFINIPATH_HWE_HTCBUSTREQPARITYERR)
-		strlcat(msg, "[HTC Treq Parity]", msgl);
-	if (hwerrs & INFINIPATH_HWE_HTCBUSTRESPPARITYERR)
-		strlcat(msg, "[HTC Tresp Parity]", msgl);
+
+	ipath_format_hwerrors(hwerrs,
+			      ipath_6110_hwerror_msgs,
+			      sizeof(ipath_6110_hwerror_msgs) /
+			      sizeof(ipath_6110_hwerror_msgs[0]),
+			      msg, msgl);
 
 	if (hwerrs & (_IPATH_HTLINK0_CRCBITS | _IPATH_HTLINK1_CRCBITS))
 		hwerr_crcbits(dd, hwerrs, msg, msgl);
 
-	if (hwerrs & INFINIPATH_HWE_HTCMISCERR5)
-		strlcat(msg, "[HT core Misc5]", msgl);
-	if (hwerrs & INFINIPATH_HWE_HTCMISCERR6)
-		strlcat(msg, "[HT core Misc6]", msgl);
-	if (hwerrs & INFINIPATH_HWE_HTCMISCERR7)
-		strlcat(msg, "[HT core Misc7]", msgl);
 	if (hwerrs & INFINIPATH_HWE_MEMBISTFAILED) {
 		strlcat(msg, "[Memory BIST test failed, InfiniPath hardware unusable]",
 			msgl);
@@ -573,11 +586,6 @@ static void ipath_ht_handle_hwerrors(struct ipath_devdata *dd, char *msg,
 				 dd->ipath_hwerrmask);
 	}
 
-	if (hwerrs & INFINIPATH_HWE_RXDSYNCMEMPARITYERR)
-		strlcat(msg, "[Rx Dsync]", msgl);
-	if (hwerrs & INFINIPATH_HWE_SERDESPLLFAILED)
-		strlcat(msg, "[SerDes PLL]", msgl);
-
 	ipath_dev_err(dd, "%s hardware error\n", msg);
 	if (isfatal && !ipath_diag_inuse && dd->ipath_freezemsg)
 		/*
@@ -1080,21 +1088,21 @@ static void ipath_setup_ht_setextled(struct ipath_devdata *dd,
 	ipath_write_kreg(dd, dd->ipath_kregs->kr_extctrl, extctl);
 }
 
-static void ipath_init_ht_variables(void)
+static void ipath_init_ht_variables(struct ipath_devdata *dd)
 {
-	ipath_gpio_sda_num = _IPATH_GPIO_SDA_NUM;
-	ipath_gpio_scl_num = _IPATH_GPIO_SCL_NUM;
-	ipath_gpio_sda = IPATH_GPIO_SDA;
-	ipath_gpio_scl = IPATH_GPIO_SCL;
+	dd->ipath_gpio_sda_num = _IPATH_GPIO_SDA_NUM;
+	dd->ipath_gpio_scl_num = _IPATH_GPIO_SCL_NUM;
+	dd->ipath_gpio_sda = IPATH_GPIO_SDA;
+	dd->ipath_gpio_scl = IPATH_GPIO_SCL;
 
-	infinipath_i_bitsextant =
+	dd->ipath_i_bitsextant =
 		(INFINIPATH_I_RCVURG_MASK << INFINIPATH_I_RCVURG_SHIFT) |
 		(INFINIPATH_I_RCVAVAIL_MASK <<
 		 INFINIPATH_I_RCVAVAIL_SHIFT) |
 		INFINIPATH_I_ERROR | INFINIPATH_I_SPIOSENT |
 		INFINIPATH_I_SPIOBUFAVAIL | INFINIPATH_I_GPIO;
 
-	infinipath_e_bitsextant =
+	dd->ipath_e_bitsextant =
 		INFINIPATH_E_RFORMATERR | INFINIPATH_E_RVCRC |
 		INFINIPATH_E_RICRC | INFINIPATH_E_RMINPKTLEN |
 		INFINIPATH_E_RMAXPKTLEN | INFINIPATH_E_RLONGPKTLEN |
@@ -1112,7 +1120,7 @@ static void ipath_init_ht_variables(void)
 		INFINIPATH_E_INVALIDADDR | INFINIPATH_E_RESET |
 		INFINIPATH_E_HARDWARE;
 
-	infinipath_hwe_bitsextant =
+	dd->ipath_hwe_bitsextant =
 		(INFINIPATH_HWE_HTCMEMPARITYERR_MASK <<
 		 INFINIPATH_HWE_HTCMEMPARITYERR_SHIFT) |
 		(INFINIPATH_HWE_TXEMEMPARITYERR_MASK <<
@@ -1141,8 +1149,8 @@ static void ipath_init_ht_variables(void)
 		INFINIPATH_HWE_IBCBUSTOSPCPARITYERR |
 		INFINIPATH_HWE_IBCBUSFRSPCPARITYERR;
 
-	infinipath_i_rcvavail_mask = INFINIPATH_I_RCVAVAIL_MASK;
-	infinipath_i_rcvurg_mask = INFINIPATH_I_RCVURG_MASK;
+	dd->ipath_i_rcvavail_mask = INFINIPATH_I_RCVAVAIL_MASK;
+	dd->ipath_i_rcvurg_mask = INFINIPATH_I_RCVURG_MASK;
 }
 
 /**
@@ -1607,5 +1615,5 @@ void ipath_init_iba6110_funcs(struct ipath_devdata *dd)
 	 * do very early init that is needed before ipath_f_bus is
 	 * called
 	 */
-	ipath_init_ht_variables();
+	ipath_init_ht_variables(dd);
 }

drivers/infiniband/hw/ipath/ipath_init_chip.c

@@ -88,13 +88,13 @@ MODULE_PARM_DESC(kpiobufs, "Set number of PIO buffers for driver");
 static int create_port0_egr(struct ipath_devdata *dd)
 {
 	unsigned e, egrcnt;
-	struct sk_buff **skbs;
+	struct ipath_skbinfo *skbinfo;
 	int ret;
 
 	egrcnt = dd->ipath_rcvegrcnt;
 
-	skbs = vmalloc(sizeof(*dd->ipath_port0_skbs) * egrcnt);
-	if (skbs == NULL) {
+	skbinfo = vmalloc(sizeof(*dd->ipath_port0_skbinfo) * egrcnt);
+	if (skbinfo == NULL) {
 		ipath_dev_err(dd, "allocation error for eager TID "
 			      "skb array\n");
 		ret = -ENOMEM;
@@ -109,13 +109,13 @@ static int create_port0_egr(struct ipath_devdata *dd)
 		 * 4 bytes so that the data buffer stays word aligned.
 		 * See ipath_kreceive() for more details.
 		 */
-		skbs[e] = ipath_alloc_skb(dd, GFP_KERNEL);
-		if (!skbs[e]) {
+		skbinfo[e].skb = ipath_alloc_skb(dd, GFP_KERNEL);
+		if (!skbinfo[e].skb) {
 			ipath_dev_err(dd, "SKB allocation error for "
 				      "eager TID %u\n", e);
 			while (e != 0)
-				dev_kfree_skb(skbs[--e]);
-			vfree(skbs);
+				dev_kfree_skb(skbinfo[--e].skb);
+			vfree(skbinfo);
 			ret = -ENOMEM;
 			goto bail;
 		}
@@ -124,14 +124,17 @@ static int create_port0_egr(struct ipath_devdata *dd)
 	 * After loop above, so we can test non-NULL to see if ready
 	 * to use at receive, etc.
 	 */
-	dd->ipath_port0_skbs = skbs;
+	dd->ipath_port0_skbinfo = skbinfo;
 
 	for (e = 0; e < egrcnt; e++) {
-		unsigned long phys =
-			virt_to_phys(dd->ipath_port0_skbs[e]->data);
+		dd->ipath_port0_skbinfo[e].phys =
+		  ipath_map_single(dd->pcidev,
+				   dd->ipath_port0_skbinfo[e].skb->data,
+				   dd->ipath_ibmaxlen, PCI_DMA_FROMDEVICE);
 		dd->ipath_f_put_tid(dd, e + (u64 __iomem *)
 				    ((char __iomem *) dd->ipath_kregbase +
-				     dd->ipath_rcvegrbase), 0, phys);
+				     dd->ipath_rcvegrbase), 0,
+				    dd->ipath_port0_skbinfo[e].phys);
 	}
 
 	ret = 0;
@@ -432,16 +435,33 @@ static int init_pioavailregs(struct ipath_devdata *dd)
  */
 static void init_shadow_tids(struct ipath_devdata *dd)
 {
-	dd->ipath_pageshadow = (struct page **)
-		vmalloc(dd->ipath_cfgports * dd->ipath_rcvtidcnt *
+	struct page **pages;
+	dma_addr_t *addrs;
+
+	pages = vmalloc(dd->ipath_cfgports * dd->ipath_rcvtidcnt *
 			sizeof(struct page *));
-	if (!dd->ipath_pageshadow)
+	if (!pages) {
 		ipath_dev_err(dd, "failed to allocate shadow page * "
 			      "array, no expected sends!\n");
-	else
-		memset(dd->ipath_pageshadow, 0,
-		       dd->ipath_cfgports * dd->ipath_rcvtidcnt *
-		       sizeof(struct page *));
+		dd->ipath_pageshadow = NULL;
+		return;
+	}
+
+	addrs = vmalloc(dd->ipath_cfgports * dd->ipath_rcvtidcnt *
+			sizeof(dma_addr_t));
+	if (!addrs) {
+		ipath_dev_err(dd, "failed to allocate shadow dma handle "
+			      "array, no expected sends!\n");
+		vfree(dd->ipath_pageshadow);
+		dd->ipath_pageshadow = NULL;
+		return;
+	}
+
+	memset(pages, 0, dd->ipath_cfgports * dd->ipath_rcvtidcnt *
+	       sizeof(struct page *));
+
+	dd->ipath_pageshadow = pages;
+	dd->ipath_physshadow = addrs;
 }
 
 static void enable_chip(struct ipath_devdata *dd,

drivers/infiniband/hw/ipath/ipath_kernel.h

@@ -39,6 +39,8 @@
  */
 
 #include <linux/interrupt.h>
+#include <linux/pci.h>
+#include <linux/dma-mapping.h>
 #include <asm/io.h>
 
 #include "ipath_common.h"
@@ -62,7 +64,7 @@ struct ipath_portdata {
 	/* rcvhdrq base, needs mmap before useful */
 	void *port_rcvhdrq;
 	/* kernel virtual address where hdrqtail is updated */
-	volatile __le64 *port_rcvhdrtail_kvaddr;
+	void *port_rcvhdrtail_kvaddr;
 	/*
 	 * temp buffer for expected send setup, allocated at open, instead
 	 * of each setup call
@@ -79,8 +81,8 @@ struct ipath_portdata {
 	dma_addr_t port_rcvhdrq_phys;
 	dma_addr_t port_rcvhdrqtailaddr_phys;
 	/*
-	 * number of opens on this instance (0 or 1; ignoring forks, dup,
-	 * etc. for now)
+	 * number of opens (including slave subports) on this instance
+	 * (ignoring forks, dup, etc. for now)
 	 */
 	int port_cnt;
 	/*
@@ -89,6 +91,10 @@ struct ipath_portdata {
 	 */
 	/* instead of calculating it */
 	unsigned port_port;
+	/* non-zero if port is being shared. */
+	u16 port_subport_cnt;
+	/* non-zero if port is being shared. */
+	u16 port_subport_id;
 	/* chip offset of PIO buffers for this port */
 	u32 port_piobufs;
 	/* how many alloc_pages() chunks in port_rcvegrbuf_pages */
@@ -121,6 +127,16 @@ struct ipath_portdata {
 	u16 port_pkeys[4];
 	/* so file ops can get at unit */
 	struct ipath_devdata *port_dd;
+	/* A page of memory for rcvhdrhead, rcvegrhead, rcvegrtail * N */
+	void *subport_uregbase;
+	/* An array of pages for the eager receive buffers * N */
+	void *subport_rcvegrbuf;
+	/* An array of pages for the eager header queue entries * N */
+	void *subport_rcvhdr_base;
+	/* The version of the library which opened this port */
+	u32 userversion;
+	/* Bitmask of active slaves */
+	u32 active_slaves;
 };
 
 struct sk_buff;
@@ -132,6 +148,11 @@ struct _ipath_layer {
 	void *l_arg;
 };
 
+struct ipath_skbinfo {
+	struct sk_buff *skb;
+	dma_addr_t phys;
+};
+
 struct ipath_devdata {
 	struct list_head ipath_list;
 
@@ -154,7 +175,7 @@ struct ipath_devdata {
 	/* ipath_cfgports pointers */
 	struct ipath_portdata **ipath_pd;
 	/* sk_buffs used by port 0 eager receive queue */
-	struct sk_buff **ipath_port0_skbs;
+	struct ipath_skbinfo *ipath_port0_skbinfo;
 	/* kvirt address of 1st 2k pio buffer */
 	void __iomem *ipath_pio2kbase;
 	/* kvirt address of 1st 4k pio buffer */
@@ -315,12 +336,16 @@ struct ipath_devdata {
 	u8 ipath_ht_slave_off;
 	/* for write combining settings */
 	unsigned long ipath_wc_cookie;
+	unsigned long ipath_wc_base;
+	unsigned long ipath_wc_len;
 	/* ref count for each pkey */
 	atomic_t ipath_pkeyrefs[4];
 	/* shadow copy of all exptids physaddr; used only by funcsim */
 	u64 *ipath_tidsimshadow;
 	/* shadow copy of struct page *'s for exp tid pages */
 	struct page **ipath_pageshadow;
+	/* shadow copy of dma handles for exp tid pages */
+	dma_addr_t *ipath_physshadow;
 	/* lock to workaround chip bug 9437 */
 	spinlock_t ipath_tid_lock;
 
@@ -402,6 +427,9 @@ struct ipath_devdata {
 	unsigned long ipath_rcvctrl;
 	/* shadow kr_sendctrl */
 	unsigned long ipath_sendctrl;
+	/* ports waiting for PIOavail intr */
+	unsigned long ipath_portpiowait;
+	unsigned long ipath_lastcancel; /* to not count armlaunch after cancel */
 
 	/* value we put in kr_rcvhdrcnt */
 	u32 ipath_rcvhdrcnt;
@@ -465,8 +493,6 @@ struct ipath_devdata {
 	u32 ipath_htwidth;
 	/* HT speed (200,400,800,1000) from HT config */
 	u32 ipath_htspeed;
-	/* ports waiting for PIOavail intr */
-	unsigned long ipath_portpiowait;
 	/*
 	 * number of sequential ibcstatus change for polling active/quiet
 	 * (i.e., link not coming up).
@@ -510,8 +536,47 @@ struct ipath_devdata {
 	u32 ipath_lli_counter;
 	/* local link integrity errors */
 	u32 ipath_lli_errors;
+	/*
+	 * Above counts only cases where _successive_ LocalLinkIntegrity
+	 * errors were seen in the receive headers of kern-packets.
+	 * Below are the three (monotonically increasing) counters
+	 * maintained via GPIO interrupts on iba6120-rev2.
+	 */
+	u32 ipath_rxfc_unsupvl_errs;
+	u32 ipath_overrun_thresh_errs;
+	u32 ipath_lli_errs;
+
+	/*
+	 * Not all devices managed by a driver instance are the same
+	 * type, so these fields must be per-device.
+	 */
+	u64 ipath_i_bitsextant;
+	ipath_err_t ipath_e_bitsextant;
+	ipath_err_t ipath_hwe_bitsextant;
+
+	/*
+	 * Below should be computable from number of ports,
+	 * since they are never modified.
+	 */
+	u32 ipath_i_rcvavail_mask;
+	u32 ipath_i_rcvurg_mask;
+
+	/*
+	 * Register bits for selecting i2c direction and values, used for
+	 * I2C serial flash.
+	 */
+	u16 ipath_gpio_sda_num;
+	u16 ipath_gpio_scl_num;
+	u64 ipath_gpio_sda;
+	u64 ipath_gpio_scl;
 };
 
+/* Private data for file operations */
+struct ipath_filedata {
+	struct ipath_portdata *pd;
+	unsigned subport;
+	unsigned tidcursor;
+};
 extern struct list_head ipath_dev_list;
 extern spinlock_t ipath_devs_lock;
 extern struct ipath_devdata *ipath_lookup(int unit);
@@ -521,6 +586,7 @@ int ipath_enable_wc(struct ipath_devdata *dd);
 void ipath_disable_wc(struct ipath_devdata *dd);
 int ipath_count_units(int *npresentp, int *nupp, u32 *maxportsp);
 void ipath_shutdown_device(struct ipath_devdata *);
+void ipath_disarm_senderrbufs(struct ipath_devdata *);
 
 struct file_operations;
 int ipath_cdev_init(int minor, char *name, struct file_operations *fops,
@@ -572,7 +638,11 @@ 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)
+#define port_fp(fp) ((struct ipath_filedata *)(fp)->private_data)->pd
+#define subport_fp(fp) \
+	((struct ipath_filedata *)(fp)->private_data)->subport
+#define tidcursor_fp(fp) \
+	((struct ipath_filedata *)(fp)->private_data)->tidcursor
 
 /*
  * values for ipath_flags
@@ -612,6 +682,15 @@ int ipath_set_rx_pol_inv(struct ipath_devdata *dd, u8 new_pol_inv);
 		/* can miss port0 rx interrupts */
 #define IPATH_POLL_RX_INTR  0x40000
 #define IPATH_DISABLED      0x80000 /* administratively disabled */
+		/* Use GPIO interrupts for new counters */
+#define IPATH_GPIO_ERRINTRS 0x100000
+
+/* Bits in GPIO for the added interrupts */
+#define IPATH_GPIO_PORT0_BIT 2
+#define IPATH_GPIO_RXUVL_BIT 3
+#define IPATH_GPIO_OVRUN_BIT 4
+#define IPATH_GPIO_LLI_BIT 5
+#define IPATH_GPIO_ERRINTR_MASK 0x38
 
 /* portdata flag bit offsets */
 		/* waiting for a packet to arrive */
@@ -798,6 +877,13 @@ void ipath_exit_ipathfs(void);
 int ipathfs_add_device(struct ipath_devdata *);
 int ipathfs_remove_device(struct ipath_devdata *);
 
+/*
+ * dma_addr wrappers - all 0's invalid for hw
+ */
+dma_addr_t ipath_map_page(struct pci_dev *, struct page *, unsigned long,
+			  size_t, int);
+dma_addr_t ipath_map_single(struct pci_dev *, void *, size_t, int);
+
 /*
  * Flush write combining store buffers (if present) and perform a write
  * barrier.
@@ -855,4 +941,20 @@ extern struct mutex ipath_mutex;
 
 #endif /* _IPATH_DEBUGGING */
 
+/*
+ * this is used for formatting hw error messages...
+ */
+struct ipath_hwerror_msgs {
+	u64 mask;
+	const char *msg;
+};
+
+#define INFINIPATH_HWE_MSG(a, b) { .mask = INFINIPATH_HWE_##a, .msg = b }
+
+/* in ipath_intr.c... */
+void ipath_format_hwerrors(u64 hwerrs,
+			   const struct ipath_hwerror_msgs *hwerrmsgs,
+			   size_t nhwerrmsgs,
+			   char *msg, size_t lmsg);
+
 #endif				/* _IPATH_KERNEL_H */

drivers/infiniband/hw/ipath/ipath_keys.c

@@ -118,9 +118,10 @@ void ipath_free_lkey(struct ipath_lkey_table *rkt, u32 lkey)
  * Check the IB SGE for validity and initialize our internal version
  * of it.
  */
-int ipath_lkey_ok(struct ipath_lkey_table *rkt, struct ipath_sge *isge,
+int ipath_lkey_ok(struct ipath_qp *qp, struct ipath_sge *isge,
 		  struct ib_sge *sge, int acc)
 {
+	struct ipath_lkey_table *rkt = &to_idev(qp->ibqp.device)->lk_table;
 	struct ipath_mregion *mr;
 	unsigned n, m;
 	size_t off;
@@ -140,7 +141,8 @@ int ipath_lkey_ok(struct ipath_lkey_table *rkt, struct ipath_sge *isge,
 		goto bail;
 	}
 	mr = rkt->table[(sge->lkey >> (32 - ib_ipath_lkey_table_size))];
-	if (unlikely(mr == NULL || mr->lkey != sge->lkey)) {
+	if (unlikely(mr == NULL || mr->lkey != sge->lkey ||
+		     qp->ibqp.pd != mr->pd)) {
 		ret = 0;
 		goto bail;
 	}
@@ -188,9 +190,10 @@ int ipath_lkey_ok(struct ipath_lkey_table *rkt, struct ipath_sge *isge,
  *
  * Return 1 if successful, otherwise 0.
  */
-int ipath_rkey_ok(struct ipath_ibdev *dev, struct ipath_sge_state *ss,
+int ipath_rkey_ok(struct ipath_qp *qp, struct ipath_sge_state *ss,
 		  u32 len, u64 vaddr, u32 rkey, int acc)
 {
+	struct ipath_ibdev *dev = to_idev(qp->ibqp.device);
 	struct ipath_lkey_table *rkt = &dev->lk_table;
 	struct ipath_sge *sge = &ss->sge;
 	struct ipath_mregion *mr;
@@ -214,7 +217,8 @@ int ipath_rkey_ok(struct ipath_ibdev *dev, struct ipath_sge_state *ss,
 	}
 
 	mr = rkt->table[(rkey >> (32 - ib_ipath_lkey_table_size))];
-	if (unlikely(mr == NULL || mr->lkey != rkey)) {
+	if (unlikely(mr == NULL || mr->lkey != rkey ||
+		     qp->ibqp.pd != mr->pd)) {
 		ret = 0;
 		goto bail;
 	}

drivers/infiniband/hw/ipath/ipath_mad.c

@@ -87,7 +87,8 @@ static int recv_subn_get_nodeinfo(struct ib_smp *smp,
 	struct ipath_devdata *dd = to_idev(ibdev)->dd;
 	u32 vendor, majrev, minrev;
 
-	if (smp->attr_mod)
+	/* GUID 0 is illegal */
+	if (smp->attr_mod || (dd->ipath_guid == 0))
 		smp->status |= IB_SMP_INVALID_FIELD;
 
 	nip->base_version = 1;
@@ -131,10 +132,15 @@ static int recv_subn_get_guidinfo(struct ib_smp *smp,
 	 * We only support one GUID for now.  If this changes, the
 	 * portinfo.guid_cap field needs to be updated too.
 	 */
-	if (startgx == 0)
-		/* The first is a copy of the read-only HW GUID. */
-		*p = to_idev(ibdev)->dd->ipath_guid;
-	else
+	if (startgx == 0) {
+		__be64 g = to_idev(ibdev)->dd->ipath_guid;
+		if (g == 0)
+			/* GUID 0 is illegal */
+			smp->status |= IB_SMP_INVALID_FIELD;
+		else
+			/* The first is a copy of the read-only HW GUID. */
+			*p = g;
+	} else
 		smp->status |= IB_SMP_INVALID_FIELD;
 
 	return reply(smp);

drivers/infiniband/hw/ipath/ipath_mr.c

@@ -138,6 +138,7 @@ struct ib_mr *ipath_reg_phys_mr(struct ib_pd *pd,
 		goto bail;
 	}
 
+	mr->mr.pd = pd;
 	mr->mr.user_base = *iova_start;
 	mr->mr.iova = *iova_start;
 	mr->mr.length = 0;
@@ -197,6 +198,7 @@ struct ib_mr *ipath_reg_user_mr(struct ib_pd *pd, struct ib_umem *region,
 		goto bail;
 	}
 
+	mr->mr.pd = pd;
 	mr->mr.user_base = region->user_base;
 	mr->mr.iova = region->virt_base;
 	mr->mr.length = region->length;
@@ -289,6 +291,7 @@ struct ib_fmr *ipath_alloc_fmr(struct ib_pd *pd, int mr_access_flags,
 	 * Resources are allocated but no valid mapping (RKEY can't be
 	 * used).
 	 */
+	fmr->mr.pd = pd;
 	fmr->mr.user_base = 0;
 	fmr->mr.iova = 0;
 	fmr->mr.length = 0;

drivers/infiniband/hw/ipath/ipath_qp.c

@@ -335,6 +335,7 @@ static void ipath_reset_qp(struct ipath_qp *qp)
 	qp->s_ack_state = IB_OPCODE_RC_ACKNOWLEDGE;
 	qp->r_ack_state = IB_OPCODE_RC_ACKNOWLEDGE;
 	qp->r_nak_state = 0;
+	qp->r_wrid_valid = 0;
 	qp->s_rnr_timeout = 0;
 	qp->s_head = 0;
 	qp->s_tail = 0;
@@ -342,6 +343,7 @@ static void ipath_reset_qp(struct ipath_qp *qp)
 	qp->s_last = 0;
 	qp->s_ssn = 1;
 	qp->s_lsn = 0;
+	qp->s_wait_credit = 0;
 	if (qp->r_rq.wq) {
 		qp->r_rq.wq->head = 0;
 		qp->r_rq.wq->tail = 0;
@@ -352,12 +354,13 @@ static void ipath_reset_qp(struct ipath_qp *qp)
 /**
  * ipath_error_qp - put a QP into an error state
  * @qp: the QP to put into an error state
+ * @err: the receive completion error to signal if a RWQE is active
  *
  * Flushes both send and receive work queues.
  * QP s_lock should be held and interrupts disabled.
  */
 
-void ipath_error_qp(struct ipath_qp *qp)
+void ipath_error_qp(struct ipath_qp *qp, enum ib_wc_status err)
 {
 	struct ipath_ibdev *dev = to_idev(qp->ibqp.device);
 	struct ib_wc wc;
@@ -373,7 +376,6 @@ void ipath_error_qp(struct ipath_qp *qp)
 		list_del_init(&qp->piowait);
 	spin_unlock(&dev->pending_lock);
 
-	wc.status = IB_WC_WR_FLUSH_ERR;
 	wc.vendor_err = 0;
 	wc.byte_len = 0;
 	wc.imm_data = 0;
@@ -385,6 +387,12 @@ void ipath_error_qp(struct ipath_qp *qp)
 	wc.sl = 0;
 	wc.dlid_path_bits = 0;
 	wc.port_num = 0;
+	if (qp->r_wrid_valid) {
+		qp->r_wrid_valid = 0;
+		wc.status = err;
+		ipath_cq_enter(to_icq(qp->ibqp.send_cq), &wc, 1);
+	}
+	wc.status = IB_WC_WR_FLUSH_ERR;
 
 	while (qp->s_last != qp->s_head) {
 		struct ipath_swqe *wqe = get_swqe_ptr(qp, qp->s_last);
@@ -501,7 +509,7 @@ int ipath_modify_qp(struct ib_qp *ibqp, struct ib_qp_attr *attr,
 		break;
 
 	case IB_QPS_ERR:
-		ipath_error_qp(qp);
+		ipath_error_qp(qp, IB_WC_GENERAL_ERR);
 		break;
 
 	default:
@@ -516,7 +524,7 @@ int ipath_modify_qp(struct ib_qp *ibqp, struct ib_qp_attr *attr,
 		qp->remote_qpn = attr->dest_qp_num;
 
 	if (attr_mask & IB_QP_SQ_PSN) {
-		qp->s_next_psn = attr->sq_psn;
+		qp->s_psn = qp->s_next_psn = attr->sq_psn;
 		qp->s_last_psn = qp->s_next_psn - 1;
 	}
 

drivers/infiniband/hw/ipath/ipath_rc.c

@@ -201,6 +201,18 @@ int ipath_make_rc_req(struct ipath_qp *qp,
 	    qp->s_rnr_timeout)
 		goto done;
 
+	/* Limit the number of packets sent without an ACK. */
+	if (ipath_cmp24(qp->s_psn, qp->s_last_psn + IPATH_PSN_CREDIT) > 0) {
+		qp->s_wait_credit = 1;
+		dev->n_rc_stalls++;
+		spin_lock(&dev->pending_lock);
+		if (list_empty(&qp->timerwait))
+			list_add_tail(&qp->timerwait,
+				      &dev->pending[dev->pending_index]);
+		spin_unlock(&dev->pending_lock);
+		goto done;
+	}
+
 	/* header size in 32-bit words LRH+BTH = (8+12)/4. */
 	hwords = 5;
 	bth0 = 0;
@@ -221,7 +233,7 @@ int ipath_make_rc_req(struct ipath_qp *qp,
 			/* Check if send work queue is empty. */
 			if (qp->s_tail == qp->s_head)
 				goto done;
-			qp->s_psn = wqe->psn = qp->s_next_psn;
+			wqe->psn = qp->s_next_psn;
 			newreq = 1;
 		}
 		/*
@@ -393,12 +405,6 @@ int ipath_make_rc_req(struct ipath_qp *qp,
 		ss = &qp->s_sge;
 		len = qp->s_len;
 		if (len > pmtu) {
-			/*
-			 * Request an ACK every 1/2 MB to avoid retransmit
-			 * timeouts.
-			 */
-			if (((wqe->length - len) % (512 * 1024)) == 0)
-				bth2 |= 1 << 31;
 			len = pmtu;
 			break;
 		}
@@ -435,12 +441,6 @@ int ipath_make_rc_req(struct ipath_qp *qp,
 		ss = &qp->s_sge;
 		len = qp->s_len;
 		if (len > pmtu) {
-			/*
-			 * Request an ACK every 1/2 MB to avoid retransmit
-			 * timeouts.
-			 */
-			if (((wqe->length - len) % (512 * 1024)) == 0)
-				bth2 |= 1 << 31;
 			len = pmtu;
 			break;
 		}
@@ -498,6 +498,8 @@ int ipath_make_rc_req(struct ipath_qp *qp,
 		 */
 		goto done;
 	}
+	if (ipath_cmp24(qp->s_psn, qp->s_last_psn + IPATH_PSN_CREDIT - 1) >= 0)
+		bth2 |= 1 << 31;	/* Request ACK. */
 	qp->s_len -= len;
 	qp->s_hdrwords = hwords;
 	qp->s_cur_sge = ss;
@@ -737,6 +739,15 @@ void ipath_restart_rc(struct ipath_qp *qp, u32 psn, struct ib_wc *wc)
 	return;
 }
 
+static inline void update_last_psn(struct ipath_qp *qp, u32 psn)
+{
+	if (qp->s_wait_credit) {
+		qp->s_wait_credit = 0;
+		tasklet_hi_schedule(&qp->s_task);
+	}
+	qp->s_last_psn = psn;
+}
+
 /**
  * do_rc_ack - process an incoming RC ACK
  * @qp: the QP the ACK came in on
@@ -805,7 +816,7 @@ static int do_rc_ack(struct ipath_qp *qp, u32 aeth, u32 psn, int opcode)
 			 * The last valid PSN seen is the previous
 			 * request's.
 			 */
-			qp->s_last_psn = wqe->psn - 1;
+			update_last_psn(qp, wqe->psn - 1);
 			/* Retry this request. */
 			ipath_restart_rc(qp, wqe->psn, &wc);
 			/*
@@ -864,7 +875,7 @@ static int do_rc_ack(struct ipath_qp *qp, u32 aeth, u32 psn, int opcode)
 		ipath_get_credit(qp, aeth);
 		qp->s_rnr_retry = qp->s_rnr_retry_cnt;
 		qp->s_retry = qp->s_retry_cnt;
-		qp->s_last_psn = psn;
+		update_last_psn(qp, psn);
 		ret = 1;
 		goto bail;
 
@@ -883,7 +894,7 @@ static int do_rc_ack(struct ipath_qp *qp, u32 aeth, u32 psn, int opcode)
 			goto bail;
 
 		/* The last valid PSN is the previous PSN. */
-		qp->s_last_psn = psn - 1;
+		update_last_psn(qp, psn - 1);
 
 		dev->n_rc_resends += (int)qp->s_psn - (int)psn;
 
@@ -898,7 +909,7 @@ static int do_rc_ack(struct ipath_qp *qp, u32 aeth, u32 psn, int opcode)
 	case 3:		/* NAK */
 		/* The last valid PSN seen is the previous request's. */
 		if (qp->s_last != qp->s_tail)
-			qp->s_last_psn = wqe->psn - 1;
+			update_last_psn(qp, wqe->psn - 1);
 		switch ((aeth >> IPATH_AETH_CREDIT_SHIFT) &
 			IPATH_AETH_CREDIT_MASK) {
 		case 0:	/* PSN sequence error */
@@ -1071,7 +1082,7 @@ static inline void ipath_rc_rcv_resp(struct ipath_ibdev *dev,
 		 * since we don't want s_sge modified.
 		 */
 		qp->s_len -= pmtu;
-		qp->s_last_psn = psn;
+		update_last_psn(qp, psn);
 		spin_unlock_irqrestore(&qp->s_lock, flags);
 		ipath_copy_sge(&qp->s_sge, data, pmtu);
 		goto bail;
@@ -1223,7 +1234,7 @@ static inline int ipath_rc_rcv_error(struct ipath_ibdev *dev,
 			 * Address range must be a subset of the original
 			 * request and start on pmtu boundaries.
 			 */
-			ok = ipath_rkey_ok(dev, &qp->s_rdma_sge,
+			ok = ipath_rkey_ok(qp, &qp->s_rdma_sge,
 					   qp->s_rdma_len, vaddr, rkey,
 					   IB_ACCESS_REMOTE_READ);
 			if (unlikely(!ok)) {
@@ -1282,6 +1293,14 @@ static inline int ipath_rc_rcv_error(struct ipath_ibdev *dev,
 	return 1;
 }
 
+static void ipath_rc_error(struct ipath_qp *qp, enum ib_wc_status err)
+{
+	spin_lock_irq(&qp->s_lock);
+	qp->state = IB_QPS_ERR;
+	ipath_error_qp(qp, err);
+	spin_unlock_irq(&qp->s_lock);
+}
+
 /**
  * ipath_rc_rcv - process an incoming RC packet
  * @dev: the device this packet came in on
@@ -1309,6 +1328,10 @@ void ipath_rc_rcv(struct ipath_ibdev *dev, struct ipath_ib_header *hdr,
 	struct ib_reth *reth;
 	int header_in_data;
 
+	/* Validate the SLID. See Ch. 9.6.1.5 */
+	if (unlikely(be16_to_cpu(hdr->lrh[3]) != qp->remote_ah_attr.dlid))
+		goto done;
+
 	/* Check for GRH */
 	if (!has_grh) {
 		ohdr = &hdr->u.oth;
@@ -1370,8 +1393,7 @@ void ipath_rc_rcv(struct ipath_ibdev *dev, struct ipath_ib_header *hdr,
 		 */
 		if (qp->r_ack_state >= OP(COMPARE_SWAP))
 			goto send_ack;
-		/* XXX Flush WQEs */
-		qp->state = IB_QPS_ERR;
+		ipath_rc_error(qp, IB_WC_REM_INV_REQ_ERR);
 		qp->r_ack_state = OP(SEND_ONLY);
 		qp->r_nak_state = IB_NAK_INVALID_REQUEST;
 		qp->r_ack_psn = qp->r_psn;
@@ -1477,9 +1499,9 @@ void ipath_rc_rcv(struct ipath_ibdev *dev, struct ipath_ib_header *hdr,
 			goto nack_inv;
 		ipath_copy_sge(&qp->r_sge, data, tlen);
 		qp->r_msn++;
-		if (opcode == OP(RDMA_WRITE_LAST) ||
-		    opcode == OP(RDMA_WRITE_ONLY))
+		if (!qp->r_wrid_valid)
 			break;
+		qp->r_wrid_valid = 0;
 		wc.wr_id = qp->r_wr_id;
 		wc.status = IB_WC_SUCCESS;
 		wc.opcode = IB_WC_RECV;
@@ -1517,7 +1539,7 @@ void ipath_rc_rcv(struct ipath_ibdev *dev, struct ipath_ib_header *hdr,
 			int ok;
 
 			/* Check rkey & NAK */
-			ok = ipath_rkey_ok(dev, &qp->r_sge,
+			ok = ipath_rkey_ok(qp, &qp->r_sge,
 					   qp->r_len, vaddr, rkey,
 					   IB_ACCESS_REMOTE_WRITE);
 			if (unlikely(!ok))
@@ -1559,7 +1581,7 @@ void ipath_rc_rcv(struct ipath_ibdev *dev, struct ipath_ib_header *hdr,
 			int ok;
 
 			/* Check rkey & NAK */
-			ok = ipath_rkey_ok(dev, &qp->s_rdma_sge,
+			ok = ipath_rkey_ok(qp, &qp->s_rdma_sge,
 					   qp->s_rdma_len, vaddr, rkey,
 					   IB_ACCESS_REMOTE_READ);
 			if (unlikely(!ok)) {
@@ -1618,7 +1640,7 @@ void ipath_rc_rcv(struct ipath_ibdev *dev, struct ipath_ib_header *hdr,
 			goto nack_inv;
 		rkey = be32_to_cpu(ateth->rkey);
 		/* Check rkey & NAK */
-		if (unlikely(!ipath_rkey_ok(dev, &qp->r_sge,
+		if (unlikely(!ipath_rkey_ok(qp, &qp->r_sge,
 					    sizeof(u64), vaddr, rkey,
 					    IB_ACCESS_REMOTE_ATOMIC)))
 			goto nack_acc;
@@ -1670,8 +1692,7 @@ void ipath_rc_rcv(struct ipath_ibdev *dev, struct ipath_ib_header *hdr,
 	 * is pending though.
 	 */
 	if (qp->r_ack_state < OP(COMPARE_SWAP)) {
-		/* XXX Flush WQEs */
-		qp->state = IB_QPS_ERR;
+		ipath_rc_error(qp, IB_WC_REM_ACCESS_ERR);
 		qp->r_ack_state = OP(RDMA_WRITE_ONLY);
 		qp->r_nak_state = IB_NAK_REMOTE_ACCESS_ERROR;
 		qp->r_ack_psn = qp->r_psn;

drivers/infiniband/hw/ipath/ipath_registers.h

@@ -134,10 +134,24 @@
 #define INFINIPATH_HWE_TXEMEMPARITYERR_SHIFT 40
 #define INFINIPATH_HWE_RXEMEMPARITYERR_MASK 0x7FULL
 #define INFINIPATH_HWE_RXEMEMPARITYERR_SHIFT 44
-#define INFINIPATH_HWE_RXDSYNCMEMPARITYERR  0x0000000400000000ULL
-#define INFINIPATH_HWE_MEMBISTFAILED        0x0040000000000000ULL
 #define INFINIPATH_HWE_IBCBUSTOSPCPARITYERR 0x4000000000000000ULL
 #define INFINIPATH_HWE_IBCBUSFRSPCPARITYERR 0x8000000000000000ULL
+/* txe mem parity errors (shift by INFINIPATH_HWE_TXEMEMPARITYERR_SHIFT) */
+#define INFINIPATH_HWE_TXEMEMPARITYERR_PIOBUF	0x1ULL
+#define INFINIPATH_HWE_TXEMEMPARITYERR_PIOPBC	0x2ULL
+#define INFINIPATH_HWE_TXEMEMPARITYERR_PIOLAUNCHFIFO 0x4ULL
+/* rxe mem parity errors (shift by INFINIPATH_HWE_RXEMEMPARITYERR_SHIFT) */
+#define INFINIPATH_HWE_RXEMEMPARITYERR_RCVBUF   0x01ULL
+#define INFINIPATH_HWE_RXEMEMPARITYERR_LOOKUPQ  0x02ULL
+#define INFINIPATH_HWE_RXEMEMPARITYERR_EAGERTID 0x04ULL
+#define INFINIPATH_HWE_RXEMEMPARITYERR_EXPTID   0x08ULL
+#define INFINIPATH_HWE_RXEMEMPARITYERR_FLAGBUF  0x10ULL
+#define INFINIPATH_HWE_RXEMEMPARITYERR_DATAINFO 0x20ULL
+#define INFINIPATH_HWE_RXEMEMPARITYERR_HDRINFO  0x40ULL
+/* waldo specific -- find the rest in ipath_6110.c */
+#define INFINIPATH_HWE_RXDSYNCMEMPARITYERR  0x0000000400000000ULL
+/* monty specific -- find the rest in ipath_6120.c */
+#define INFINIPATH_HWE_MEMBISTFAILED	0x0040000000000000ULL
 
 /* kr_hwdiagctrl bits */
 #define INFINIPATH_DC_FORCETXEMEMPARITYERR_MASK 0xFULL
@@ -209,9 +223,9 @@
 
 /* combination link status states that we use with some frequency */
 #define IPATH_IBSTATE_MASK ((INFINIPATH_IBCS_LINKTRAININGSTATE_MASK \
-		<< INFINIPATH_IBCS_LINKSTATE_SHIFT) | \
+		<< INFINIPATH_IBCS_LINKTRAININGSTATE_SHIFT) | \
 		(INFINIPATH_IBCS_LINKSTATE_MASK \
-		<<INFINIPATH_IBCS_LINKTRAININGSTATE_SHIFT))
+		<<INFINIPATH_IBCS_LINKSTATE_SHIFT))
 #define IPATH_IBSTATE_INIT ((INFINIPATH_IBCS_L_STATE_INIT \
 		<< INFINIPATH_IBCS_LINKSTATE_SHIFT) | \
 		(INFINIPATH_IBCS_LT_STATE_LINKUP \
@@ -302,6 +316,17 @@
 
 typedef u64 ipath_err_t;
 
+/* The following change with the type of device, so
+ * need to be part of the ipath_devdata struct, or
+ * we could have problems plugging in devices of
+ * different types (e.g. one HT, one PCIE)
+ * in one system, to be managed by one driver.
+ * On the other hand, this file is may also be included
+ * by other code, so leave the declarations here
+ * temporarily. Minor footprint issue if common-model
+ * linker used, none if C89+ linker used.
+ */
+
 /* mask of defined bits for various registers */
 extern u64 infinipath_i_bitsextant;
 extern ipath_err_t infinipath_e_bitsextant, infinipath_hwe_bitsextant;
@@ -309,13 +334,6 @@ extern ipath_err_t infinipath_e_bitsextant, infinipath_hwe_bitsextant;
 /* masks that are different in various chips, or only exist in some chips */
 extern u32 infinipath_i_rcvavail_mask, infinipath_i_rcvurg_mask;
 
-/*
- * register bits for selecting i2c direction and values, used for I2C serial
- * flash
- */
-extern u16 ipath_gpio_sda_num, ipath_gpio_scl_num;
-extern u64 ipath_gpio_sda, ipath_gpio_scl;
-
 /*
  * These are the infinipath general register numbers (not offsets).
  * The kernel registers are used directly, those beyond the kernel

drivers/infiniband/hw/ipath/ipath_ruc.c

@@ -108,7 +108,6 @@ void ipath_insert_rnr_queue(struct ipath_qp *qp)
 
 static int init_sge(struct ipath_qp *qp, struct ipath_rwqe *wqe)
 {
-	struct ipath_ibdev *dev = to_idev(qp->ibqp.device);
 	int user = to_ipd(qp->ibqp.pd)->user;
 	int i, j, ret;
 	struct ib_wc wc;
@@ -119,8 +118,7 @@ static int init_sge(struct ipath_qp *qp, struct ipath_rwqe *wqe)
 			continue;
 		/* Check LKEY */
 		if ((user && wqe->sg_list[i].lkey == 0) ||
-		    !ipath_lkey_ok(&dev->lk_table,
-				   &qp->r_sg_list[j], &wqe->sg_list[i],
+		    !ipath_lkey_ok(qp, &qp->r_sg_list[j], &wqe->sg_list[i],
 				   IB_ACCESS_LOCAL_WRITE))
 			goto bad_lkey;
 		qp->r_len += wqe->sg_list[i].length;
@@ -231,6 +229,7 @@ int ipath_get_rwqe(struct ipath_qp *qp, int wr_id_only)
 		}
 	}
 	spin_unlock_irqrestore(&rq->lock, flags);
+	qp->r_wrid_valid = 1;
 
 bail:
 	return ret;
@@ -326,7 +325,7 @@ static void ipath_ruc_loopback(struct ipath_qp *sqp)
 	case IB_WR_RDMA_WRITE:
 		if (wqe->length == 0)
 			break;
-		if (unlikely(!ipath_rkey_ok(dev, &qp->r_sge, wqe->length,
+		if (unlikely(!ipath_rkey_ok(qp, &qp->r_sge, wqe->length,
 					    wqe->wr.wr.rdma.remote_addr,
 					    wqe->wr.wr.rdma.rkey,
 					    IB_ACCESS_REMOTE_WRITE))) {
@@ -350,7 +349,7 @@ static void ipath_ruc_loopback(struct ipath_qp *sqp)
 		break;
 
 	case IB_WR_RDMA_READ:
-		if (unlikely(!ipath_rkey_ok(dev, &sqp->s_sge, wqe->length,
+		if (unlikely(!ipath_rkey_ok(qp, &sqp->s_sge, wqe->length,
 					    wqe->wr.wr.rdma.remote_addr,
 					    wqe->wr.wr.rdma.rkey,
 					    IB_ACCESS_REMOTE_READ)))
@@ -365,7 +364,7 @@ static void ipath_ruc_loopback(struct ipath_qp *sqp)
 
 	case IB_WR_ATOMIC_CMP_AND_SWP:
 	case IB_WR_ATOMIC_FETCH_AND_ADD:
-		if (unlikely(!ipath_rkey_ok(dev, &qp->r_sge, sizeof(u64),
+		if (unlikely(!ipath_rkey_ok(qp, &qp->r_sge, sizeof(u64),
 					    wqe->wr.wr.rdma.remote_addr,
 					    wqe->wr.wr.rdma.rkey,
 					    IB_ACCESS_REMOTE_ATOMIC)))
@@ -575,8 +574,7 @@ int ipath_post_ruc_send(struct ipath_qp *qp, struct ib_send_wr *wr)
 		}
 		if (wr->sg_list[i].length == 0)
 			continue;
-		if (!ipath_lkey_ok(&to_idev(qp->ibqp.device)->lk_table,
-				   &wqe->sg_list[j], &wr->sg_list[i],
+		if (!ipath_lkey_ok(qp, &wqe->sg_list[j], &wr->sg_list[i],
 				   acc)) {
 			spin_unlock_irqrestore(&qp->s_lock, flags);
 			ret = -EINVAL;

drivers/infiniband/hw/ipath/ipath_srq.c

@@ -104,11 +104,6 @@ struct ib_srq *ipath_create_srq(struct ib_pd *ibpd,
 	u32 sz;
 	struct ib_srq *ret;
 
-	if (dev->n_srqs_allocated == ib_ipath_max_srqs) {
-		ret = ERR_PTR(-ENOMEM);
-		goto done;
-	}
-
 	if (srq_init_attr->attr.max_wr == 0) {
 		ret = ERR_PTR(-EINVAL);
 		goto done;
@@ -180,10 +175,17 @@ struct ib_srq *ipath_create_srq(struct ib_pd *ibpd,
 	spin_lock_init(&srq->rq.lock);
 	srq->rq.wq->head = 0;
 	srq->rq.wq->tail = 0;
-	srq->rq.max_sge = srq_init_attr->attr.max_sge;
 	srq->limit = srq_init_attr->attr.srq_limit;
 
-	dev->n_srqs_allocated++;
+	spin_lock(&dev->n_srqs_lock);
+	if (dev->n_srqs_allocated == ib_ipath_max_srqs) {
+		spin_unlock(&dev->n_srqs_lock);
+		ret = ERR_PTR(-ENOMEM);
+		goto bail_wq;
+	}
+
+ 	dev->n_srqs_allocated++;
+	spin_unlock(&dev->n_srqs_lock);
 
 	ret = &srq->ibsrq;
 	goto done;
@@ -351,8 +353,13 @@ int ipath_destroy_srq(struct ib_srq *ibsrq)
 	struct ipath_srq *srq = to_isrq(ibsrq);
 	struct ipath_ibdev *dev = to_idev(ibsrq->device);
 
+	spin_lock(&dev->n_srqs_lock);
 	dev->n_srqs_allocated--;
-	vfree(srq->rq.wq);
+	spin_unlock(&dev->n_srqs_lock);
+	if (srq->ip)
+		kref_put(&srq->ip->ref, ipath_release_mmap_info);
+	else
+		vfree(srq->rq.wq);
 	kfree(srq);
 
 	return 0;

drivers/infiniband/hw/ipath/ipath_sysfs.c

@@ -257,7 +257,7 @@ static ssize_t store_guid(struct device *dev,
 	struct ipath_devdata *dd = dev_get_drvdata(dev);
 	ssize_t ret;
 	unsigned short guid[8];
-	__be64 nguid;
+	__be64 new_guid;
 	u8 *ng;
 	int i;
 
@@ -266,7 +266,7 @@ static ssize_t store_guid(struct device *dev,
 		   &guid[4], &guid[5], &guid[6], &guid[7]) != 8)
 		goto invalid;
 
-	ng = (u8 *) &nguid;
+	ng = (u8 *) &new_guid;
 
 	for (i = 0; i < 8; i++) {
 		if (guid[i] > 0xff)
@@ -274,7 +274,10 @@ static ssize_t store_guid(struct device *dev,
 		ng[i] = guid[i];
 	}
 
-	dd->ipath_guid = nguid;
+	if (new_guid == 0)
+		goto invalid;
+
+	dd->ipath_guid = new_guid;
 	dd->ipath_nguid = 1;
 
 	ret = strlen(buf);
@@ -297,6 +300,16 @@ static ssize_t show_nguid(struct device *dev,
 	return scnprintf(buf, PAGE_SIZE, "%u\n", dd->ipath_nguid);
 }
 
+static ssize_t show_nports(struct device *dev,
+			   struct device_attribute *attr,
+			   char *buf)
+{
+	struct ipath_devdata *dd = dev_get_drvdata(dev);
+
+	/* Return the number of user ports available. */
+	return scnprintf(buf, PAGE_SIZE, "%u\n", dd->ipath_cfgports - 1);
+}
+
 static ssize_t show_serial(struct device *dev,
 			   struct device_attribute *attr,
 			   char *buf)
@@ -608,6 +621,7 @@ static DEVICE_ATTR(mlid, S_IWUSR | S_IRUGO, show_mlid, store_mlid);
 static DEVICE_ATTR(mtu, S_IWUSR | S_IRUGO, show_mtu, store_mtu);
 static DEVICE_ATTR(enabled, S_IWUSR | S_IRUGO, show_enabled, store_enabled);
 static DEVICE_ATTR(nguid, S_IRUGO, show_nguid, NULL);
+static DEVICE_ATTR(nports, S_IRUGO, show_nports, NULL);
 static DEVICE_ATTR(reset, S_IWUSR, NULL, store_reset);
 static DEVICE_ATTR(serial, S_IRUGO, show_serial, NULL);
 static DEVICE_ATTR(status, S_IRUGO, show_status, NULL);
@@ -623,6 +637,7 @@ static struct attribute *dev_attributes[] = {
 	&dev_attr_mlid.attr,
 	&dev_attr_mtu.attr,
 	&dev_attr_nguid.attr,
+	&dev_attr_nports.attr,
 	&dev_attr_serial.attr,
 	&dev_attr_status.attr,
 	&dev_attr_status_str.attr,

drivers/infiniband/hw/ipath/ipath_uc.c

@@ -246,6 +246,10 @@ void ipath_uc_rcv(struct ipath_ibdev *dev, struct ipath_ib_header *hdr,
 	struct ib_reth *reth;
 	int header_in_data;
 
+	/* Validate the SLID. See Ch. 9.6.1.5 */
+	if (unlikely(be16_to_cpu(hdr->lrh[3]) != qp->remote_ah_attr.dlid))
+		goto done;
+
 	/* Check for GRH */
 	if (!has_grh) {
 		ohdr = &hdr->u.oth;
@@ -440,7 +444,7 @@ void ipath_uc_rcv(struct ipath_ibdev *dev, struct ipath_ib_header *hdr,
 			int ok;
 
 			/* Check rkey */
-			ok = ipath_rkey_ok(dev, &qp->r_sge, qp->r_len,
+			ok = ipath_rkey_ok(qp, &qp->r_sge, qp->r_len,
 					   vaddr, rkey,
 					   IB_ACCESS_REMOTE_WRITE);
 			if (unlikely(!ok)) {

drivers/infiniband/hw/ipath/ipath_ud.c

@@ -39,7 +39,6 @@
 static int init_sge(struct ipath_qp *qp, struct ipath_rwqe *wqe,
 		    u32 *lengthp, struct ipath_sge_state *ss)
 {
-	struct ipath_ibdev *dev = to_idev(qp->ibqp.device);
 	int user = to_ipd(qp->ibqp.pd)->user;
 	int i, j, ret;
 	struct ib_wc wc;
@@ -50,8 +49,7 @@ static int init_sge(struct ipath_qp *qp, struct ipath_rwqe *wqe,
 			continue;
 		/* Check LKEY */
 		if ((user && wqe->sg_list[i].lkey == 0) ||
-		    !ipath_lkey_ok(&dev->lk_table,
-				   j ? &ss->sg_list[j - 1] : &ss->sge,
+		    !ipath_lkey_ok(qp, j ? &ss->sg_list[j - 1] : &ss->sge,
 				   &wqe->sg_list[i], IB_ACCESS_LOCAL_WRITE))
 			goto bad_lkey;
 		*lengthp += wqe->sg_list[i].length;
@@ -343,7 +341,7 @@ int ipath_post_ud_send(struct ipath_qp *qp, struct ib_send_wr *wr)
 
 		if (wr->sg_list[i].length == 0)
 			continue;
-		if (!ipath_lkey_ok(&dev->lk_table, ss.num_sge ?
+		if (!ipath_lkey_ok(qp, ss.num_sge ?
 				   sg_list + ss.num_sge - 1 : &ss.sge,
 				   &wr->sg_list[i], 0)) {
 			ret = -EINVAL;

drivers/infiniband/hw/ipath/ipath_user_pages.c

@@ -89,6 +89,62 @@ static int __get_user_pages(unsigned long start_page, size_t num_pages,
 	return ret;
 }
 
+/**
+ * ipath_map_page - a safety wrapper around pci_map_page()
+ *
+ * A dma_addr of all 0's is interpreted by the chip as "disabled".
+ * Unfortunately, it can also be a valid dma_addr returned on some
+ * architectures.
+ *
+ * The powerpc iommu assigns dma_addrs in ascending order, so we don't
+ * have to bother with retries or mapping a dummy page to insure we
+ * don't just get the same mapping again.
+ *
+ * I'm sure we won't be so lucky with other iommu's, so FIXME.
+ */
+dma_addr_t ipath_map_page(struct pci_dev *hwdev, struct page *page,
+	unsigned long offset, size_t size, int direction)
+{
+	dma_addr_t phys;
+
+	phys = pci_map_page(hwdev, page, offset, size, direction);
+
+	if (phys == 0) {
+		pci_unmap_page(hwdev, phys, size, direction);
+		phys = pci_map_page(hwdev, page, offset, size, direction);
+		/*
+		 * FIXME: If we get 0 again, we should keep this page,
+		 * map another, then free the 0 page.
+		 */
+	}
+
+	return phys;
+}
+
+/**
+ * ipath_map_single - a safety wrapper around pci_map_single()
+ *
+ * Same idea as ipath_map_page().
+ */
+dma_addr_t ipath_map_single(struct pci_dev *hwdev, void *ptr, size_t size,
+	int direction)
+{
+	dma_addr_t phys;
+
+	phys = pci_map_single(hwdev, ptr, size, direction);
+
+	if (phys == 0) {
+		pci_unmap_single(hwdev, phys, size, direction);
+		phys = pci_map_single(hwdev, ptr, size, direction);
+		/*
+		 * FIXME: If we get 0 again, we should keep this page,
+		 * map another, then free the 0 page.
+		 */
+	}
+
+	return phys;
+}
+
 /**
  * ipath_get_user_pages - lock user pages into memory
  * @start_page: the start page

drivers/infiniband/hw/ipath/ipath_verbs.c

@@ -898,7 +898,8 @@ int ipath_get_counters(struct ipath_devdata *dd,
 		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);
+		ipath_snap_cntr(dd, dd->ipath_cregs->cr_badformatcnt) +
+		dd->ipath_rxfc_unsupvl_errs;
 	cntrs->port_rcv_remphys_errors =
 		ipath_snap_cntr(dd, dd->ipath_cregs->cr_rcvebpcnt);
 	cntrs->port_xmit_discards =
@@ -911,8 +912,10 @@ int ipath_get_counters(struct ipath_devdata *dd,
 		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 */
+	cntrs->local_link_integrity_errors =
+		(dd->ipath_flags & IPATH_GPIO_ERRINTRS) ?
+		dd->ipath_lli_errs : dd->ipath_lli_errors;
+	cntrs->excessive_buffer_overrun_errors = dd->ipath_overrun_thresh_errs;
 
 	ret = 0;
 
@@ -1199,6 +1202,7 @@ static struct ib_ah *ipath_create_ah(struct ib_pd *pd,
 	struct ipath_ah *ah;
 	struct ib_ah *ret;
 	struct ipath_ibdev *dev = to_idev(pd->device);
+	unsigned long flags;
 
 	/* A multicast address requires a GRH (see ch. 8.4.1). */
 	if (ah_attr->dlid >= IPATH_MULTICAST_LID_BASE &&
@@ -1225,16 +1229,16 @@ static struct ib_ah *ipath_create_ah(struct ib_pd *pd,
 		goto bail;
 	}
 
-	spin_lock(&dev->n_ahs_lock);
+	spin_lock_irqsave(&dev->n_ahs_lock, flags);
 	if (dev->n_ahs_allocated == ib_ipath_max_ahs) {
-		spin_unlock(&dev->n_ahs_lock);
+		spin_unlock_irqrestore(&dev->n_ahs_lock, flags);
 		kfree(ah);
 		ret = ERR_PTR(-ENOMEM);
 		goto bail;
 	}
 
 	dev->n_ahs_allocated++;
-	spin_unlock(&dev->n_ahs_lock);
+	spin_unlock_irqrestore(&dev->n_ahs_lock, flags);
 
 	/* ib_create_ah() will initialize ah->ibah. */
 	ah->attr = *ah_attr;
@@ -1255,10 +1259,11 @@ static int ipath_destroy_ah(struct ib_ah *ibah)
 {
 	struct ipath_ibdev *dev = to_idev(ibah->device);
 	struct ipath_ah *ah = to_iah(ibah);
+	unsigned long flags;
 
-	spin_lock(&dev->n_ahs_lock);
+	spin_lock_irqsave(&dev->n_ahs_lock, flags);
 	dev->n_ahs_allocated--;
-	spin_unlock(&dev->n_ahs_lock);
+	spin_unlock_irqrestore(&dev->n_ahs_lock, flags);
 
 	kfree(ah);
 
@@ -1380,11 +1385,13 @@ static int enable_timer(struct ipath_devdata *dd)
 	 * processing.
 	 */
 	if (dd->ipath_flags & IPATH_GPIO_INTR) {
+		u64 val;
 		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));
+		val = ipath_read_kreg64(dd, dd->ipath_kregs->kr_gpio_mask);
+		val |= (u64) (1 << IPATH_GPIO_PORT0_BIT);
+		ipath_write_kreg( dd, dd->ipath_kregs->kr_gpio_mask, val);
 	}
 
 	init_timer(&dd->verbs_timer);
@@ -1399,8 +1406,17 @@ static int enable_timer(struct ipath_devdata *dd)
 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);
+	if (dd->ipath_flags & IPATH_GPIO_INTR) {
+                u64 val;
+                /* Disable GPIO bit 2 interrupt */
+                val = ipath_read_kreg64(dd, dd->ipath_kregs->kr_gpio_mask);
+                val &= ~((u64) (1 << IPATH_GPIO_PORT0_BIT));
+                ipath_write_kreg( dd, dd->ipath_kregs->kr_gpio_mask, val);
+		/*
+		 * We might want to undo changes to debugportselect,
+		 * but how?
+		 */
+	}
 
 	del_timer_sync(&dd->verbs_timer);
 
@@ -1683,6 +1699,7 @@ static ssize_t show_stats(struct class_device *cdev, char *buf)
 		      "RC OTH NAKs %d\n"
 		      "RC timeouts %d\n"
 		      "RC RDMA dup %d\n"
+		      "RC stalls   %d\n"
 		      "piobuf wait %d\n"
 		      "no piobuf   %d\n"
 		      "PKT drops   %d\n"
@@ -1690,7 +1707,7 @@ static ssize_t show_stats(struct class_device *cdev, char *buf)
 		      dev->n_rc_resends, dev->n_rc_qacks, dev->n_rc_acks,
 		      dev->n_seq_naks, dev->n_rdma_seq, dev->n_rnr_naks,
 		      dev->n_other_naks, dev->n_timeouts,
-		      dev->n_rdma_dup_busy, dev->n_piowait,
+		      dev->n_rdma_dup_busy, dev->n_rc_stalls, dev->n_piowait,
 		      dev->n_no_piobuf, dev->n_pkt_drops, dev->n_wqe_errs);
 	for (i = 0; i < ARRAY_SIZE(dev->opstats); i++) {
 		const struct ipath_opcode_stats *si = &dev->opstats[i];

drivers/infiniband/hw/ipath/ipath_verbs.h

@@ -220,6 +220,7 @@ struct ipath_segarray {
 };
 
 struct ipath_mregion {
+	struct ib_pd *pd;	/* shares refcnt of ibmr.pd */
 	u64 user_base;		/* User's address for this region */
 	u64 iova;		/* IB start address of this region */
 	size_t length;
@@ -364,12 +365,14 @@ struct ipath_qp {
 	u8 r_min_rnr_timer;	/* retry timeout value for RNR NAKs */
 	u8 r_reuse_sge;		/* for UC receive errors */
 	u8 r_sge_inx;		/* current index into sg_list */
+	u8 r_wrid_valid;	/* r_wrid set but CQ entry not yet made */
 	u8 qp_access_flags;
 	u8 s_max_sge;		/* size of s_wq->sg_list */
 	u8 s_retry_cnt;		/* number of times to retry */
 	u8 s_rnr_retry_cnt;
 	u8 s_retry;		/* requester retry counter */
 	u8 s_rnr_retry;		/* requester RNR retry counter */
+	u8 s_wait_credit;	/* limit number of unacked packets sent */
 	u8 s_pkey_index;	/* PKEY index to use */
 	u8 timeout;		/* Timeout for this QP */
 	enum ib_mtu path_mtu;
@@ -393,6 +396,8 @@ struct ipath_qp {
 #define IPATH_S_BUSY		0
 #define IPATH_S_SIGNAL_REQ_WR	1
 
+#define IPATH_PSN_CREDIT	2048
+
 /*
  * Since struct ipath_swqe is not a fixed size, we can't simply index into
  * struct ipath_qp.s_wq.  This function does the array index computation.
@@ -521,6 +526,7 @@ struct ipath_ibdev {
 	u32 n_rnr_naks;
 	u32 n_other_naks;
 	u32 n_timeouts;
+	u32 n_rc_stalls;
 	u32 n_pkt_drops;
 	u32 n_vl15_dropped;
 	u32 n_wqe_errs;
@@ -634,6 +640,8 @@ struct ib_qp *ipath_create_qp(struct ib_pd *ibpd,
 
 int ipath_destroy_qp(struct ib_qp *ibqp);
 
+void ipath_error_qp(struct ipath_qp *qp, enum ib_wc_status err);
+
 int ipath_modify_qp(struct ib_qp *ibqp, struct ib_qp_attr *attr,
 		    int attr_mask, struct ib_udata *udata);
 
@@ -653,12 +661,6 @@ int ipath_verbs_send(struct ipath_devdata *dd, u32 hdrwords,
 
 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,
-		  u32 len, u64 vaddr, u32 rkey, int acc);
-
-int ipath_lkey_ok(struct ipath_lkey_table *rkt, struct ipath_sge *isge,
-		  struct ib_sge *sge, int acc);
-
 void ipath_copy_sge(struct ipath_sge_state *ss, void *data, u32 length);
 
 void ipath_skip_sge(struct ipath_sge_state *ss, u32 length);
@@ -683,10 +685,10 @@ int ipath_alloc_lkey(struct ipath_lkey_table *rkt,
 
 void ipath_free_lkey(struct ipath_lkey_table *rkt, u32 lkey);
 
-int ipath_lkey_ok(struct ipath_lkey_table *rkt, struct ipath_sge *isge,
+int ipath_lkey_ok(struct ipath_qp *qp, struct ipath_sge *isge,
 		  struct ib_sge *sge, int acc);
 
-int ipath_rkey_ok(struct ipath_ibdev *dev, struct ipath_sge_state *ss,
+int ipath_rkey_ok(struct ipath_qp *qp, struct ipath_sge_state *ss,
 		  u32 len, u64 vaddr, u32 rkey, int acc);
 
 int ipath_post_srq_receive(struct ib_srq *ibsrq, struct ib_recv_wr *wr,

drivers/infiniband/hw/ipath/ipath_wc_ppc64.c

@@ -38,13 +38,23 @@
 #include "ipath_kernel.h"
 
 /**
- * ipath_unordered_wc - indicate whether write combining is ordered
+ * ipath_enable_wc - enable write combining for MMIO writes to the device
+ * @dd: infinipath device
  *
- * 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.
+ * Nothing to do on PowerPC, so just return without error.
+ */
+int ipath_enable_wc(struct ipath_devdata *dd)
+{
+	return 0;
+}
+
+/**
+ * ipath_unordered_wc - indicate whether write combining is unordered
  *
+ * Because our performance depends on our ability to do write
+ * combining mmio writes in the most efficient way, we need to
+ * know if we are on a processor that may reorder stores when
+ * write combining.
  */
 int ipath_unordered_wc(void)
 {

drivers/infiniband/hw/ipath/ipath_wc_x86_64.c

@@ -123,6 +123,8 @@ int ipath_enable_wc(struct ipath_devdata *dd)
 			ipath_cdbg(VERBOSE, "Set mtrr for chip to WC, "
 				   "cookie is %d\n", cookie);
 			dd->ipath_wc_cookie = cookie;
+			dd->ipath_wc_base = (unsigned long) pioaddr;
+			dd->ipath_wc_len = (unsigned long) piolen;
 		}
 	}
 
@@ -136,9 +138,16 @@ int ipath_enable_wc(struct ipath_devdata *dd)
 void ipath_disable_wc(struct ipath_devdata *dd)
 {
 	if (dd->ipath_wc_cookie) {
+		int r;
 		ipath_cdbg(VERBOSE, "undoing WCCOMB on pio buffers\n");
-		mtrr_del(dd->ipath_wc_cookie, 0, 0);
-		dd->ipath_wc_cookie = 0;
+		r = mtrr_del(dd->ipath_wc_cookie, dd->ipath_wc_base,
+			     dd->ipath_wc_len);
+		if (r < 0)
+			dev_info(&dd->pcidev->dev,
+				 "mtrr_del(%lx, %lx, %lx) failed: %d\n",
+				 dd->ipath_wc_cookie, dd->ipath_wc_base,
+				 dd->ipath_wc_len, r);
+		dd->ipath_wc_cookie = 0; /* even on failure */
 	}
 }
 

drivers/infiniband/ulp/iser/Kconfig

 config INFINIBAND_ISER
-	tristate "ISCSI RDMA Protocol"
+	tristate "iSCSI Extensions for RDMA (iSER)"
 	depends on INFINIBAND && SCSI && INET
 	select SCSI_ISCSI_ATTRS
 	---help---
-	  Support for the ISCSI RDMA Protocol over InfiniBand.  This
-	  allows you to access storage devices that speak ISER/ISCSI
-	  over InfiniBand.
+	  Support for the iSCSI Extensions for RDMA (iSER) Protocol
+          over InfiniBand. This allows you to access storage devices
+          that speak iSCSI over iSER over InfiniBand.
 
-	  The ISER protocol is defined by IETF.
-	  See <http://www.ietf.org/>.
+	  The iSER protocol is defined by IETF.
+	  See <http://www.ietf.org/internet-drafts/draft-ietf-ips-iser-05.txt>
+	  and <http://www.infinibandta.org/members/spec/iser_annex_060418.pdf>

drivers/infiniband/ulp/iser/iscsi_iser.c

@@ -317,6 +317,8 @@ iscsi_iser_conn_destroy(struct iscsi_cls_conn *cls_conn)
 	struct iscsi_iser_conn *iser_conn = conn->dd_data;
 
 	iscsi_conn_teardown(cls_conn);
+	if (iser_conn->ib_conn)
+		iser_conn->ib_conn->iser_conn = NULL;
 	kfree(iser_conn);
 }
 

drivers/infiniband/ulp/iser/iscsi_iser.h

@@ -192,7 +192,7 @@ struct iser_regd_buf {
 
 struct iser_dto {
 	struct iscsi_iser_cmd_task *ctask;
-	struct iscsi_iser_conn     *conn;
+	struct iser_conn *ib_conn;
 	int                        notify_enable;
 
 	/* vector of registered buffers */
@@ -355,4 +355,11 @@ int  iser_post_send(struct iser_desc *tx_desc);
 
 int iser_conn_state_comp(struct iser_conn *ib_conn,
 			 enum iser_ib_conn_state comp);
+
+int iser_dma_map_task_data(struct iscsi_iser_cmd_task *iser_ctask,
+			    struct iser_data_buf       *data,
+			    enum   iser_data_dir       iser_dir,
+			    enum   dma_data_direction  dma_dir);
+
+void iser_dma_unmap_task_data(struct iscsi_iser_cmd_task *iser_ctask);
 #endif

drivers/infiniband/ulp/iser/iser_initiator.c

@@ -66,42 +66,6 @@ static void iser_dto_add_regd_buff(struct iser_dto *dto,
 	dto->regd_vector_len++;
 }
 
-static int iser_dma_map_task_data(struct iscsi_iser_cmd_task *iser_ctask,
-				  struct iser_data_buf       *data,
-				  enum   iser_data_dir       iser_dir,
-				  enum   dma_data_direction  dma_dir)
-{
-	struct device *dma_device;
-
-	iser_ctask->dir[iser_dir] = 1;
-	dma_device = iser_ctask->iser_conn->ib_conn->device->ib_device->dma_device;
-
-	data->dma_nents = dma_map_sg(dma_device, data->buf, data->size, dma_dir);
-	if (data->dma_nents == 0) {
-		iser_err("dma_map_sg failed!!!\n");
-		return -EINVAL;
-	}
-	return 0;
-}
-
-static void iser_dma_unmap_task_data(struct iscsi_iser_cmd_task *iser_ctask)
-{
-	struct device  *dma_device;
-	struct iser_data_buf *data;
-
-	dma_device = iser_ctask->iser_conn->ib_conn->device->ib_device->dma_device;
-
-	if (iser_ctask->dir[ISER_DIR_IN]) {
-		data = &iser_ctask->data[ISER_DIR_IN];
-		dma_unmap_sg(dma_device, data->buf, data->size, DMA_FROM_DEVICE);
-	}
-
-	if (iser_ctask->dir[ISER_DIR_OUT]) {
-		data = &iser_ctask->data[ISER_DIR_OUT];
-		dma_unmap_sg(dma_device, data->buf, data->size, DMA_TO_DEVICE);
-	}
-}
-
 /* Register user buffer memory and initialize passive rdma
  *  dto descriptor. Total data size is stored in
  *  iser_ctask->data[ISER_DIR_IN].data_len
@@ -249,7 +213,7 @@ static int iser_post_receive_control(struct iscsi_conn *conn)
 	}
 
 	recv_dto = &rx_desc->dto;
-	recv_dto->conn          = iser_conn;
+	recv_dto->ib_conn = iser_conn->ib_conn;
 	recv_dto->regd_vector_len = 0;
 
 	regd_hdr = &rx_desc->hdr_regd_buf;
@@ -296,7 +260,7 @@ static void iser_create_send_desc(struct iscsi_iser_conn *iser_conn,
 	regd_hdr->virt_addr  = tx_desc; /* == &tx_desc->iser_header */
 	regd_hdr->data_size  = ISER_TOTAL_HEADERS_LEN;
 
-	send_dto->conn          = iser_conn;
+	send_dto->ib_conn         = iser_conn->ib_conn;
 	send_dto->notify_enable   = 1;
 	send_dto->regd_vector_len = 0;
 
@@ -588,7 +552,7 @@ void iser_rcv_completion(struct iser_desc *rx_desc,
 			 unsigned long dto_xfer_len)
 {
 	struct iser_dto        *dto = &rx_desc->dto;
-	struct iscsi_iser_conn *conn = dto->conn;
+	struct iscsi_iser_conn *conn = dto->ib_conn->iser_conn;
 	struct iscsi_session *session = conn->iscsi_conn->session;
 	struct iscsi_cmd_task *ctask;
 	struct iscsi_iser_cmd_task *iser_ctask;
@@ -641,7 +605,8 @@ void iser_rcv_completion(struct iser_desc *rx_desc,
 void iser_snd_completion(struct iser_desc *tx_desc)
 {
 	struct iser_dto        *dto = &tx_desc->dto;
-	struct iscsi_iser_conn *iser_conn = dto->conn;
+	struct iser_conn       *ib_conn = dto->ib_conn;
+	struct iscsi_iser_conn *iser_conn = ib_conn->iser_conn;
 	struct iscsi_conn      *conn = iser_conn->iscsi_conn;
 	struct iscsi_mgmt_task *mtask;
 
@@ -652,7 +617,7 @@ void iser_snd_completion(struct iser_desc *tx_desc)
 	if (tx_desc->type == ISCSI_TX_DATAOUT)
 		kmem_cache_free(ig.desc_cache, tx_desc);
 
-	atomic_dec(&iser_conn->ib_conn->post_send_buf_count);
+	atomic_dec(&ib_conn->post_send_buf_count);
 
 	write_lock(conn->recv_lock);
 	if (conn->suspend_tx) {
@@ -698,14 +663,19 @@ void iser_ctask_rdma_init(struct iscsi_iser_cmd_task *iser_ctask)
 void iser_ctask_rdma_finalize(struct iscsi_iser_cmd_task *iser_ctask)
 {
 	int deferred;
+	int is_rdma_aligned = 1;
 
 	/* if we were reading, copy back to unaligned sglist,
 	 * anyway dma_unmap and free the copy
 	 */
-	if (iser_ctask->data_copy[ISER_DIR_IN].copy_buf != NULL)
+	if (iser_ctask->data_copy[ISER_DIR_IN].copy_buf != NULL) {
+		is_rdma_aligned = 0;
 		iser_finalize_rdma_unaligned_sg(iser_ctask, ISER_DIR_IN);
-	if (iser_ctask->data_copy[ISER_DIR_OUT].copy_buf != NULL)
+	}
+	if (iser_ctask->data_copy[ISER_DIR_OUT].copy_buf != NULL) {
+		is_rdma_aligned = 0;
 		iser_finalize_rdma_unaligned_sg(iser_ctask, ISER_DIR_OUT);
+	}
 
 	if (iser_ctask->dir[ISER_DIR_IN]) {
 		deferred = iser_regd_buff_release
@@ -725,7 +695,9 @@ void iser_ctask_rdma_finalize(struct iscsi_iser_cmd_task *iser_ctask)
 		}
 	}
 
-	iser_dma_unmap_task_data(iser_ctask);
+       /* if the data was unaligned, it was already unmapped and then copied */
+       if (is_rdma_aligned)
+		iser_dma_unmap_task_data(iser_ctask);
 }
 
 void iser_dto_buffs_release(struct iser_dto *dto)

drivers/infiniband/ulp/iser/iser_memory.c

@@ -369,6 +369,44 @@ static void iser_page_vec_build(struct iser_data_buf *data,
 	}
 }
 
+int iser_dma_map_task_data(struct iscsi_iser_cmd_task *iser_ctask,
+			    struct iser_data_buf       *data,
+			    enum   iser_data_dir       iser_dir,
+			    enum   dma_data_direction  dma_dir)
+{
+	struct device *dma_device;
+
+	iser_ctask->dir[iser_dir] = 1;
+	dma_device =
+		iser_ctask->iser_conn->ib_conn->device->ib_device->dma_device;
+
+	data->dma_nents = dma_map_sg(dma_device, data->buf, data->size, dma_dir);
+	if (data->dma_nents == 0) {
+		iser_err("dma_map_sg failed!!!\n");
+		return -EINVAL;
+	}
+	return 0;
+}
+
+void iser_dma_unmap_task_data(struct iscsi_iser_cmd_task *iser_ctask)
+{
+	struct device  *dma_device;
+	struct iser_data_buf *data;
+
+	dma_device =
+		iser_ctask->iser_conn->ib_conn->device->ib_device->dma_device;
+
+	if (iser_ctask->dir[ISER_DIR_IN]) {
+		data = &iser_ctask->data[ISER_DIR_IN];
+		dma_unmap_sg(dma_device, data->buf, data->size, DMA_FROM_DEVICE);
+	}
+
+	if (iser_ctask->dir[ISER_DIR_OUT]) {
+		data = &iser_ctask->data[ISER_DIR_OUT];
+		dma_unmap_sg(dma_device, data->buf, data->size, DMA_TO_DEVICE);
+	}
+}
+
 /**
  * iser_reg_rdma_mem - Registers memory intended for RDMA,
  * obtaining rkey and va
@@ -394,6 +432,10 @@ int iser_reg_rdma_mem(struct iscsi_iser_cmd_task *iser_ctask,
 		iser_err("rdma alignment violation %d/%d aligned\n",
 			 aligned_len, mem->size);
 		iser_data_buf_dump(mem);
+
+		/* unmap the command data before accessing it */
+		iser_dma_unmap_task_data(iser_ctask);
+
 		/* allocate copy buf, if we are writing, copy the */
 		/* unaligned scatterlist, dma map the copy        */
 		if (iser_start_rdma_unaligned_sg(iser_ctask, cmd_dir) != 0)

drivers/infiniband/ulp/iser/iser_verbs.c

@@ -571,6 +571,8 @@ void iser_conn_release(struct iser_conn *ib_conn)
 	/* on EVENT_ADDR_ERROR there's no device yet for this conn */
 	if (device != NULL)
 		iser_device_try_release(device);
+	if (ib_conn->iser_conn)
+		ib_conn->iser_conn->ib_conn = NULL;
 	kfree(ib_conn);
 }
 
@@ -694,7 +696,7 @@ int iser_post_recv(struct iser_desc *rx_desc)
 	struct iser_dto   *recv_dto = &rx_desc->dto;
 
 	/* Retrieve conn */
-	ib_conn = recv_dto->conn->ib_conn;
+	ib_conn = recv_dto->ib_conn;
 
 	iser_dto_to_iov(recv_dto, iov, 2);
 
@@ -727,7 +729,7 @@ int iser_post_send(struct iser_desc *tx_desc)
 	struct iser_conn  *ib_conn;
 	struct iser_dto   *dto = &tx_desc->dto;
 
-	ib_conn = dto->conn->ib_conn;
+	ib_conn = dto->ib_conn;
 
 	iser_dto_to_iov(dto, iov, MAX_REGD_BUF_VECTOR_LEN);
 
@@ -774,7 +776,7 @@ static void iser_comp_error_worker(void *data)
 static void iser_handle_comp_error(struct iser_desc *desc)
 {
 	struct iser_dto  *dto     = &desc->dto;
-	struct iser_conn *ib_conn = dto->conn->ib_conn;
+	struct iser_conn *ib_conn = dto->ib_conn;
 
 	iser_dto_buffs_release(dto);