4 files changed, 320 insertions, 187 deletions

diff --git a/drivers/ntb/ntb_hw.c b/drivers/ntb/ntb_hw.c
index 1cb6e51e6bd..372e08c4ffe 100644
--- a/drivers/ntb/ntb_hw.c
+++ b/drivers/ntb/ntb_hw.c
@@ -91,7 +91,7 @@ static struct dentry *debugfs_dir;
 /* Translate memory window 0,1 to BAR 2,4 */
 #define MW_TO_BAR(mw)	(mw * NTB_MAX_NUM_MW + 2)
 
-static DEFINE_PCI_DEVICE_TABLE(ntb_pci_tbl) = {
+static const struct pci_device_id ntb_pci_tbl[] = {
 	{PCI_VDEVICE(INTEL, PCI_DEVICE_ID_INTEL_NTB_B2B_BWD)},
 	{PCI_VDEVICE(INTEL, PCI_DEVICE_ID_INTEL_NTB_B2B_JSF)},
 	{PCI_VDEVICE(INTEL, PCI_DEVICE_ID_INTEL_NTB_B2B_SNB)},
@@ -120,7 +120,8 @@ MODULE_DEVICE_TABLE(pci, ntb_pci_tbl);
  * RETURNS: An appropriate -ERRNO error value on error, or zero for success.
  */
 int ntb_register_event_callback(struct ntb_device *ndev,
-			    void (*func)(void *handle, enum ntb_hw_event event))
+				void (*func)(void *handle,
+					     enum ntb_hw_event event))
 {
 	if (ndev->event_cb)
 		return -EINVAL;
@@ -141,6 +142,24 @@ void ntb_unregister_event_callback(struct ntb_device *ndev)
 	ndev->event_cb = NULL;
 }
 
+static void ntb_irq_work(unsigned long data)
+{
+	struct ntb_db_cb *db_cb = (struct ntb_db_cb *)data;
+	int rc;
+
+	rc = db_cb->callback(db_cb->data, db_cb->db_num);
+	if (rc)
+		tasklet_schedule(&db_cb->irq_work);
+	else {
+		struct ntb_device *ndev = db_cb->ndev;
+		unsigned long mask;
+
+		mask = readw(ndev->reg_ofs.ldb_mask);
+		clear_bit(db_cb->db_num * ndev->bits_per_vector, &mask);
+		writew(mask, ndev->reg_ofs.ldb_mask);
+	}
+}
+
 /**
  * ntb_register_db_callback() - register a callback for doorbell interrupt
  * @ndev: pointer to ntb_device instance
@@ -155,7 +174,7 @@ void ntb_unregister_event_callback(struct ntb_device *ndev)
  * RETURNS: An appropriate -ERRNO error value on error, or zero for success.
  */
 int ntb_register_db_callback(struct ntb_device *ndev, unsigned int idx,
-			     void *data, void (*func)(void *data, int db_num))
+			     void *data, int (*func)(void *data, int db_num))
 {
 	unsigned long mask;
 
@@ -166,6 +185,10 @@ int ntb_register_db_callback(struct ntb_device *ndev, unsigned int idx,
 
 	ndev->db_cb[idx].callback = func;
 	ndev->db_cb[idx].data = data;
+	ndev->db_cb[idx].ndev = ndev;
+
+	tasklet_init(&ndev->db_cb[idx].irq_work, ntb_irq_work,
+		     (unsigned long) &ndev->db_cb[idx]);
 
 	/* unmask interrupt */
 	mask = readw(ndev->reg_ofs.ldb_mask);
@@ -194,6 +217,8 @@ void ntb_unregister_db_callback(struct ntb_device *ndev, unsigned int idx)
 	set_bit(idx * ndev->bits_per_vector, &mask);
 	writew(mask, ndev->reg_ofs.ldb_mask);
 
+	tasklet_disable(&ndev->db_cb[idx].irq_work);
+
 	ndev->db_cb[idx].callback = NULL;
 }
 
@@ -678,6 +703,7 @@ static int ntb_xeon_setup(struct ntb_device *ndev)
 				return -EINVAL;
 
 			ndev->limits.max_mw = SNB_ERRATA_MAX_MW;
+			ndev->limits.max_db_bits = SNB_MAX_DB_BITS;
 			ndev->reg_ofs.spad_write = ndev->mw[1].vbase +
 						   SNB_SPAD_OFFSET;
 			ndev->reg_ofs.rdb = ndev->mw[1].vbase +
@@ -688,8 +714,21 @@ static int ntb_xeon_setup(struct ntb_device *ndev)
 			 */
 			writeq(ndev->mw[1].bar_sz + 0x1000, ndev->reg_base +
 			       SNB_PBAR4LMT_OFFSET);
+			/* HW errata on the Limit registers.  They can only be
+			 * written when the base register is 4GB aligned and
+			 * < 32bit.  This should already be the case based on
+			 * the driver defaults, but write the Limit registers
+			 * first just in case.
+			 */
 		} else {
 			ndev->limits.max_mw = SNB_MAX_MW;
+
+			/* HW Errata on bit 14 of b2bdoorbell register.  Writes
+			 * will not be mirrored to the remote system.  Shrink
+			 * the number of bits by one, since bit 14 is the last
+			 * bit.
+			 */
+			ndev->limits.max_db_bits = SNB_MAX_DB_BITS - 1;
 			ndev->reg_ofs.spad_write = ndev->reg_base +
 						   SNB_B2B_SPAD_OFFSET;
 			ndev->reg_ofs.rdb = ndev->reg_base +
@@ -699,6 +738,12 @@ static int ntb_xeon_setup(struct ntb_device *ndev)
 			 * something silly
 			 */
 			writeq(0, ndev->reg_base + SNB_PBAR4LMT_OFFSET);
+			/* HW errata on the Limit registers.  They can only be
+			 * written when the base register is 4GB aligned and
+			 * < 32bit.  This should already be the case based on
+			 * the driver defaults, but write the Limit registers
+			 * first just in case.
+			 */
 		}
 
 		/* The Xeon errata workaround requires setting SBAR Base
@@ -741,7 +786,7 @@ static int ntb_xeon_setup(struct ntb_device *ndev)
 				/* B2B_XLAT_OFFSET is a 64bit register, but can
 				 * only take 32bit writes
 				 */
-				writel(SNB_MBAR01_DSD_ADDR & 0xffffffff,
+				writel(SNB_MBAR01_USD_ADDR & 0xffffffff,
 				       ndev->reg_base + SNB_B2B_XLAT_OFFSETL);
 				writel(SNB_MBAR01_USD_ADDR >> 32,
 				       ndev->reg_base + SNB_B2B_XLAT_OFFSETU);
@@ -759,7 +804,7 @@ static int ntb_xeon_setup(struct ntb_device *ndev)
 		ndev->conn_type = NTB_CONN_RP;
 
 		if (xeon_errata_workaround) {
-			dev_err(&ndev->pdev->dev, 
+			dev_err(&ndev->pdev->dev,
 				"NTB-RP disabled due to hardware errata.  To disregard this warning and potentially lock-up the system, add the parameter 'xeon_errata_workaround=0'.\n");
 			return -EINVAL;
 		}
@@ -769,6 +814,7 @@ static int ntb_xeon_setup(struct ntb_device *ndev)
 		 * have an equal amount.
 		 */
 		ndev->limits.max_spads = SNB_MAX_COMPAT_SPADS / 2;
+		ndev->limits.max_db_bits = SNB_MAX_DB_BITS;
 		/* Note: The SDOORBELL is the cause of the errata.  You REALLY
 		 * don't want to touch it.
 		 */
@@ -793,6 +839,7 @@ static int ntb_xeon_setup(struct ntb_device *ndev)
 		 * have an equal amount.
 		 */
 		ndev->limits.max_spads = SNB_MAX_COMPAT_SPADS / 2;
+		ndev->limits.max_db_bits = SNB_MAX_DB_BITS;
 		ndev->reg_ofs.rdb = ndev->reg_base + SNB_PDOORBELL_OFFSET;
 		ndev->reg_ofs.ldb = ndev->reg_base + SNB_SDOORBELL_OFFSET;
 		ndev->reg_ofs.ldb_mask = ndev->reg_base + SNB_SDBMSK_OFFSET;
@@ -819,7 +866,6 @@ static int ntb_xeon_setup(struct ntb_device *ndev)
 	ndev->reg_ofs.lnk_stat = ndev->reg_base + SNB_SLINK_STATUS_OFFSET;
 	ndev->reg_ofs.spci_cmd = ndev->reg_base + SNB_PCICMD_OFFSET;
 
-	ndev->limits.max_db_bits = SNB_MAX_DB_BITS;
 	ndev->limits.msix_cnt = SNB_MSIX_CNT;
 	ndev->bits_per_vector = SNB_DB_BITS_PER_VEC;
 
@@ -934,12 +980,16 @@ static irqreturn_t bwd_callback_msix_irq(int irq, void *data)
 {
 	struct ntb_db_cb *db_cb = data;
 	struct ntb_device *ndev = db_cb->ndev;
+	unsigned long mask;
 
 	dev_dbg(&ndev->pdev->dev, "MSI-X irq %d received for DB %d\n", irq,
 		db_cb->db_num);
 
-	if (db_cb->callback)
-		db_cb->callback(db_cb->data, db_cb->db_num);
+	mask = readw(ndev->reg_ofs.ldb_mask);
+	set_bit(db_cb->db_num * ndev->bits_per_vector, &mask);
+	writew(mask, ndev->reg_ofs.ldb_mask);
+
+	tasklet_schedule(&db_cb->irq_work);
 
 	/* No need to check for the specific HB irq, any interrupt means
 	 * we're connected.
@@ -955,12 +1005,16 @@ static irqreturn_t xeon_callback_msix_irq(int irq, void *data)
 {
 	struct ntb_db_cb *db_cb = data;
 	struct ntb_device *ndev = db_cb->ndev;
+	unsigned long mask;
 
 	dev_dbg(&ndev->pdev->dev, "MSI-X irq %d received for DB %d\n", irq,
 		db_cb->db_num);
 
-	if (db_cb->callback)
-		db_cb->callback(db_cb->data, db_cb->db_num);
+	mask = readw(ndev->reg_ofs.ldb_mask);
+	set_bit(db_cb->db_num * ndev->bits_per_vector, &mask);
+	writew(mask, ndev->reg_ofs.ldb_mask);
+
+	tasklet_schedule(&db_cb->irq_work);
 
 	/* On Sandybridge, there are 16 bits in the interrupt register
 	 * but only 4 vectors.  So, 5 bits are assigned to the first 3
@@ -986,7 +1040,7 @@ static irqreturn_t xeon_event_msix_irq(int irq, void *dev)
 		dev_err(&ndev->pdev->dev, "Error determining link status\n");
 
 	/* bit 15 is always the link bit */
-	writew(1 << ndev->limits.max_db_bits, ndev->reg_ofs.ldb);
+	writew(1 << SNB_LINK_DB, ndev->reg_ofs.ldb);
 
 	return IRQ_HANDLED;
 }
@@ -1026,25 +1080,104 @@ static irqreturn_t ntb_interrupt(int irq, void *dev)
 	return IRQ_HANDLED;
 }
 
-static int ntb_setup_msix(struct ntb_device *ndev)
+static int ntb_setup_snb_msix(struct ntb_device *ndev, int msix_entries)
 {
 	struct pci_dev *pdev = ndev->pdev;
 	struct msix_entry *msix;
-	int msix_entries;
 	int rc, i;
-	u16 val;
 
-	if (!pdev->msix_cap) {
-		rc = -EIO;
-		goto err;
+	if (msix_entries < ndev->limits.msix_cnt)
+		return -ENOSPC;
+
+	rc = pci_enable_msix_exact(pdev, ndev->msix_entries, msix_entries);
+	if (rc < 0)
+		return rc;
+
+	for (i = 0; i < msix_entries; i++) {
+		msix = &ndev->msix_entries[i];
+		WARN_ON(!msix->vector);
+
+		if (i == msix_entries - 1) {
+			rc = request_irq(msix->vector,
+					 xeon_event_msix_irq, 0,
+					 "ntb-event-msix", ndev);
+			if (rc)
+				goto err;
+		} else {
+			rc = request_irq(msix->vector,
+					 xeon_callback_msix_irq, 0,
+					 "ntb-callback-msix",
+					 &ndev->db_cb[i]);
+			if (rc)
+				goto err;
+		}
 	}
 
-	rc = pci_read_config_word(pdev, pdev->msix_cap + PCI_MSIX_FLAGS, &val);
-	if (rc)
-		goto err;
+	ndev->num_msix = msix_entries;
+	ndev->max_cbs = msix_entries - 1;
+
+	return 0;
+
+err:
+	while (--i >= 0) {
+		/* Code never reaches here for entry nr 'ndev->num_msix - 1' */
+		msix = &ndev->msix_entries[i];
+		free_irq(msix->vector, &ndev->db_cb[i]);
+	}
 
-	msix_entries = msix_table_size(val);
-	if (msix_entries > ndev->limits.msix_cnt) {
+	pci_disable_msix(pdev);
+	ndev->num_msix = 0;
+
+	return rc;
+}
+
+static int ntb_setup_bwd_msix(struct ntb_device *ndev, int msix_entries)
+{
+	struct pci_dev *pdev = ndev->pdev;
+	struct msix_entry *msix;
+	int rc, i;
+
+	msix_entries = pci_enable_msix_range(pdev, ndev->msix_entries,
+					     1, msix_entries);
+	if (msix_entries < 0)
+		return msix_entries;
+
+	for (i = 0; i < msix_entries; i++) {
+		msix = &ndev->msix_entries[i];
+		WARN_ON(!msix->vector);
+
+		rc = request_irq(msix->vector, bwd_callback_msix_irq, 0,
+				 "ntb-callback-msix", &ndev->db_cb[i]);
+		if (rc)
+			goto err;
+	}
+
+	ndev->num_msix = msix_entries;
+	ndev->max_cbs = msix_entries;
+
+	return 0;
+
+err:
+	while (--i >= 0)
+		free_irq(msix->vector, &ndev->db_cb[i]);
+
+	pci_disable_msix(pdev);
+	ndev->num_msix = 0;
+
+	return rc;
+}
+
+static int ntb_setup_msix(struct ntb_device *ndev)
+{
+	struct pci_dev *pdev = ndev->pdev;
+	int msix_entries;
+	int rc, i;
+
+	msix_entries = pci_msix_vec_count(pdev);
+	if (msix_entries < 0) {
+		rc = msix_entries;
+		goto err;
+	} else if (msix_entries > ndev->limits.msix_cnt) {
 		rc = -EINVAL;
 		goto err;
 	}
@@ -1059,74 +1192,19 @@ static int ntb_setup_msix(struct ntb_device *ndev)
 	for (i = 0; i < msix_entries; i++)
 		ndev->msix_entries[i].entry = i;
 
-	rc = pci_enable_msix(pdev, ndev->msix_entries, msix_entries);
-	if (rc < 0)
-		goto err1;
-	if (rc > 0) {
-		/* On SNB, the link interrupt is always tied to 4th vector.  If
-		 * we can't get all 4, then we can't use MSI-X.
-		 */
-		if (ndev->hw_type != BWD_HW) {
-			rc = -EIO;
-			goto err1;
-		}
-
-		dev_warn(&pdev->dev,
-			 "Only %d MSI-X vectors.  Limiting the number of queues to that number.\n",
-			 rc);
-		msix_entries = rc;
-	}
-
-	for (i = 0; i < msix_entries; i++) {
-		msix = &ndev->msix_entries[i];
-		WARN_ON(!msix->vector);
-
-		/* Use the last MSI-X vector for Link status */
-		if (ndev->hw_type == BWD_HW) {
-			rc = request_irq(msix->vector, bwd_callback_msix_irq, 0,
-					 "ntb-callback-msix", &ndev->db_cb[i]);
-			if (rc)
-				goto err2;
-		} else {
-			if (i == msix_entries - 1) {
-				rc = request_irq(msix->vector,
-						 xeon_event_msix_irq, 0,
-						 "ntb-event-msix", ndev);
-				if (rc)
-					goto err2;
-			} else {
-				rc = request_irq(msix->vector,
-						 xeon_callback_msix_irq, 0,
-						 "ntb-callback-msix",
-						 &ndev->db_cb[i]);
-				if (rc)
-					goto err2;
-			}
-		}
-	}
-
-	ndev->num_msix = msix_entries;
 	if (ndev->hw_type == BWD_HW)
-		ndev->max_cbs = msix_entries;
+		rc = ntb_setup_bwd_msix(ndev, msix_entries);
 	else
-		ndev->max_cbs = msix_entries - 1;
+		rc = ntb_setup_snb_msix(ndev, msix_entries);
+	if (rc)
+		goto err1;
 
 	return 0;
 
-err2:
-	while (--i >= 0) {
-		msix = &ndev->msix_entries[i];
-		if (ndev->hw_type != BWD_HW && i == ndev->num_msix - 1)
-			free_irq(msix->vector, ndev);
-		else
-			free_irq(msix->vector, &ndev->db_cb[i]);
-	}
-	pci_disable_msix(pdev);
 err1:
 	kfree(ndev->msix_entries);
-	dev_err(&pdev->dev, "Error allocating MSI-X interrupt\n");
 err:
-	ndev->num_msix = 0;
+	dev_err(&pdev->dev, "Error allocating MSI-X interrupt\n");
 	return rc;
 }
 
@@ -1176,9 +1254,10 @@ static int ntb_setup_interrupts(struct ntb_device *ndev)
 	 */
 	if (ndev->hw_type == BWD_HW)
 		writeq(~0, ndev->reg_ofs.ldb_mask);
-	else
-		writew(~(1 << ndev->limits.max_db_bits),
-		       ndev->reg_ofs.ldb_mask);
+	else {
+		u16 var = 1 << SNB_LINK_DB;
+		writew(~var, ndev->reg_ofs.ldb_mask);
+	}
 
 	rc = ntb_setup_msix(ndev);
 	if (!rc)
@@ -1223,6 +1302,7 @@ static void ntb_free_interrupts(struct ntb_device *ndev)
 				free_irq(msix->vector, &ndev->db_cb[i]);
 		}
 		pci_disable_msix(pdev);
+		kfree(ndev->msix_entries);
 	} else {
 		free_irq(pdev->irq, ndev);
 
@@ -1286,6 +1366,39 @@ static void ntb_free_debugfs(struct ntb_device *ndev)
 	}
 }
 
+static void ntb_hw_link_up(struct ntb_device *ndev)
+{
+	if (ndev->conn_type == NTB_CONN_TRANSPARENT)
+		ntb_link_event(ndev, NTB_LINK_UP);
+	else {
+		u32 ntb_cntl;
+
+		/* Let's bring the NTB link up */
+		ntb_cntl = readl(ndev->reg_ofs.lnk_cntl);
+		ntb_cntl &= ~(NTB_CNTL_LINK_DISABLE | NTB_CNTL_CFG_LOCK);
+		ntb_cntl |= NTB_CNTL_P2S_BAR23_SNOOP | NTB_CNTL_S2P_BAR23_SNOOP;
+		ntb_cntl |= NTB_CNTL_P2S_BAR45_SNOOP | NTB_CNTL_S2P_BAR45_SNOOP;
+		writel(ntb_cntl, ndev->reg_ofs.lnk_cntl);
+	}
+}
+
+static void ntb_hw_link_down(struct ntb_device *ndev)
+{
+	u32 ntb_cntl;
+
+	if (ndev->conn_type == NTB_CONN_TRANSPARENT) {
+		ntb_link_event(ndev, NTB_LINK_DOWN);
+		return;
+	}
+
+	/* Bring NTB link down */
+	ntb_cntl = readl(ndev->reg_ofs.lnk_cntl);
+	ntb_cntl &= ~(NTB_CNTL_P2S_BAR23_SNOOP | NTB_CNTL_S2P_BAR23_SNOOP);
+	ntb_cntl &= ~(NTB_CNTL_P2S_BAR45_SNOOP | NTB_CNTL_S2P_BAR45_SNOOP);
+	ntb_cntl |= NTB_CNTL_LINK_DISABLE | NTB_CNTL_CFG_LOCK;
+	writel(ntb_cntl, ndev->reg_ofs.lnk_cntl);
+}
+
 static int ntb_pci_probe(struct pci_dev *pdev, const struct pci_device_id *id)
 {
 	struct ntb_device *ndev;
@@ -1374,9 +1487,7 @@ static int ntb_pci_probe(struct pci_dev *pdev, const struct pci_device_id *id)
 	if (rc)
 		goto err6;
 
-	/* Let's bring the NTB link up */
-	writel(NTB_CNTL_BAR23_SNOOP | NTB_CNTL_BAR45_SNOOP,
-	       ndev->reg_ofs.lnk_cntl);
+	ntb_hw_link_up(ndev);
 
 	return 0;
 
@@ -1406,12 +1517,8 @@ static void ntb_pci_remove(struct pci_dev *pdev)
 {
 	struct ntb_device *ndev = pci_get_drvdata(pdev);
 	int i;
-	u32 ntb_cntl;
 
-	/* Bring NTB link down */
-	ntb_cntl = readl(ndev->reg_ofs.lnk_cntl);
-	ntb_cntl |= NTB_CNTL_LINK_DISABLE;
-	writel(ntb_cntl, ndev->reg_ofs.lnk_cntl);
+	ntb_hw_link_down(ndev);
 
 	ntb_transport_free(ndev->ntb_transport);
 
diff --git a/drivers/ntb/ntb_hw.h b/drivers/ntb/ntb_hw.h
index 0a31cedae7d..465517b7393 100644
--- a/drivers/ntb/ntb_hw.h
+++ b/drivers/ntb/ntb_hw.h
@@ -45,6 +45,7 @@
  * Contact Information:
  * Jon Mason <jon.mason@intel.com>
  */
+#include <linux/ntb.h>
 
 #define PCI_DEVICE_ID_INTEL_NTB_B2B_JSF		0x3725
 #define PCI_DEVICE_ID_INTEL_NTB_PS_JSF		0x3726
@@ -60,8 +61,6 @@
 #define PCI_DEVICE_ID_INTEL_NTB_SS_HSX		0x2F0F
 #define PCI_DEVICE_ID_INTEL_NTB_B2B_BWD		0x0C4E
 
-#define msix_table_size(control)	((control & PCI_MSIX_FLAGS_QSIZE)+1)
-
 #ifndef readq
 static inline u64 readq(void __iomem *addr)
 {
@@ -83,9 +82,6 @@ static inline void writeq(u64 val, void __iomem *addr)
 #define NTB_BAR_MASK		((1 << NTB_BAR_MMIO) | (1 << NTB_BAR_23) |\
 				 (1 << NTB_BAR_45))
 
-#define NTB_LINK_DOWN		0
-#define NTB_LINK_UP		1
-
 #define NTB_HB_TIMEOUT		msecs_to_jiffies(1000)
 
 #define NTB_MAX_NUM_MW		2
@@ -106,10 +102,11 @@ struct ntb_mw {
 };
 
 struct ntb_db_cb {
-	void (*callback) (void *data, int db_num);
+	int (*callback)(void *data, int db_num);
 	unsigned int db_num;
 	void *data;
 	struct ntb_device *ndev;
+	struct tasklet_struct irq_work;
 };
 
 struct ntb_device {
@@ -228,11 +225,11 @@ struct ntb_device *ntb_register_transport(struct pci_dev *pdev,
 void ntb_unregister_transport(struct ntb_device *ndev);
 void ntb_set_mw_addr(struct ntb_device *ndev, unsigned int mw, u64 addr);
 int ntb_register_db_callback(struct ntb_device *ndev, unsigned int idx,
-			     void *data, void (*db_cb_func) (void *data,
-							     int db_num));
+			     void *data, int (*db_cb_func)(void *data,
+							   int db_num));
 void ntb_unregister_db_callback(struct ntb_device *ndev, unsigned int idx);
 int ntb_register_event_callback(struct ntb_device *ndev,
-				void (*event_cb_func) (void *handle,
+				void (*event_cb_func)(void *handle,
 						      enum ntb_hw_event event));
 void ntb_unregister_event_callback(struct ntb_device *ndev);
 int ntb_get_max_spads(struct ntb_device *ndev);
diff --git a/drivers/ntb/ntb_regs.h b/drivers/ntb/ntb_regs.h
index aa4bdd393c5..9774506419d 100644
--- a/drivers/ntb/ntb_regs.h
+++ b/drivers/ntb/ntb_regs.h
@@ -55,6 +55,7 @@
 #define SNB_MAX_COMPAT_SPADS	16
 /* Reserve the uppermost bit for link interrupt */
 #define SNB_MAX_DB_BITS		15
+#define SNB_LINK_DB		15
 #define SNB_DB_BITS_PER_VEC	5
 #define SNB_MAX_MW		2
 #define SNB_ERRATA_MAX_MW	1
@@ -75,9 +76,6 @@
 #define SNB_SBAR2XLAT_OFFSET	0x0030
 #define SNB_SBAR4XLAT_OFFSET	0x0038
 #define SNB_SBAR0BASE_OFFSET	0x0040
-#define SNB_SBAR0BASE_OFFSET	0x0040
-#define SNB_SBAR2BASE_OFFSET	0x0048
-#define SNB_SBAR4BASE_OFFSET	0x0050
 #define SNB_SBAR2BASE_OFFSET	0x0048
 #define SNB_SBAR4BASE_OFFSET	0x0050
 #define SNB_NTBCNTL_OFFSET	0x0058
@@ -145,11 +143,13 @@
 #define BWD_LTSSMSTATEJMP_FORCEDETECT	(1 << 2)
 #define BWD_IBIST_ERR_OFLOW	0x7FFF7FFF
 
-#define NTB_CNTL_CFG_LOCK	(1 << 0)
-#define NTB_CNTL_LINK_DISABLE	(1 << 1)
-#define NTB_CNTL_BAR23_SNOOP	(1 << 2)
-#define NTB_CNTL_BAR45_SNOOP	(1 << 6)
-#define BWD_CNTL_LINK_DOWN	(1 << 16)
+#define NTB_CNTL_CFG_LOCK		(1 << 0)
+#define NTB_CNTL_LINK_DISABLE		(1 << 1)
+#define NTB_CNTL_S2P_BAR23_SNOOP	(1 << 2)
+#define NTB_CNTL_P2S_BAR23_SNOOP	(1 << 4)
+#define NTB_CNTL_S2P_BAR45_SNOOP	(1 << 6)
+#define NTB_CNTL_P2S_BAR45_SNOOP	(1 << 8)
+#define BWD_CNTL_LINK_DOWN		(1 << 16)
 
 #define NTB_PPD_OFFSET		0x00D4
 #define SNB_PPD_CONN_TYPE	0x0003
diff --git a/drivers/ntb/ntb_transport.c b/drivers/ntb/ntb_transport.c
index 12a9e83c008..9dd63b82202 100644
--- a/drivers/ntb/ntb_transport.c
+++ b/drivers/ntb/ntb_transport.c
@@ -56,7 +56,6 @@
 #include <linux/pci.h>
 #include <linux/slab.h>
 #include <linux/types.h>
-#include <linux/ntb.h>
 #include "ntb_hw.h"
 
 #define NTB_TRANSPORT_VERSION	3
@@ -107,8 +106,8 @@ struct ntb_transport_qp {
 	struct ntb_rx_info __iomem *rx_info;
 	struct ntb_rx_info *remote_rx_info;
 
-	void (*tx_handler) (struct ntb_transport_qp *qp, void *qp_data,
-			    void *data, int len);
+	void (*tx_handler)(struct ntb_transport_qp *qp, void *qp_data,
+			   void *data, int len);
 	struct list_head tx_free_q;
 	spinlock_t ntb_tx_free_q_lock;
 	void __iomem *tx_mw;
@@ -117,9 +116,8 @@ struct ntb_transport_qp {
 	unsigned int tx_max_entry;
 	unsigned int tx_max_frame;
 
-	void (*rx_handler) (struct ntb_transport_qp *qp, void *qp_data,
-			    void *data, int len);
-	struct tasklet_struct rx_work;
+	void (*rx_handler)(struct ntb_transport_qp *qp, void *qp_data,
+			   void *data, int len);
 	struct list_head rx_pend_q;
 	struct list_head rx_free_q;
 	spinlock_t ntb_rx_pend_q_lock;
@@ -130,7 +128,7 @@ struct ntb_transport_qp {
 	unsigned int rx_max_frame;
 	dma_cookie_t last_cookie;
 
-	void (*event_handler) (void *data, int status);
+	void (*event_handler)(void *data, int status);
 	struct delayed_work link_work;
 	struct work_struct link_cleanup;
 
@@ -481,7 +479,7 @@ static void ntb_list_add(spinlock_t *lock, struct list_head *entry,
 }
 
 static struct ntb_queue_entry *ntb_list_rm(spinlock_t *lock,
-						struct list_head *list)
+					   struct list_head *list)
 {
 	struct ntb_queue_entry *entry;
 	unsigned long flags;
@@ -584,11 +582,8 @@ static int ntb_set_mw(struct ntb_transport *nt, int num_mw, unsigned int size)
 	return 0;
 }
 
-static void ntb_qp_link_cleanup(struct work_struct *work)
+static void ntb_qp_link_cleanup(struct ntb_transport_qp *qp)
 {
-	struct ntb_transport_qp *qp = container_of(work,
-						   struct ntb_transport_qp,
-						   link_cleanup);
 	struct ntb_transport *nt = qp->transport;
 	struct pci_dev *pdev = ntb_query_pdev(nt->ndev);
 
@@ -602,6 +597,16 @@ static void ntb_qp_link_cleanup(struct work_struct *work)
 
 	dev_info(&pdev->dev, "qp %d: Link Down\n", qp->qp_num);
 	qp->qp_link = NTB_LINK_DOWN;
+}
+
+static void ntb_qp_link_cleanup_work(struct work_struct *work)
+{
+	struct ntb_transport_qp *qp = container_of(work,
+						   struct ntb_transport_qp,
+						   link_cleanup);
+	struct ntb_transport *nt = qp->transport;
+
+	ntb_qp_link_cleanup(qp);
 
 	if (nt->transport_link == NTB_LINK_UP)
 		schedule_delayed_work(&qp->link_work,
@@ -613,22 +618,20 @@ static void ntb_qp_link_down(struct ntb_transport_qp *qp)
 	schedule_work(&qp->link_cleanup);
 }
 
-static void ntb_transport_link_cleanup(struct work_struct *work)
+static void ntb_transport_link_cleanup(struct ntb_transport *nt)
 {
-	struct ntb_transport *nt = container_of(work, struct ntb_transport,
-						link_cleanup);
 	int i;
 
+	/* Pass along the info to any clients */
+	for (i = 0; i < nt->max_qps; i++)
+		if (!test_bit(i, &nt->qp_bitmap))
+			ntb_qp_link_cleanup(&nt->qps[i]);
+
 	if (nt->transport_link == NTB_LINK_DOWN)
 		cancel_delayed_work_sync(&nt->link_work);
 	else
 		nt->transport_link = NTB_LINK_DOWN;
 
-	/* Pass along the info to any clients */
-	for (i = 0; i < nt->max_qps; i++)
-		if (!test_bit(i, &nt->qp_bitmap))
-			ntb_qp_link_down(&nt->qps[i]);
-
 	/* The scratchpad registers keep the values if the remote side
 	 * goes down, blast them now to give them a sane value the next
 	 * time they are accessed
@@ -637,6 +640,14 @@ static void ntb_transport_link_cleanup(struct work_struct *work)
 		ntb_write_local_spad(nt->ndev, i, 0);
 }
 
+static void ntb_transport_link_cleanup_work(struct work_struct *work)
+{
+	struct ntb_transport *nt = container_of(work, struct ntb_transport,
+						link_cleanup);
+
+	ntb_transport_link_cleanup(nt);
+}
+
 static void ntb_transport_event_callback(void *data, enum ntb_hw_event event)
 {
 	struct ntb_transport *nt = data;
@@ -827,7 +838,7 @@ static void ntb_qp_link_work(struct work_struct *work)
 }
 
 static int ntb_transport_init_queue(struct ntb_transport *nt,
-				     unsigned int qp_num)
+				    unsigned int qp_num)
 {
 	struct ntb_transport_qp *qp;
 	unsigned int num_qps_mw, tx_size;
@@ -880,7 +891,7 @@ static int ntb_transport_init_queue(struct ntb_transport *nt,
 	}
 
 	INIT_DELAYED_WORK(&qp->link_work, ntb_qp_link_work);
-	INIT_WORK(&qp->link_cleanup, ntb_qp_link_cleanup);
+	INIT_WORK(&qp->link_cleanup, ntb_qp_link_cleanup_work);
 
 	spin_lock_init(&qp->ntb_rx_pend_q_lock);
 	spin_lock_init(&qp->ntb_rx_free_q_lock);
@@ -936,7 +947,7 @@ int ntb_transport_init(struct pci_dev *pdev)
 	}
 
 	INIT_DELAYED_WORK(&nt->link_work, ntb_transport_link_work);
-	INIT_WORK(&nt->link_cleanup, ntb_transport_link_cleanup);
+	INIT_WORK(&nt->link_cleanup, ntb_transport_link_cleanup_work);
 
 	rc = ntb_register_event_callback(nt->ndev,
 					 ntb_transport_event_callback);
@@ -972,7 +983,7 @@ void ntb_transport_free(void *transport)
 	struct ntb_device *ndev = nt->ndev;
 	int i;
 
-	nt->transport_link = NTB_LINK_DOWN;
+	ntb_transport_link_cleanup(nt);
 
 	/* verify that all the qp's are freed */
 	for (i = 0; i < nt->max_qps; i++) {
@@ -1034,46 +1045,59 @@ static void ntb_async_rx(struct ntb_queue_entry *entry, void *offset,
 	struct dma_chan *chan = qp->dma_chan;
 	struct dma_device *device;
 	size_t pay_off, buff_off;
-	dma_addr_t src, dest;
+	struct dmaengine_unmap_data *unmap;
 	dma_cookie_t cookie;
 	void *buf = entry->buf;
-	unsigned long flags;
 
 	entry->len = len;
 
 	if (!chan)
 		goto err;
 
-	if (len < copy_bytes) 
-		goto err1;
+	if (len < copy_bytes)
+		goto err_wait;
 
 	device = chan->device;
 	pay_off = (size_t) offset & ~PAGE_MASK;
 	buff_off = (size_t) buf & ~PAGE_MASK;
 
 	if (!is_dma_copy_aligned(device, pay_off, buff_off, len))
-		goto err1;
+		goto err_wait;
 
-	dest = dma_map_single(device->dev, buf, len, DMA_FROM_DEVICE);
-	if (dma_mapping_error(device->dev, dest))
-		goto err1;
+	unmap = dmaengine_get_unmap_data(device->dev, 2, GFP_NOWAIT);
+	if (!unmap)
+		goto err_wait;
 
-	src = dma_map_single(device->dev, offset, len, DMA_TO_DEVICE);
-	if (dma_mapping_error(device->dev, src))
-		goto err2;
+	unmap->len = len;
+	unmap->addr[0] = dma_map_page(device->dev, virt_to_page(offset),
+				      pay_off, len, DMA_TO_DEVICE);
+	if (dma_mapping_error(device->dev, unmap->addr[0]))
+		goto err_get_unmap;
+
+	unmap->to_cnt = 1;
+
+	unmap->addr[1] = dma_map_page(device->dev, virt_to_page(buf),
+				      buff_off, len, DMA_FROM_DEVICE);
+	if (dma_mapping_error(device->dev, unmap->addr[1]))
+		goto err_get_unmap;
+
+	unmap->from_cnt = 1;
 
-	flags = DMA_COMPL_DEST_UNMAP_SINGLE | DMA_COMPL_SRC_UNMAP_SINGLE |
-		DMA_PREP_INTERRUPT;
-	txd = device->device_prep_dma_memcpy(chan, dest, src, len, flags);
+	txd = device->device_prep_dma_memcpy(chan, unmap->addr[1],
+					     unmap->addr[0], len,
+					     DMA_PREP_INTERRUPT);
 	if (!txd)
-		goto err3;
+		goto err_get_unmap;
 
 	txd->callback = ntb_rx_copy_callback;
 	txd->callback_param = entry;
+	dma_set_unmap(txd, unmap);
 
 	cookie = dmaengine_submit(txd);
 	if (dma_submit_error(cookie))
-		goto err3;
+		goto err_set_unmap;
+
+	dmaengine_unmap_put(unmap);
 
 	qp->last_cookie = cookie;
 
@@ -1081,11 +1105,11 @@ static void ntb_async_rx(struct ntb_queue_entry *entry, void *offset,
 
 	return;
 
-err3:
-	dma_unmap_single(device->dev, src, len, DMA_TO_DEVICE);
-err2:
-	dma_unmap_single(device->dev, dest, len, DMA_FROM_DEVICE);
-err1:
+err_set_unmap:
+	dmaengine_unmap_put(unmap);
+err_get_unmap:
+	dmaengine_unmap_put(unmap);
+err_wait:
 	/* If the callbacks come out of order, the writing of the index to the
 	 * last completed will be out of order.  This may result in the
 	 * receive stalling forever.
@@ -1165,8 +1189,7 @@ out:
 	return 0;
 
 err:
-	ntb_list_add(&qp->ntb_rx_pend_q_lock, &entry->entry,
-		     &qp->rx_pend_q);
+	ntb_list_add(&qp->ntb_rx_pend_q_lock, &entry->entry, &qp->rx_pend_q);
 	/* Ensure that the data is fully copied out before clearing the flag */
 	wmb();
 	hdr->flags = 0;
@@ -1175,11 +1198,14 @@ err:
 	goto out;
 }
 
-static void ntb_transport_rx(unsigned long data)
+static int ntb_transport_rxc_db(void *data, int db_num)
 {
-	struct ntb_transport_qp *qp = (struct ntb_transport_qp *)data;
+	struct ntb_transport_qp *qp = data;
 	int rc, i;
 
+	dev_dbg(&ntb_query_pdev(qp->ndev)->dev, "%s: doorbell %d received\n",
+		__func__, db_num);
+
 	/* Limit the number of packets processed in a single interrupt to
 	 * provide fairness to others
 	 */
@@ -1191,16 +1217,8 @@ static void ntb_transport_rx(unsigned long data)
 
 	if (qp->dma_chan)
 		dma_async_issue_pending(qp->dma_chan);
-}
-
-static void ntb_transport_rxc_db(void *data, int db_num)
-{
-	struct ntb_transport_qp *qp = data;
-
-	dev_dbg(&ntb_query_pdev(qp->ndev)->dev, "%s: doorbell %d received\n",
-		__func__, db_num);
 
-	tasklet_schedule(&qp->rx_work);
+	return i;
 }
 
 static void ntb_tx_copy_callback(void *data)
@@ -1245,12 +1263,12 @@ static void ntb_async_tx(struct ntb_transport_qp *qp,
 	struct dma_chan *chan = qp->dma_chan;
 	struct dma_device *device;
 	size_t dest_off, buff_off;
-	dma_addr_t src, dest;
+	struct dmaengine_unmap_data *unmap;
+	dma_addr_t dest;
 	dma_cookie_t cookie;
 	void __iomem *offset;
 	size_t len = entry->len;
 	void *buf = entry->buf;
-	unsigned long flags;
 
 	offset = qp->tx_mw + qp->tx_max_frame * qp->tx_index;
 	hdr = offset + qp->tx_max_frame - sizeof(struct ntb_payload_header);
@@ -1273,28 +1291,41 @@ static void ntb_async_tx(struct ntb_transport_qp *qp,
 	if (!is_dma_copy_aligned(device, buff_off, dest_off, len))
 		goto err;
 
-	src = dma_map_single(device->dev, buf, len, DMA_TO_DEVICE);
-	if (dma_mapping_error(device->dev, src))
+	unmap = dmaengine_get_unmap_data(device->dev, 1, GFP_NOWAIT);
+	if (!unmap)
 		goto err;
 
-	flags = DMA_COMPL_SRC_UNMAP_SINGLE | DMA_PREP_INTERRUPT;
-	txd = device->device_prep_dma_memcpy(chan, dest, src, len, flags);
+	unmap->len = len;
+	unmap->addr[0] = dma_map_page(device->dev, virt_to_page(buf),
+				      buff_off, len, DMA_TO_DEVICE);
+	if (dma_mapping_error(device->dev, unmap->addr[0]))
+		goto err_get_unmap;
+
+	unmap->to_cnt = 1;
+
+	txd = device->device_prep_dma_memcpy(chan, dest, unmap->addr[0], len,
+					     DMA_PREP_INTERRUPT);
 	if (!txd)
-		goto err1;
+		goto err_get_unmap;
 
 	txd->callback = ntb_tx_copy_callback;
 	txd->callback_param = entry;
+	dma_set_unmap(txd, unmap);
 
 	cookie = dmaengine_submit(txd);
 	if (dma_submit_error(cookie))
-		goto err1;
+		goto err_set_unmap;
+
+	dmaengine_unmap_put(unmap);
 
 	dma_async_issue_pending(chan);
 	qp->tx_async++;
 
 	return;
-err1:
-	dma_unmap_single(device->dev, src, len, DMA_TO_DEVICE);
+err_set_unmap:
+	dmaengine_unmap_put(unmap);
+err_get_unmap:
+	dmaengine_unmap_put(unmap);
 err:
 	ntb_memcpy_tx(entry, offset);
 	qp->tx_memcpy++;
@@ -1406,11 +1437,12 @@ ntb_transport_create_queue(void *data, struct pci_dev *pdev,
 	qp->tx_handler = handlers->tx_handler;
 	qp->event_handler = handlers->event_handler;
 
+	dmaengine_get();
 	qp->dma_chan = dma_find_channel(DMA_MEMCPY);
-	if (!qp->dma_chan)
+	if (!qp->dma_chan) {
+		dmaengine_put();
 		dev_info(&pdev->dev, "Unable to allocate DMA channel, using CPU instead\n");
-	else
-		dmaengine_get();
+	}
 
 	for (i = 0; i < NTB_QP_DEF_NUM_ENTRIES; i++) {
 		entry = kzalloc(sizeof(struct ntb_queue_entry), GFP_ATOMIC);
@@ -1432,25 +1464,23 @@ ntb_transport_create_queue(void *data, struct pci_dev *pdev,
 			     &qp->tx_free_q);
 	}
 
-	tasklet_init(&qp->rx_work, ntb_transport_rx, (unsigned long) qp);
-
 	rc = ntb_register_db_callback(qp->ndev, free_queue, qp,
 				      ntb_transport_rxc_db);
 	if (rc)
-		goto err3;
+		goto err2;
 
 	dev_info(&pdev->dev, "NTB Transport QP %d created\n", qp->qp_num);
 
 	return qp;
 
-err3:
-	tasklet_disable(&qp->rx_work);
 err2:
 	while ((entry = ntb_list_rm(&qp->ntb_tx_free_q_lock, &qp->tx_free_q)))
 		kfree(entry);
 err1:
 	while ((entry = ntb_list_rm(&qp->ntb_rx_free_q_lock, &qp->rx_free_q)))
 		kfree(entry);
+	if (qp->dma_chan)
+		dmaengine_put();
 	set_bit(free_queue, &nt->qp_bitmap);
 err:
 	return NULL;
@@ -1489,7 +1519,6 @@ void ntb_transport_free_queue(struct ntb_transport_qp *qp)
 	}
 
 	ntb_unregister_db_callback(qp->ndev, qp->qp_num);
-	tasklet_disable(&qp->rx_work);
 
 	cancel_delayed_work_sync(&qp->link_work);