Merge tag 'ntb-5.8' of git://github.com/jonmason/ntb

Pull NTB updates from Jon Mason:
 "Intel Icelake NTB support, Intel driver bug fixes, and lots of bug
  fixes for ntb tests"

* tag 'ntb-5.8' of git://github.com/jonmason/ntb:
  NTB: ntb_test: Fix bug when counting remote files
  NTB: perf: Fix race condition when run with ntb_test
  NTB: perf: Fix support for hardware that doesn't have port numbers
  NTB: perf: Don't require one more memory window than number of peers
  NTB: ntb_pingpong: Choose doorbells based on port number
  NTB: Fix the default port and peer numbers for legacy drivers
  NTB: Revert the change to use the NTB device dev for DMA allocations
  NTB: ntb_tool: reading the link file should not end in a NULL byte
  ntb_perf: avoid false dma unmap of destination address
  ntb_perf: increase sleep time from one milli sec to one sec
  ntb_tool: pass correct struct device to dma_alloc_coherent
  ntb_perf: pass correct struct device to dma_alloc_coherent
  ntb: hw: remove the code that sets the DMA mask
  NTB: correct ntb_peer_spad_addr and ntb_peer_spad_read comment typos
  ntb: intel: fix static declaration
  ntb: intel: add hw workaround for NTB BAR alignment
  ntb: intel: Add Icelake (gen4) support for Intel NTB
  NTB: Fix static check warning in perf_clear_test
  include/ntb: Fix typo in ntb_unregister_device description

14 files changed, 747 insertions, 81 deletions

diff --git a/drivers/ntb/core.c b/drivers/ntb/core.c
index 2581ab724c34..f8f75a504a58 100644
--- a/drivers/ntb/core.c
+++ b/drivers/ntb/core.c
@@ -214,10 +214,8 @@ int ntb_default_port_number(struct ntb_dev *ntb)
 	case NTB_TOPO_B2B_DSD:
 		return NTB_PORT_SEC_DSD;
 	default:
-		break;
+		return 0;
 	}
-
-	return -EINVAL;
 }
 EXPORT_SYMBOL(ntb_default_port_number);
 
@@ -240,10 +238,8 @@ int ntb_default_peer_port_number(struct ntb_dev *ntb, int pidx)
 	case NTB_TOPO_B2B_DSD:
 		return NTB_PORT_PRI_USD;
 	default:
-		break;
+		return 0;
 	}
-
-	return -EINVAL;
 }
 EXPORT_SYMBOL(ntb_default_peer_port_number);
 
@@ -315,4 +311,3 @@ static void __exit ntb_driver_exit(void)
 	bus_unregister(&ntb_bus);
 }
 module_exit(ntb_driver_exit);
-
diff --git a/drivers/ntb/hw/amd/ntb_hw_amd.c b/drivers/ntb/hw/amd/ntb_hw_amd.c
index 9e310e1ad4d0..88e1db65be02 100644
--- a/drivers/ntb/hw/amd/ntb_hw_amd.c
+++ b/drivers/ntb/hw/amd/ntb_hw_amd.c
@@ -1191,10 +1191,6 @@ static int amd_ntb_init_pci(struct amd_ntb_dev *ndev,
 			goto err_dma_mask;
 		dev_warn(&pdev->dev, "Cannot DMA consistent highmem\n");
 	}
-	rc = dma_coerce_mask_and_coherent(&ndev->ntb.dev,
-					  dma_get_mask(&pdev->dev));
-	if (rc)
-		goto err_dma_mask;
 
 	ndev->self_mmio = pci_iomap(pdev, 0, 0);
 	if (!ndev->self_mmio) {
diff --git a/drivers/ntb/hw/idt/ntb_hw_idt.c b/drivers/ntb/hw/idt/ntb_hw_idt.c
index edae52384b8a..d54261f50851 100644
--- a/drivers/ntb/hw/idt/ntb_hw_idt.c
+++ b/drivers/ntb/hw/idt/ntb_hw_idt.c
@@ -2660,12 +2660,6 @@ static int idt_init_pci(struct idt_ntb_dev *ndev)
 		dev_warn(&pdev->dev,
 			"Cannot set consistent DMA highmem bit mask\n");
 	}
-	ret = dma_coerce_mask_and_coherent(&ndev->ntb.dev,
-					   dma_get_mask(&pdev->dev));
-	if (ret != 0) {
-		dev_err(&pdev->dev, "Failed to set NTB device DMA bit mask\n");
-		return ret;
-	}
 
 	/*
 	 * Enable the device advanced error reporting. It's not critical to
diff --git a/drivers/ntb/hw/intel/Makefile b/drivers/ntb/hw/intel/Makefile
index 60ec8a773eea..f80da0ba15b2 100644
--- a/drivers/ntb/hw/intel/Makefile
+++ b/drivers/ntb/hw/intel/Makefile
@@ -1,3 +1,3 @@
 # SPDX-License-Identifier: GPL-2.0-only
 obj-$(CONFIG_NTB_INTEL) += ntb_hw_intel.o
-ntb_hw_intel-y := ntb_hw_gen1.o ntb_hw_gen3.o
+ntb_hw_intel-y := ntb_hw_gen1.o ntb_hw_gen3.o ntb_hw_gen4.o
diff --git a/drivers/ntb/hw/intel/ntb_hw_gen1.c b/drivers/ntb/hw/intel/ntb_hw_gen1.c
index bb57ec239029..423f9b8fbbcf 100644
--- a/drivers/ntb/hw/intel/ntb_hw_gen1.c
+++ b/drivers/ntb/hw/intel/ntb_hw_gen1.c
@@ -60,6 +60,7 @@
 #include "ntb_hw_intel.h"
 #include "ntb_hw_gen1.h"
 #include "ntb_hw_gen3.h"
+#include "ntb_hw_gen4.h"
 
 #define NTB_NAME	"ntb_hw_intel"
 #define NTB_DESC	"Intel(R) PCI-E Non-Transparent Bridge Driver"
@@ -762,6 +763,8 @@ static ssize_t ndev_debugfs_read(struct file *filp, char __user *ubuf,
 		return ndev_ntb_debugfs_read(filp, ubuf, count, offp);
 	else if (pdev_is_gen3(ndev->ntb.pdev))
 		return ndev_ntb3_debugfs_read(filp, ubuf, count, offp);
+	else if (pdev_is_gen4(ndev->ntb.pdev))
+		return ndev_ntb4_debugfs_read(filp, ubuf, count, offp);
 
 	return -ENXIO;
 }
@@ -1783,10 +1786,6 @@ static int intel_ntb_init_pci(struct intel_ntb_dev *ndev, struct pci_dev *pdev)
 			goto err_dma_mask;
 		dev_warn(&pdev->dev, "Cannot DMA consistent highmem\n");
 	}
-	rc = dma_coerce_mask_and_coherent(&ndev->ntb.dev,
-					  dma_get_mask(&pdev->dev));
-	if (rc)
-		goto err_dma_mask;
 
 	ndev->self_mmio = pci_iomap(pdev, 0, 0);
 	if (!ndev->self_mmio) {
@@ -1858,16 +1857,15 @@ static int intel_ntb_pci_probe(struct pci_dev *pdev,
 	int rc, node;
 
 	node = dev_to_node(&pdev->dev);
+	ndev = kzalloc_node(sizeof(*ndev), GFP_KERNEL, node);
+	if (!ndev) {
+		rc = -ENOMEM;
+		goto err_ndev;
+	}
 
-	if (pdev_is_gen1(pdev)) {
-		ndev = kzalloc_node(sizeof(*ndev), GFP_KERNEL, node);
-		if (!ndev) {
-			rc = -ENOMEM;
-			goto err_ndev;
-		}
-
-		ndev_init_struct(ndev, pdev);
+	ndev_init_struct(ndev, pdev);
 
+	if (pdev_is_gen1(pdev)) {
 		rc = intel_ntb_init_pci(ndev, pdev);
 		if (rc)
 			goto err_init_pci;
@@ -1875,17 +1873,8 @@ static int intel_ntb_pci_probe(struct pci_dev *pdev,
 		rc = xeon_init_dev(ndev);
 		if (rc)
 			goto err_init_dev;
-
 	} else if (pdev_is_gen3(pdev)) {
-		ndev = kzalloc_node(sizeof(*ndev), GFP_KERNEL, node);
-		if (!ndev) {
-			rc = -ENOMEM;
-			goto err_ndev;
-		}
-
-		ndev_init_struct(ndev, pdev);
 		ndev->ntb.ops = &intel_ntb3_ops;
-
 		rc = intel_ntb_init_pci(ndev, pdev);
 		if (rc)
 			goto err_init_pci;
@@ -1893,7 +1882,15 @@ static int intel_ntb_pci_probe(struct pci_dev *pdev,
 		rc = gen3_init_dev(ndev);
 		if (rc)
 			goto err_init_dev;
+	} else if (pdev_is_gen4(pdev)) {
+		ndev->ntb.ops = &intel_ntb4_ops;
+		rc = intel_ntb_init_pci(ndev, pdev);
+		if (rc)
+			goto err_init_pci;
 
+		rc = gen4_init_dev(ndev);
+		if (rc)
+			goto err_init_dev;
 	} else {
 		rc = -EINVAL;
 		goto err_ndev;
@@ -1915,7 +1912,7 @@ static int intel_ntb_pci_probe(struct pci_dev *pdev,
 
 err_register:
 	ndev_deinit_debugfs(ndev);
-	if (pdev_is_gen1(pdev) || pdev_is_gen3(pdev))
+	if (pdev_is_gen1(pdev) || pdev_is_gen3(pdev) || pdev_is_gen4(pdev))
 		xeon_deinit_dev(ndev);
 err_init_dev:
 	intel_ntb_deinit_pci(ndev);
@@ -1931,7 +1928,7 @@ static void intel_ntb_pci_remove(struct pci_dev *pdev)
 
 	ntb_unregister_device(&ndev->ntb);
 	ndev_deinit_debugfs(ndev);
-	if (pdev_is_gen1(pdev) || pdev_is_gen3(pdev))
+	if (pdev_is_gen1(pdev) || pdev_is_gen3(pdev) || pdev_is_gen4(pdev))
 		xeon_deinit_dev(ndev);
 	intel_ntb_deinit_pci(ndev);
 	kfree(ndev);
@@ -2036,6 +2033,7 @@ static const struct file_operations intel_ntb_debugfs_info = {
 };
 
 static const struct pci_device_id intel_ntb_pci_tbl[] = {
+	/* GEN1 */
 	{PCI_VDEVICE(INTEL, PCI_DEVICE_ID_INTEL_NTB_B2B_JSF)},
 	{PCI_VDEVICE(INTEL, PCI_DEVICE_ID_INTEL_NTB_B2B_SNB)},
 	{PCI_VDEVICE(INTEL, PCI_DEVICE_ID_INTEL_NTB_B2B_IVT)},
@@ -2051,7 +2049,12 @@ static const struct pci_device_id intel_ntb_pci_tbl[] = {
 	{PCI_VDEVICE(INTEL, PCI_DEVICE_ID_INTEL_NTB_SS_IVT)},
 	{PCI_VDEVICE(INTEL, PCI_DEVICE_ID_INTEL_NTB_SS_HSX)},
 	{PCI_VDEVICE(INTEL, PCI_DEVICE_ID_INTEL_NTB_SS_BDX)},
+
+	/* GEN3 */
 	{PCI_VDEVICE(INTEL, PCI_DEVICE_ID_INTEL_NTB_B2B_SKX)},
+
+	/* GEN4 */
+	{PCI_VDEVICE(INTEL, PCI_DEVICE_ID_INTEL_NTB_B2B_ICX)},
 	{0}
 };
 MODULE_DEVICE_TABLE(pci, intel_ntb_pci_tbl);
diff --git a/drivers/ntb/hw/intel/ntb_hw_gen1.h b/drivers/ntb/hw/intel/ntb_hw_gen1.h
index 544cf5c06f4d..1b759942d8af 100644
--- a/drivers/ntb/hw/intel/ntb_hw_gen1.h
+++ b/drivers/ntb/hw/intel/ntb_hw_gen1.h
@@ -140,6 +140,7 @@
 #define NTB_HWERR_SB01BASE_LOCKUP	BIT_ULL(1)
 #define NTB_HWERR_B2BDOORBELL_BIT14	BIT_ULL(2)
 #define NTB_HWERR_MSIX_VECTOR32_BAD	BIT_ULL(3)
+#define NTB_HWERR_BAR_ALIGN		BIT_ULL(4)
 
 extern struct intel_b2b_addr xeon_b2b_usd_addr;
 extern struct intel_b2b_addr xeon_b2b_dsd_addr;
diff --git a/drivers/ntb/hw/intel/ntb_hw_gen3.c b/drivers/ntb/hw/intel/ntb_hw_gen3.c
index c3397160db7f..ffcfc3e02c35 100644
--- a/drivers/ntb/hw/intel/ntb_hw_gen3.c
+++ b/drivers/ntb/hw/intel/ntb_hw_gen3.c
@@ -415,9 +415,8 @@ ssize_t ndev_ntb3_debugfs_read(struct file *filp, char __user *ubuf,
 	return ret;
 }
 
-static int intel_ntb3_link_enable(struct ntb_dev *ntb,
-				  enum ntb_speed max_speed,
-				  enum ntb_width max_width)
+int intel_ntb3_link_enable(struct ntb_dev *ntb, enum ntb_speed max_speed,
+		enum ntb_width max_width)
 {
 	struct intel_ntb_dev *ndev;
 	u32 ntb_ctl;
@@ -532,7 +531,7 @@ static int intel_ntb3_mw_set_trans(struct ntb_dev *ntb, int pidx, int idx,
 	return 0;
 }
 
-static int intel_ntb3_peer_db_addr(struct ntb_dev *ntb, phys_addr_t *db_addr,
+int intel_ntb3_peer_db_addr(struct ntb_dev *ntb, phys_addr_t *db_addr,
 				   resource_size_t *db_size,
 				   u64 *db_data, int db_bit)
 {
@@ -563,7 +562,7 @@ static int intel_ntb3_peer_db_addr(struct ntb_dev *ntb, phys_addr_t *db_addr,
 	return 0;
 }
 
-static int intel_ntb3_peer_db_set(struct ntb_dev *ntb, u64 db_bits)
+int intel_ntb3_peer_db_set(struct ntb_dev *ntb, u64 db_bits)
 {
 	struct intel_ntb_dev *ndev = ntb_ndev(ntb);
 	int bit;
@@ -581,7 +580,7 @@ static int intel_ntb3_peer_db_set(struct ntb_dev *ntb, u64 db_bits)
 	return 0;
 }
 
-static u64 intel_ntb3_db_read(struct ntb_dev *ntb)
+u64 intel_ntb3_db_read(struct ntb_dev *ntb)
 {
 	struct intel_ntb_dev *ndev = ntb_ndev(ntb);
 
@@ -590,7 +589,7 @@ static u64 intel_ntb3_db_read(struct ntb_dev *ntb)
 			    ndev->self_reg->db_clear);
 }
 
-static int intel_ntb3_db_clear(struct ntb_dev *ntb, u64 db_bits)
+int intel_ntb3_db_clear(struct ntb_dev *ntb, u64 db_bits)
 {
 	struct intel_ntb_dev *ndev = ntb_ndev(ntb);
 
diff --git a/drivers/ntb/hw/intel/ntb_hw_gen3.h b/drivers/ntb/hw/intel/ntb_hw_gen3.h
index 75fb86ca27bb..2bc5d8356045 100644
--- a/drivers/ntb/hw/intel/ntb_hw_gen3.h
+++ b/drivers/ntb/hw/intel/ntb_hw_gen3.h
@@ -104,6 +104,14 @@ static inline void gen3_db_iowrite(u64 bits, void __iomem *mmio)
 ssize_t ndev_ntb3_debugfs_read(struct file *filp, char __user *ubuf,
 				      size_t count, loff_t *offp);
 int gen3_init_dev(struct intel_ntb_dev *ndev);
+int intel_ntb3_link_enable(struct ntb_dev *ntb, enum ntb_speed max_speed,
+		enum ntb_width max_width);
+u64 intel_ntb3_db_read(struct ntb_dev *ntb);
+int intel_ntb3_db_clear(struct ntb_dev *ntb, u64 db_bits);
+int intel_ntb3_peer_db_set(struct ntb_dev *ntb, u64 db_bits);
+int intel_ntb3_peer_db_addr(struct ntb_dev *ntb, phys_addr_t *db_addr,
+				resource_size_t *db_size,
+				u64 *db_data, int db_bit);
 
 extern const struct ntb_dev_ops intel_ntb3_ops;
 
diff --git a/drivers/ntb/hw/intel/ntb_hw_gen4.c b/drivers/ntb/hw/intel/ntb_hw_gen4.c
new file mode 100644
index 000000000000..bc4541cbf8c6
--- /dev/null
+++ b/drivers/ntb/hw/intel/ntb_hw_gen4.c
@@ -0,0 +1,552 @@
+// SPDX-License-Identifier: (GPL-2.0 OR BSD-3-Clause)
+/* Copyright(c) 2020 Intel Corporation. All rights reserved. */
+#include <linux/debugfs.h>
+#include <linux/delay.h>
+#include <linux/init.h>
+#include <linux/interrupt.h>
+#include <linux/module.h>
+#include <linux/pci.h>
+#include <linux/random.h>
+#include <linux/slab.h>
+#include <linux/ntb.h>
+#include <linux/log2.h>
+
+#include "ntb_hw_intel.h"
+#include "ntb_hw_gen1.h"
+#include "ntb_hw_gen3.h"
+#include "ntb_hw_gen4.h"
+
+static int gen4_poll_link(struct intel_ntb_dev *ndev);
+static int gen4_link_is_up(struct intel_ntb_dev *ndev);
+
+static const struct intel_ntb_reg gen4_reg = {
+	.poll_link		= gen4_poll_link,
+	.link_is_up		= gen4_link_is_up,
+	.db_ioread		= gen3_db_ioread,
+	.db_iowrite		= gen3_db_iowrite,
+	.db_size		= sizeof(u32),
+	.ntb_ctl		= GEN4_NTBCNTL_OFFSET,
+	.mw_bar			= {2, 4},
+};
+
+static const struct intel_ntb_alt_reg gen4_pri_reg = {
+	.db_clear		= GEN4_IM_INT_STATUS_OFFSET,
+	.db_mask		= GEN4_IM_INT_DISABLE_OFFSET,
+	.spad			= GEN4_IM_SPAD_OFFSET,
+};
+
+static const struct intel_ntb_xlat_reg gen4_sec_xlat = {
+	.bar2_limit		= GEN4_IM23XLMT_OFFSET,
+	.bar2_xlat		= GEN4_IM23XBASE_OFFSET,
+	.bar2_idx		= GEN4_IM23XBASEIDX_OFFSET,
+};
+
+static const struct intel_ntb_alt_reg gen4_b2b_reg = {
+	.db_bell		= GEN4_IM_DOORBELL_OFFSET,
+	.spad			= GEN4_EM_SPAD_OFFSET,
+};
+
+static int gen4_poll_link(struct intel_ntb_dev *ndev)
+{
+	u16 reg_val;
+
+	/*
+	 * We need to write to DLLSCS bit in the SLOTSTS before we
+	 * can clear the hardware link interrupt on ICX NTB.
+	 */
+	iowrite16(GEN4_SLOTSTS_DLLSCS, ndev->self_mmio + GEN4_SLOTSTS);
+	ndev->reg->db_iowrite(ndev->db_link_mask,
+			      ndev->self_mmio +
+			      ndev->self_reg->db_clear);
+
+	reg_val = ioread16(ndev->self_mmio + GEN4_LINK_STATUS_OFFSET);
+	if (reg_val == ndev->lnk_sta)
+		return 0;
+
+	ndev->lnk_sta = reg_val;
+
+	return 1;
+}
+
+static int gen4_link_is_up(struct intel_ntb_dev *ndev)
+{
+	return NTB_LNK_STA_ACTIVE(ndev->lnk_sta);
+}
+
+static int gen4_init_isr(struct intel_ntb_dev *ndev)
+{
+	int i;
+
+	/*
+	 * The MSIX vectors and the interrupt status bits are not lined up
+	 * on Gen3 (Skylake) and Gen4. By default the link status bit is bit
+	 * 32, however it is by default MSIX vector0. We need to fixup to
+	 * line them up. The vectors at reset is 1-32,0. We need to reprogram
+	 * to 0-32.
+	 */
+	for (i = 0; i < GEN4_DB_MSIX_VECTOR_COUNT; i++)
+		iowrite8(i, ndev->self_mmio + GEN4_INTVEC_OFFSET + i);
+
+	return ndev_init_isr(ndev, GEN4_DB_MSIX_VECTOR_COUNT,
+			     GEN4_DB_MSIX_VECTOR_COUNT,
+			     GEN4_DB_MSIX_VECTOR_SHIFT,
+			     GEN4_DB_TOTAL_SHIFT);
+}
+
+static int gen4_setup_b2b_mw(struct intel_ntb_dev *ndev,
+			    const struct intel_b2b_addr *addr,
+			    const struct intel_b2b_addr *peer_addr)
+{
+	struct pci_dev *pdev;
+	void __iomem *mmio;
+	phys_addr_t bar_addr;
+
+	pdev = ndev->ntb.pdev;
+	mmio = ndev->self_mmio;
+
+	/* setup incoming bar limits == base addrs (zero length windows) */
+	bar_addr = addr->bar2_addr64;
+	iowrite64(bar_addr, mmio + GEN4_IM23XLMT_OFFSET);
+	bar_addr = ioread64(mmio + GEN4_IM23XLMT_OFFSET);
+	dev_dbg(&pdev->dev, "IM23XLMT %#018llx\n", bar_addr);
+
+	bar_addr = addr->bar4_addr64;
+	iowrite64(bar_addr, mmio + GEN4_IM45XLMT_OFFSET);
+	bar_addr = ioread64(mmio + GEN4_IM45XLMT_OFFSET);
+	dev_dbg(&pdev->dev, "IM45XLMT %#018llx\n", bar_addr);
+
+	/* zero incoming translation addrs */
+	iowrite64(0, mmio + GEN4_IM23XBASE_OFFSET);
+	iowrite64(0, mmio + GEN4_IM45XBASE_OFFSET);
+
+	ndev->peer_mmio = ndev->self_mmio;
+
+	return 0;
+}
+
+static int gen4_init_ntb(struct intel_ntb_dev *ndev)
+{
+	int rc;
+
+
+	ndev->mw_count = XEON_MW_COUNT;
+	ndev->spad_count = GEN4_SPAD_COUNT;
+	ndev->db_count = GEN4_DB_COUNT;
+	ndev->db_link_mask = GEN4_DB_LINK_BIT;
+
+	ndev->self_reg = &gen4_pri_reg;
+	ndev->xlat_reg = &gen4_sec_xlat;
+	ndev->peer_reg = &gen4_b2b_reg;
+
+	if (ndev->ntb.topo == NTB_TOPO_B2B_USD)
+		rc = gen4_setup_b2b_mw(ndev, &xeon_b2b_dsd_addr,
+				&xeon_b2b_usd_addr);
+	else
+		rc = gen4_setup_b2b_mw(ndev, &xeon_b2b_usd_addr,
+				&xeon_b2b_dsd_addr);
+	if (rc)
+		return rc;
+
+	ndev->db_valid_mask = BIT_ULL(ndev->db_count) - 1;
+
+	ndev->reg->db_iowrite(ndev->db_valid_mask,
+			      ndev->self_mmio +
+			      ndev->self_reg->db_mask);
+
+	return 0;
+}
+
+static enum ntb_topo gen4_ppd_topo(struct intel_ntb_dev *ndev, u32 ppd)
+{
+	switch (ppd & GEN4_PPD_TOPO_MASK) {
+	case GEN4_PPD_TOPO_B2B_USD:
+		return NTB_TOPO_B2B_USD;
+	case GEN4_PPD_TOPO_B2B_DSD:
+		return NTB_TOPO_B2B_DSD;
+	}
+
+	return NTB_TOPO_NONE;
+}
+
+int gen4_init_dev(struct intel_ntb_dev *ndev)
+{
+	struct pci_dev *pdev = ndev->ntb.pdev;
+	u32 ppd1/*, ppd0*/;
+	u16 lnkctl;
+	int rc;
+
+	ndev->reg = &gen4_reg;
+
+	if (pdev_is_ICX(pdev))
+		ndev->hwerr_flags |= NTB_HWERR_BAR_ALIGN;
+
+	ppd1 = ioread32(ndev->self_mmio + GEN4_PPD1_OFFSET);
+	ndev->ntb.topo = gen4_ppd_topo(ndev, ppd1);
+	dev_dbg(&pdev->dev, "ppd %#x topo %s\n", ppd1,
+		ntb_topo_string(ndev->ntb.topo));
+	if (ndev->ntb.topo == NTB_TOPO_NONE)
+		return -EINVAL;
+
+	rc = gen4_init_ntb(ndev);
+	if (rc)
+		return rc;
+
+	/* init link setup */
+	lnkctl = ioread16(ndev->self_mmio + GEN4_LINK_CTRL_OFFSET);
+	lnkctl |= GEN4_LINK_CTRL_LINK_DISABLE;
+	iowrite16(lnkctl, ndev->self_mmio + GEN4_LINK_CTRL_OFFSET);
+
+	return gen4_init_isr(ndev);
+}
+
+ssize_t ndev_ntb4_debugfs_read(struct file *filp, char __user *ubuf,
+				      size_t count, loff_t *offp)
+{
+	struct intel_ntb_dev *ndev;
+	void __iomem *mmio;
+	char *buf;
+	size_t buf_size;
+	ssize_t ret, off;
+	union { u64 v64; u32 v32; u16 v16; } u;
+
+	ndev = filp->private_data;
+	mmio = ndev->self_mmio;
+
+	buf_size = min(count, 0x800ul);
+
+	buf = kmalloc(buf_size, GFP_KERNEL);
+	if (!buf)
+		return -ENOMEM;
+
+	off = 0;
+
+	off += scnprintf(buf + off, buf_size - off,
+			 "NTB Device Information:\n");
+
+	off += scnprintf(buf + off, buf_size - off,
+			 "Connection Topology -\t%s\n",
+			 ntb_topo_string(ndev->ntb.topo));
+
+	off += scnprintf(buf + off, buf_size - off,
+			 "NTB CTL -\t\t%#06x\n", ndev->ntb_ctl);
+	off += scnprintf(buf + off, buf_size - off,
+			 "LNK STA (cached) -\t\t%#06x\n", ndev->lnk_sta);
+
+	if (!ndev->reg->link_is_up(ndev))
+		off += scnprintf(buf + off, buf_size - off,
+				 "Link Status -\t\tDown\n");
+	else {
+		off += scnprintf(buf + off, buf_size - off,
+				 "Link Status -\t\tUp\n");
+		off += scnprintf(buf + off, buf_size - off,
+				 "Link Speed -\t\tPCI-E Gen %u\n",
+				 NTB_LNK_STA_SPEED(ndev->lnk_sta));
+		off += scnprintf(buf + off, buf_size - off,
+				 "Link Width -\t\tx%u\n",
+				 NTB_LNK_STA_WIDTH(ndev->lnk_sta));
+	}
+
+	off += scnprintf(buf + off, buf_size - off,
+			 "Memory Window Count -\t%u\n", ndev->mw_count);
+	off += scnprintf(buf + off, buf_size - off,
+			 "Scratchpad Count -\t%u\n", ndev->spad_count);
+	off += scnprintf(buf + off, buf_size - off,
+			 "Doorbell Count -\t%u\n", ndev->db_count);
+	off += scnprintf(buf + off, buf_size - off,
+			 "Doorbell Vector Count -\t%u\n", ndev->db_vec_count);
+	off += scnprintf(buf + off, buf_size - off,
+			 "Doorbell Vector Shift -\t%u\n", ndev->db_vec_shift);
+
+	off += scnprintf(buf + off, buf_size - off,
+			 "Doorbell Valid Mask -\t%#llx\n", ndev->db_valid_mask);
+	off += scnprintf(buf + off, buf_size - off,
+			 "Doorbell Link Mask -\t%#llx\n", ndev->db_link_mask);
+	off += scnprintf(buf + off, buf_size - off,
+			 "Doorbell Mask Cached -\t%#llx\n", ndev->db_mask);
+
+	u.v64 = ndev_db_read(ndev, mmio + ndev->self_reg->db_mask);
+	off += scnprintf(buf + off, buf_size - off,
+			 "Doorbell Mask -\t\t%#llx\n", u.v64);
+
+	off += scnprintf(buf + off, buf_size - off,
+			 "\nNTB Incoming XLAT:\n");
+
+	u.v64 = ioread64(mmio + GEN4_IM23XBASE_OFFSET);
+	off += scnprintf(buf + off, buf_size - off,
+			 "IM23XBASE -\t\t%#018llx\n", u.v64);
+
+	u.v64 = ioread64(mmio + GEN4_IM45XBASE_OFFSET);
+	off += scnprintf(buf + off, buf_size - off,
+			 "IM45XBASE -\t\t%#018llx\n", u.v64);
+
+	u.v64 = ioread64(mmio + GEN4_IM23XLMT_OFFSET);
+	off += scnprintf(buf + off, buf_size - off,
+			 "IM23XLMT -\t\t\t%#018llx\n", u.v64);
+
+	u.v64 = ioread64(mmio + GEN4_IM45XLMT_OFFSET);
+	off += scnprintf(buf + off, buf_size - off,
+			 "IM45XLMT -\t\t\t%#018llx\n", u.v64);
+
+	off += scnprintf(buf + off, buf_size - off,
+			 "\nNTB Statistics:\n");
+
+	off += scnprintf(buf + off, buf_size - off,
+			 "\nNTB Hardware Errors:\n");
+
+	if (!pci_read_config_word(ndev->ntb.pdev,
+				  GEN4_DEVSTS_OFFSET, &u.v16))
+		off += scnprintf(buf + off, buf_size - off,
+				"DEVSTS -\t\t%#06x\n", u.v16);
+
+	u.v16 = ioread16(mmio + GEN4_LINK_STATUS_OFFSET);
+	off += scnprintf(buf + off, buf_size - off,
+			"LNKSTS -\t\t%#06x\n", u.v16);
+
+	if (!pci_read_config_dword(ndev->ntb.pdev,
+				   GEN4_UNCERRSTS_OFFSET, &u.v32))
+		off += scnprintf(buf + off, buf_size - off,
+				 "UNCERRSTS -\t\t%#06x\n", u.v32);
+
+	if (!pci_read_config_dword(ndev->ntb.pdev,
+				   GEN4_CORERRSTS_OFFSET, &u.v32))
+		off += scnprintf(buf + off, buf_size - off,
+				 "CORERRSTS -\t\t%#06x\n", u.v32);
+
+	ret = simple_read_from_buffer(ubuf, count, offp, buf, off);
+	kfree(buf);
+	return ret;
+}
+
+static int intel_ntb4_mw_set_trans(struct ntb_dev *ntb, int pidx, int idx,
+				   dma_addr_t addr, resource_size_t size)
+{
+	struct intel_ntb_dev *ndev = ntb_ndev(ntb);
+	unsigned long xlat_reg, limit_reg, idx_reg;
+	unsigned short base_idx, reg_val16;
+	resource_size_t bar_size, mw_size;
+	void __iomem *mmio;
+	u64 base, limit, reg_val;
+	int bar;
+
+	if (pidx != NTB_DEF_PEER_IDX)
+		return -EINVAL;
+
+	if (idx >= ndev->b2b_idx && !ndev->b2b_off)
+		idx += 1;
+
+	bar = ndev_mw_to_bar(ndev, idx);
+	if (bar < 0)
+		return bar;
+
+	bar_size = pci_resource_len(ndev->ntb.pdev, bar);
+
+	if (idx == ndev->b2b_idx)
+		mw_size = bar_size - ndev->b2b_off;
+	else
+		mw_size = bar_size;
+
+	if (ndev->hwerr_flags & NTB_HWERR_BAR_ALIGN) {
+		/* hardware requires that addr is aligned to bar size */
+		if (addr & (bar_size - 1))
+			return -EINVAL;
+	} else {
+		if (addr & (PAGE_SIZE - 1))
+			return -EINVAL;
+	}
+
+	/* make sure the range fits in the usable mw size */
+	if (size > mw_size)
+		return -EINVAL;
+
+	mmio = ndev->self_mmio;
+	xlat_reg = ndev->xlat_reg->bar2_xlat + (idx * 0x10);
+	limit_reg = ndev->xlat_reg->bar2_limit + (idx * 0x10);
+	base = pci_resource_start(ndev->ntb.pdev, bar);
+
+	/* Set the limit if supported, if size is not mw_size */
+	if (limit_reg && size != mw_size) {
+		limit = base + size;
+		base_idx = __ilog2_u64(size);
+	} else {
+		limit = base + mw_size;
+		base_idx = __ilog2_u64(mw_size);
+	}
+
+
+	/* set and verify setting the translation address */
+	iowrite64(addr, mmio + xlat_reg);
+	reg_val = ioread64(mmio + xlat_reg);
+	if (reg_val != addr) {
+		iowrite64(0, mmio + xlat_reg);
+		return -EIO;
+	}
+
+	dev_dbg(&ntb->pdev->dev, "BAR %d IMXBASE: %#Lx\n", bar, reg_val);
+
+	/* set and verify setting the limit */
+	iowrite64(limit, mmio + limit_reg);
+	reg_val = ioread64(mmio + limit_reg);
+	if (reg_val != limit) {
+		iowrite64(base, mmio + limit_reg);
+		iowrite64(0, mmio + xlat_reg);
+		return -EIO;
+	}
+
+	dev_dbg(&ntb->pdev->dev, "BAR %d IMXLMT: %#Lx\n", bar, reg_val);
+
+	if (ndev->hwerr_flags & NTB_HWERR_BAR_ALIGN) {
+		idx_reg = ndev->xlat_reg->bar2_idx + (idx * 0x2);
+		iowrite16(base_idx, mmio + idx_reg);
+		reg_val16 = ioread16(mmio + idx_reg);
+		if (reg_val16 != base_idx) {
+			iowrite64(base, mmio + limit_reg);
+			iowrite64(0, mmio + xlat_reg);
+			iowrite16(0, mmio + idx_reg);
+			return -EIO;
+		}
+		dev_dbg(&ntb->pdev->dev, "BAR %d IMBASEIDX: %#x\n", bar, reg_val16);
+	}
+
+
+	return 0;
+}
+
+static int intel_ntb4_link_enable(struct ntb_dev *ntb,
+		enum ntb_speed max_speed, enum ntb_width max_width)
+{
+	struct intel_ntb_dev *ndev;
+	u32 ntb_ctl, ppd0;
+	u16 lnkctl;
+
+	ndev = container_of(ntb, struct intel_ntb_dev, ntb);
+
+	dev_dbg(&ntb->pdev->dev,
+			"Enabling link with max_speed %d max_width %d\n",
+			max_speed, max_width);
+
+	if (max_speed != NTB_SPEED_AUTO)
+		dev_dbg(&ntb->pdev->dev,
+				"ignoring max_speed %d\n", max_speed);
+	if (max_width != NTB_WIDTH_AUTO)
+		dev_dbg(&ntb->pdev->dev,
+				"ignoring max_width %d\n", max_width);
+
+	ntb_ctl = NTB_CTL_E2I_BAR23_SNOOP | NTB_CTL_I2E_BAR23_SNOOP;
+	ntb_ctl |= NTB_CTL_E2I_BAR45_SNOOP | NTB_CTL_I2E_BAR45_SNOOP;
+	iowrite32(ntb_ctl, ndev->self_mmio + ndev->reg->ntb_ctl);
+
+	lnkctl = ioread16(ndev->self_mmio + GEN4_LINK_CTRL_OFFSET);
+	lnkctl &= ~GEN4_LINK_CTRL_LINK_DISABLE;
+	iowrite16(lnkctl, ndev->self_mmio + GEN4_LINK_CTRL_OFFSET);
+
+	/* start link training in PPD0 */
+	ppd0 = ioread32(ndev->self_mmio + GEN4_PPD0_OFFSET);
+	ppd0 |= GEN4_PPD_LINKTRN;
+	iowrite32(ppd0, ndev->self_mmio + GEN4_PPD0_OFFSET);
+
+	/* make sure link training has started */
+	ppd0 = ioread32(ndev->self_mmio + GEN4_PPD0_OFFSET);
+	if (!(ppd0 & GEN4_PPD_LINKTRN)) {
+		dev_warn(&ntb->pdev->dev, "Link is not training\n");
+		return -ENXIO;
+	}
+
+	ndev->dev_up = 1;
+
+	return 0;
+}
+
+static int intel_ntb4_link_disable(struct ntb_dev *ntb)
+{
+	struct intel_ntb_dev *ndev;
+	u32 ntb_cntl;
+	u16 lnkctl;
+
+	ndev = container_of(ntb, struct intel_ntb_dev, ntb);
+
+	dev_dbg(&ntb->pdev->dev, "Disabling link\n");
+
+	/* clear the snoop bits */
+	ntb_cntl = ioread32(ndev->self_mmio + ndev->reg->ntb_ctl);
+	ntb_cntl &= ~(NTB_CTL_E2I_BAR23_SNOOP | NTB_CTL_I2E_BAR23_SNOOP);
+	ntb_cntl &= ~(NTB_CTL_E2I_BAR45_SNOOP | NTB_CTL_I2E_BAR45_SNOOP);
+	iowrite32(ntb_cntl, ndev->self_mmio + ndev->reg->ntb_ctl);
+
+	lnkctl = ioread16(ndev->self_mmio + GEN4_LINK_CTRL_OFFSET);
+	lnkctl |= GEN4_LINK_CTRL_LINK_DISABLE;
+	iowrite16(lnkctl, ndev->self_mmio + GEN4_LINK_CTRL_OFFSET);
+
+	ndev->dev_up = 0;
+
+	return 0;
+}
+
+static int intel_ntb4_mw_get_align(struct ntb_dev *ntb, int pidx, int idx,
+				   resource_size_t *addr_align,
+				   resource_size_t *size_align,
+				   resource_size_t *size_max)
+{
+	struct intel_ntb_dev *ndev = ntb_ndev(ntb);
+	resource_size_t bar_size, mw_size;
+	int bar;
+
+	if (pidx != NTB_DEF_PEER_IDX)
+		return -EINVAL;
+
+	if (idx >= ndev->b2b_idx && !ndev->b2b_off)
+		idx += 1;
+
+	bar = ndev_mw_to_bar(ndev, idx);
+	if (bar < 0)
+		return bar;
+
+	bar_size = pci_resource_len(ndev->ntb.pdev, bar);
+
+	if (idx == ndev->b2b_idx)
+		mw_size = bar_size - ndev->b2b_off;
+	else
+		mw_size = bar_size;
+
+	if (addr_align) {
+		if (ndev->hwerr_flags & NTB_HWERR_BAR_ALIGN)
+			*addr_align = pci_resource_len(ndev->ntb.pdev, bar);
+		else
+			*addr_align = PAGE_SIZE;
+	}
+
+	if (size_align)
+		*size_align = 1;
+
+	if (size_max)
+		*size_max = mw_size;
+
+	return 0;
+}
+
+const struct ntb_dev_ops intel_ntb4_ops = {
+	.mw_count		= intel_ntb_mw_count,
+	.mw_get_align		= intel_ntb4_mw_get_align,
+	.mw_set_trans		= intel_ntb4_mw_set_trans,
+	.peer_mw_count		= intel_ntb_peer_mw_count,
+	.peer_mw_get_addr	= intel_ntb_peer_mw_get_addr,
+	.link_is_up		= intel_ntb_link_is_up,
+	.link_enable		= intel_ntb4_link_enable,
+	.link_disable		= intel_ntb4_link_disable,
+	.db_valid_mask		= intel_ntb_db_valid_mask,
+	.db_vector_count	= intel_ntb_db_vector_count,
+	.db_vector_mask		= intel_ntb_db_vector_mask,
+	.db_read		= intel_ntb3_db_read,
+	.db_clear		= intel_ntb3_db_clear,
+	.db_set_mask		= intel_ntb_db_set_mask,
+	.db_clear_mask		= intel_ntb_db_clear_mask,
+	.peer_db_addr		= intel_ntb3_peer_db_addr,
+	.peer_db_set		= intel_ntb3_peer_db_set,
+	.spad_is_unsafe		= intel_ntb_spad_is_unsafe,
+	.spad_count		= intel_ntb_spad_count,
+	.spad_read		= intel_ntb_spad_read,
+	.spad_write		= intel_ntb_spad_write,
+	.peer_spad_addr		= intel_ntb_peer_spad_addr,
+	.peer_spad_read		= intel_ntb_peer_spad_read,
+	.peer_spad_write	= intel_ntb_peer_spad_write,
+};
+
diff --git a/drivers/ntb/hw/intel/ntb_hw_gen4.h b/drivers/ntb/hw/intel/ntb_hw_gen4.h
new file mode 100644
index 000000000000..a868c788de02
--- /dev/null
+++ b/drivers/ntb/hw/intel/ntb_hw_gen4.h
@@ -0,0 +1,100 @@
+/* SPDX-License-Identifier: (GPL-2.0 OR BSD-3-Clause)          */
+/* Copyright(c) 2020 Intel Corporation. All rights reserved.   */
+#ifndef _NTB_INTEL_GEN4_H_
+#define _NTB_INTEL_GEN4_H_
+
+#include "ntb_hw_intel.h"
+
+/* Supported PCI device revision range for ICX */
+#define PCI_DEVICE_REVISION_ICX_MIN	0x2
+#define PCI_DEVICE_REVISION_ICX_MAX	0xF
+
+/* Intel Gen4 NTB hardware */
+/* PCIe config space */
+#define GEN4_IMBAR23SZ_OFFSET		0x00c4
+#define GEN4_IMBAR45SZ_OFFSET		0x00c5
+#define GEN4_EMBAR23SZ_OFFSET		0x00c6
+#define GEN4_EMBAR45SZ_OFFSET		0x00c7
+#define GEN4_DEVCTRL_OFFSET		0x0048
+#define GEN4_DEVSTS_OFFSET		0x004a
+#define GEN4_UNCERRSTS_OFFSET		0x0104
+#define GEN4_CORERRSTS_OFFSET		0x0110
+
+/* BAR0 MMIO */
+#define GEN4_NTBCNTL_OFFSET		0x0000
+#define GEN4_IM23XBASE_OFFSET		0x0010	/* IMBAR1XBASE */
+#define GEN4_IM23XLMT_OFFSET		0x0018  /* IMBAR1XLMT */
+#define GEN4_IM45XBASE_OFFSET		0x0020	/* IMBAR2XBASE */
+#define GEN4_IM45XLMT_OFFSET		0x0028  /* IMBAR2XLMT */
+#define GEN4_IM_INT_STATUS_OFFSET	0x0040
+#define GEN4_IM_INT_DISABLE_OFFSET	0x0048
+#define GEN4_INTVEC_OFFSET		0x0050  /* 0-32 vecs */
+#define GEN4_IM23XBASEIDX_OFFSET	0x0074
+#define GEN4_IM45XBASEIDX_OFFSET	0x0076
+#define GEN4_IM_SPAD_OFFSET		0x0080  /* 0-15 SPADs */
+#define GEN4_IM_SPAD_SEM_OFFSET		0x00c0	/* SPAD hw semaphore */
+#define GEN4_IM_SPAD_STICKY_OFFSET	0x00c4  /* sticky SPAD */
+#define GEN4_IM_DOORBELL_OFFSET		0x0100  /* 0-31 doorbells */
+#define GEN4_EM_SPAD_OFFSET		0x8080
+/* note, link status is now in MMIO and not config space for NTB */
+#define GEN4_LINK_CTRL_OFFSET		0xb050
+#define GEN4_LINK_STATUS_OFFSET		0xb052
+#define GEN4_PPD0_OFFSET		0xb0d4
+#define GEN4_PPD1_OFFSET		0xb4c0
+#define GEN4_LTSSMSTATEJMP		0xf040
+
+#define GEN4_PPD_CLEAR_TRN		0x0001
+#define GEN4_PPD_LINKTRN		0x0008
+#define GEN4_PPD_CONN_MASK		0x0300
+#define GEN4_PPD_CONN_B2B		0x0200
+#define GEN4_PPD_DEV_MASK		0x1000
+#define GEN4_PPD_DEV_DSD		0x1000
+#define GEN4_PPD_DEV_USD		0x0000
+#define GEN4_LINK_CTRL_LINK_DISABLE	0x0010
+
+#define GEN4_SLOTSTS			0xb05a
+#define GEN4_SLOTSTS_DLLSCS		0x100
+
+#define GEN4_PPD_TOPO_MASK	(GEN4_PPD_CONN_MASK | GEN4_PPD_DEV_MASK)
+#define GEN4_PPD_TOPO_B2B_USD	(GEN4_PPD_CONN_B2B | GEN4_PPD_DEV_USD)
+#define GEN4_PPD_TOPO_B2B_DSD	(GEN4_PPD_CONN_B2B | GEN4_PPD_DEV_DSD)
+
+#define GEN4_DB_COUNT			32
+#define GEN4_DB_LINK			32
+#define GEN4_DB_LINK_BIT		BIT_ULL(GEN4_DB_LINK)
+#define GEN4_DB_MSIX_VECTOR_COUNT	33
+#define GEN4_DB_MSIX_VECTOR_SHIFT	1
+#define GEN4_DB_TOTAL_SHIFT		33
+#define GEN4_SPAD_COUNT			16
+
+#define NTB_CTL_E2I_BAR23_SNOOP		0x000004
+#define NTB_CTL_E2I_BAR23_NOSNOOP	0x000008
+#define NTB_CTL_I2E_BAR23_SNOOP		0x000010
+#define NTB_CTL_I2E_BAR23_NOSNOOP	0x000020
+#define NTB_CTL_E2I_BAR45_SNOOP		0x000040
+#define NTB_CTL_E2I_BAR45_NOSNOO	0x000080
+#define NTB_CTL_I2E_BAR45_SNOOP		0x000100
+#define NTB_CTL_I2E_BAR45_NOSNOOP	0x000200
+#define NTB_CTL_BUSNO_DIS_INC		0x000400
+#define NTB_CTL_LINK_DOWN		0x010000
+
+#define NTB_SJC_FORCEDETECT		0x000004
+
+ssize_t ndev_ntb4_debugfs_read(struct file *filp, char __user *ubuf,
+				      size_t count, loff_t *offp);
+int gen4_init_dev(struct intel_ntb_dev *ndev);
+ssize_t ndev_ntb4_debugfs_read(struct file *filp, char __user *ubuf,
+				      size_t count, loff_t *offp);
+
+extern const struct ntb_dev_ops intel_ntb4_ops;
+
+static inline int pdev_is_ICX(struct pci_dev *pdev)
+{
+	if (pdev_is_gen4(pdev) &&
+	    pdev->revision >= PCI_DEVICE_REVISION_ICX_MIN &&
+	    pdev->revision <= PCI_DEVICE_REVISION_ICX_MAX)
+		return 1;
+	return 0;
+}
+
+#endif
diff --git a/drivers/ntb/hw/intel/ntb_hw_intel.h b/drivers/ntb/hw/intel/ntb_hw_intel.h
index e071e28bca3f..d61fcd91714b 100644
--- a/drivers/ntb/hw/intel/ntb_hw_intel.h
+++ b/drivers/ntb/hw/intel/ntb_hw_intel.h
@@ -72,6 +72,7 @@
 #define PCI_DEVICE_ID_INTEL_NTB_PS_BDX	0x6F0E
 #define PCI_DEVICE_ID_INTEL_NTB_SS_BDX	0x6F0F
 #define PCI_DEVICE_ID_INTEL_NTB_B2B_SKX	0x201C
+#define PCI_DEVICE_ID_INTEL_NTB_B2B_ICX	0x347e
 
 /* Ntb control and link status */
 #define NTB_CTL_CFG_LOCK		BIT(0)
@@ -120,6 +121,7 @@ struct intel_ntb_xlat_reg {
 	unsigned long			bar0_base;
 	unsigned long			bar2_xlat;
 	unsigned long			bar2_limit;
+	unsigned short			bar2_idx;
 };
 
 struct intel_b2b_addr {
@@ -182,6 +184,9 @@ struct intel_ntb_dev {
 
 	struct dentry			*debugfs_dir;
 	struct dentry			*debugfs_info;
+
+	/* gen4 entries */
+	int				dev_up;
 };
 
 #define ntb_ndev(__ntb) container_of(__ntb, struct intel_ntb_dev, ntb)
@@ -219,4 +224,11 @@ static inline int pdev_is_gen3(struct pci_dev *pdev)
 	return 0;
 }
 
+static inline int pdev_is_gen4(struct pci_dev *pdev)
+{
+	if (pdev->device == PCI_DEVICE_ID_INTEL_NTB_B2B_ICX)
+		return 1;
+
+	return 0;
+}
 #endif
diff --git a/drivers/ntb/test/ntb_perf.c b/drivers/ntb/test/ntb_perf.c
index 972f6d984f6d..89df1350fefd 100644
--- a/drivers/ntb/test/ntb_perf.c
+++ b/drivers/ntb/test/ntb_perf.c
@@ -101,8 +101,8 @@ MODULE_DESCRIPTION("PCIe NTB Performance Measurement Tool");
 #define DMA_MDELAY		10
 
 #define MSG_TRIES		1000
-#define MSG_UDELAY_LOW		1000
-#define MSG_UDELAY_HIGH		2000
+#define MSG_UDELAY_LOW		1000000
+#define MSG_UDELAY_HIGH		2000000
 
 #define PERF_BUF_LEN 1024
 
@@ -159,6 +159,8 @@ struct perf_peer {
 	/* NTB connection setup service */
 	struct work_struct	service;
 	unsigned long		sts;
+
+	struct completion init_comp;
 };
 #define to_peer_service(__work) \
 	container_of(__work, struct perf_peer, service)
@@ -547,6 +549,7 @@ static int perf_setup_outbuf(struct perf_peer *peer)
 
 	/* Initialization is finally done */
 	set_bit(PERF_STS_DONE, &peer->sts);
+	complete_all(&peer->init_comp);
 
 	return 0;
 }
@@ -557,7 +560,7 @@ static void perf_free_inbuf(struct perf_peer *peer)
 		return;
 
 	(void)ntb_mw_clear_trans(peer->perf->ntb, peer->pidx, peer->gidx);
-	dma_free_coherent(&peer->perf->ntb->dev, peer->inbuf_size,
+	dma_free_coherent(&peer->perf->ntb->pdev->dev, peer->inbuf_size,
 			  peer->inbuf, peer->inbuf_xlat);
 	peer->inbuf = NULL;
 }
@@ -586,8 +589,9 @@ static int perf_setup_inbuf(struct perf_peer *peer)
 
 	perf_free_inbuf(peer);
 
-	peer->inbuf = dma_alloc_coherent(&perf->ntb->dev, peer->inbuf_size,
-					 &peer->inbuf_xlat, GFP_KERNEL);
+	peer->inbuf = dma_alloc_coherent(&perf->ntb->pdev->dev,
+					 peer->inbuf_size, &peer->inbuf_xlat,
+					 GFP_KERNEL);
 	if (!peer->inbuf) {
 		dev_err(&perf->ntb->dev, "Failed to alloc inbuf of %pa\n",
 			&peer->inbuf_size);
@@ -637,6 +641,7 @@ static void perf_service_work(struct work_struct *work)
 		perf_setup_outbuf(peer);
 
 	if (test_and_clear_bit(PERF_CMD_CLEAR, &peer->sts)) {
+		init_completion(&peer->init_comp);
 		clear_bit(PERF_STS_DONE, &peer->sts);
 		if (test_bit(0, &peer->perf->busy_flag) &&
 		    peer == peer->perf->test_peer) {
@@ -653,7 +658,7 @@ static int perf_init_service(struct perf_ctx *perf)
 {
 	u64 mask;
 
-	if (ntb_peer_mw_count(perf->ntb) < perf->pcnt + 1) {
+	if (ntb_peer_mw_count(perf->ntb) < perf->pcnt) {
 		dev_err(&perf->ntb->dev, "Not enough memory windows\n");
 		return -EINVAL;
 	}
@@ -803,7 +808,7 @@ static int perf_copy_chunk(struct perf_thread *pthr,
 	dst_vaddr = dst;
 	dst_dma_addr = peer->dma_dst_addr + (dst_vaddr - vbase);
 
-	unmap = dmaengine_get_unmap_data(dma_dev, 2, GFP_NOWAIT);
+	unmap = dmaengine_get_unmap_data(dma_dev, 1, GFP_NOWAIT);
 	if (!unmap)
 		return -ENOMEM;
 
@@ -816,15 +821,8 @@ static int perf_copy_chunk(struct perf_thread *pthr,
 	}
 	unmap->to_cnt = 1;
 
-	unmap->addr[1] = dst_dma_addr;
-	if (dma_mapping_error(dma_dev, unmap->addr[1])) {
-		ret = -EIO;
-		goto err_free_resource;
-	}
-	unmap->from_cnt = 1;
-
 	do {
-		tx = dmaengine_prep_dma_memcpy(pthr->dma_chan, unmap->addr[1],
+		tx = dmaengine_prep_dma_memcpy(pthr->dma_chan, dst_dma_addr,
 			unmap->addr[0], len, DMA_PREP_INTERRUPT | DMA_CTRL_ACK);
 		if (!tx)
 			msleep(DMA_MDELAY);
@@ -1010,8 +1008,8 @@ static void perf_clear_test(struct perf_thread *pthr)
 				   pthr->perf->test_peer->dma_dst_addr,
 				   pthr->perf->test_peer->outbuf_size,
 				   DMA_FROM_DEVICE, 0);
-	if (pthr->dma_chan)
-		dma_release_channel(pthr->dma_chan);
+
+	dma_release_channel(pthr->dma_chan);
 
 no_dma_notify:
 	atomic_dec(&perf->tsync);
@@ -1083,8 +1081,9 @@ static int perf_submit_test(struct perf_peer *peer)
 	struct perf_thread *pthr;
 	int tidx, ret;
 
-	if (!test_bit(PERF_STS_DONE, &peer->sts))
-		return -ENOLINK;
+	ret = wait_for_completion_interruptible(&peer->init_comp);
+	if (ret < 0)
+		return ret;
 
 	if (test_and_set_bit_lock(0, &perf->busy_flag))
 		return -EBUSY;
@@ -1455,10 +1454,21 @@ static int perf_init_peers(struct perf_ctx *perf)
 			peer->gidx = pidx;
 		}
 		INIT_WORK(&peer->service, perf_service_work);
+		init_completion(&peer->init_comp);
 	}
 	if (perf->gidx == -1)
 		perf->gidx = pidx;
 
+	/*
+	 * Hardware with only two ports may not have unique port
+	 * numbers. In this case, the gidxs should all be zero.
+	 */
+	if (perf->pcnt == 1 &&  ntb_port_number(perf->ntb) == 0 &&
+	    ntb_peer_port_number(perf->ntb, 0) == 0) {
+		perf->gidx = 0;
+		perf->peers[0].gidx = 0;
+	}
+
 	for (pidx = 0; pidx < perf->pcnt; pidx++) {
 		ret = perf_setup_peer_mw(&perf->peers[pidx]);
 		if (ret)
@@ -1554,4 +1564,3 @@ static void __exit perf_exit(void)
 	destroy_workqueue(perf_wq);
 }
 module_exit(perf_exit);
-
diff --git a/drivers/ntb/test/ntb_pingpong.c b/drivers/ntb/test/ntb_pingpong.c
index 04dd46647db3..2164e8492772 100644
--- a/drivers/ntb/test/ntb_pingpong.c
+++ b/drivers/ntb/test/ntb_pingpong.c
@@ -121,15 +121,14 @@ static int pp_find_next_peer(struct pp_ctx *pp)
 	link = ntb_link_is_up(pp->ntb, NULL, NULL);
 
 	/* Find next available peer */
-	if (link & pp->nmask) {
+	if (link & pp->nmask)
 		pidx = __ffs64(link & pp->nmask);
-		out_db = BIT_ULL(pidx + 1);
-	} else if (link & pp->pmask) {
+	else if (link & pp->pmask)
 		pidx = __ffs64(link & pp->pmask);
-		out_db = BIT_ULL(pidx);
-	} else {
+	else
 		return -ENODEV;
-	}
+
+	out_db = BIT_ULL(ntb_peer_port_number(pp->ntb, pidx));
 
 	spin_lock(&pp->lock);
 	pp->out_pidx = pidx;
@@ -303,7 +302,7 @@ static void pp_init_flds(struct pp_ctx *pp)
 			break;
 	}
 
-	pp->in_db = BIT_ULL(pidx);
+	pp->in_db = BIT_ULL(lport);
 	pp->pmask = GENMASK_ULL(pidx, 0) >> 1;
 	pp->nmask = GENMASK_ULL(pcnt - 1, pidx);
 
@@ -432,4 +431,3 @@ static void __exit pp_exit(void)
 	debugfs_remove_recursive(pp_dbgfs_topdir);
 }
 module_exit(pp_exit);
-
diff --git a/drivers/ntb/test/ntb_tool.c b/drivers/ntb/test/ntb_tool.c
index 69da758fe64c..b7bf3f863d79 100644
--- a/drivers/ntb/test/ntb_tool.c
+++ b/drivers/ntb/test/ntb_tool.c
@@ -504,7 +504,7 @@ static ssize_t tool_peer_link_read(struct file *filep, char __user *ubuf,
 	buf[1] = '\n';
 	buf[2] = '\0';
 
-	return simple_read_from_buffer(ubuf, size, offp, buf, 3);
+	return simple_read_from_buffer(ubuf, size, offp, buf, 2);
 }
 
 static TOOL_FOPS_RDWR(tool_peer_link_fops,
@@ -590,7 +590,7 @@ static int tool_setup_mw(struct tool_ctx *tc, int pidx, int widx,
 	inmw->size = min_t(resource_size_t, req_size, size);
 	inmw->size = round_up(inmw->size, addr_align);
 	inmw->size = round_up(inmw->size, size_align);
-	inmw->mm_base = dma_alloc_coherent(&tc->ntb->dev, inmw->size,
+	inmw->mm_base = dma_alloc_coherent(&tc->ntb->pdev->dev, inmw->size,
 					   &inmw->dma_base, GFP_KERNEL);
 	if (!inmw->mm_base)
 		return -ENOMEM;
@@ -612,7 +612,7 @@ static int tool_setup_mw(struct tool_ctx *tc, int pidx, int widx,
 	return 0;
 
 err_free_dma:
-	dma_free_coherent(&tc->ntb->dev, inmw->size, inmw->mm_base,
+	dma_free_coherent(&tc->ntb->pdev->dev, inmw->size, inmw->mm_base,
 			  inmw->dma_base);
 	inmw->mm_base = NULL;
 	inmw->dma_base = 0;
@@ -629,7 +629,7 @@ static void tool_free_mw(struct tool_ctx *tc, int pidx, int widx)
 
 	if (inmw->mm_base != NULL) {
 		ntb_mw_clear_trans(tc->ntb, pidx, widx);
-		dma_free_coherent(&tc->ntb->dev, inmw->size,
+		dma_free_coherent(&tc->ntb->pdev->dev, inmw->size,
 				  inmw->mm_base, inmw->dma_base);
 	}
 
@@ -1690,4 +1690,3 @@ static void __exit tool_exit(void)
 	debugfs_remove_recursive(tool_dbgfs_topdir);
 }
 module_exit(tool_exit);
-