lists.openwall.net   lists  /  announce  owl-users  owl-dev  john-users  john-dev  passwdqc-users  yescrypt  popa3d-users  /  oss-security  kernel-hardening  musl  sabotage  tlsify  passwords  /  crypt-dev  xvendor  /  Bugtraq  Full-Disclosure  linux-kernel  linux-netdev  linux-ext4  linux-hardening  linux-cve-announce  PHC 
Open Source and information security mailing list archives
 
Hash Suite: Windows password security audit tool. GUI, reports in PDF.
[<prev] [next>] [day] [month] [year] [list]
Message-Id: <20200616020323.1227-1-yunaixin03610@163.com>
Date:   Tue, 16 Jun 2020 10:03:23 +0800
From:   yunaixin03610@....com
To:     linux-kernel@...r.kernel.org
Cc:     yunaixin <yunaixin@...wei.com>
Subject: [PATCH 4/5] Huawei BMA: Adding Huawei BMA driver: cdev_veth_drv

From: yunaixin <yunaixin@...wei.com>

The BMA software is a system management software offered by Huawei. It supports the status monitoring, performance monitoring, and event monitoring of various components, including server CPUs, memory, hard disks, NICs, IB cards, PCIe cards, RAID controller cards, and optical modules.

This cdev_veth_drv driver is one of the communication drivers used by BMA software. It depends on the host_edma_drv driver. It will create a char device once loaded, offering interfaces (open, close, read, write and poll) to BMA to send/receive RESTful messages between BMC software. When the message is longer than 1KB, it will be cut into packets of 1KB. The other side, BMC's cdev_veth driver, will assemble these packets back into original mesages.

Signed-off-by: yunaixin <yunaixin@...wei.com>
---
 drivers/net/ethernet/huawei/bma/Kconfig       |    3 +-
 drivers/net/ethernet/huawei/bma/Makefile      |    3 +-
 .../ethernet/huawei/bma/cdev_veth_drv/Kconfig |   11 +
 .../huawei/bma/cdev_veth_drv/Makefile         |    2 +
 .../bma/cdev_veth_drv/virtual_cdev_eth_net.c  | 1839 +++++++++++++++++
 .../bma/cdev_veth_drv/virtual_cdev_eth_net.h  |  300 +++
 6 files changed, 2156 insertions(+), 2 deletions(-)
 create mode 100644 drivers/net/ethernet/huawei/bma/cdev_veth_drv/Kconfig
 create mode 100644 drivers/net/ethernet/huawei/bma/cdev_veth_drv/Makefile
 create mode 100644 drivers/net/ethernet/huawei/bma/cdev_veth_drv/virtual_cdev_eth_net.c
 create mode 100644 drivers/net/ethernet/huawei/bma/cdev_veth_drv/virtual_cdev_eth_net.h

diff --git a/drivers/net/ethernet/huawei/bma/Kconfig b/drivers/net/ethernet/huawei/bma/Kconfig
index 21e9f7ee5be6..90b267c265e0 100644
--- a/drivers/net/ethernet/huawei/bma/Kconfig
+++ b/drivers/net/ethernet/huawei/bma/Kconfig
@@ -1,3 +1,4 @@
 source "drivers/net/ethernet/huawei/bma/edma_drv/Kconfig"
 source "drivers/net/ethernet/huawei/bma/cdev_drv/Kconfig"
-source "drivers/net/ethernet/huawei/bma/veth_drv/Kconfig"
\ No newline at end of file
+source "drivers/net/ethernet/huawei/bma/veth_drv/Kconfig"
+source "drivers/net/ethernet/huawei/bma/cdev_veth_drv/Kconfig"
\ No newline at end of file
diff --git a/drivers/net/ethernet/huawei/bma/Makefile b/drivers/net/ethernet/huawei/bma/Makefile
index 6ff24d31c650..c626618f47fb 100644
--- a/drivers/net/ethernet/huawei/bma/Makefile
+++ b/drivers/net/ethernet/huawei/bma/Makefile
@@ -4,4 +4,5 @@
 
 obj-$(CONFIG_BMA) += edma_drv/
 obj-$(CONFIG_BMA) += cdev_drv/
-obj-$(CONFIG_BMA) += veth_drv/
\ No newline at end of file
+obj-$(CONFIG_BMA) += veth_drv/
+obj-$(CONFIG_BMA) += cdev_veth_drv/
\ No newline at end of file
diff --git a/drivers/net/ethernet/huawei/bma/cdev_veth_drv/Kconfig b/drivers/net/ethernet/huawei/bma/cdev_veth_drv/Kconfig
new file mode 100644
index 000000000000..97829c5487c2
--- /dev/null
+++ b/drivers/net/ethernet/huawei/bma/cdev_veth_drv/Kconfig
@@ -0,0 +1,11 @@
+#
+# Huawei BMA software driver configuration
+#
+
+config BMA
+	tristate "Huawei BMA Software Communication Driver"
+
+	---help---
+	  This driver supports Huawei BMA Software. It is used 
+	  to communication between Huawei BMA and BMC software.
+
diff --git a/drivers/net/ethernet/huawei/bma/cdev_veth_drv/Makefile b/drivers/net/ethernet/huawei/bma/cdev_veth_drv/Makefile
new file mode 100644
index 000000000000..a0f73e4009f7
--- /dev/null
+++ b/drivers/net/ethernet/huawei/bma/cdev_veth_drv/Makefile
@@ -0,0 +1,2 @@
+obj-$(CONFIG_BMA) += cdev_veth_drv.o
+cdev_veth_drv-y := virtual_cdev_eth_net.o
diff --git a/drivers/net/ethernet/huawei/bma/cdev_veth_drv/virtual_cdev_eth_net.c b/drivers/net/ethernet/huawei/bma/cdev_veth_drv/virtual_cdev_eth_net.c
new file mode 100644
index 000000000000..1ab720bb7412
--- /dev/null
+++ b/drivers/net/ethernet/huawei/bma/cdev_veth_drv/virtual_cdev_eth_net.c
@@ -0,0 +1,1839 @@
+// SPDX-License-Identifier: GPL-2.0
+/* Huawei iBMA driver.
+ * Copyright (c) 2019, Huawei Technologies Co., Ltd.
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public License
+ * as published by the Free Software Foundation; either version 2
+ * of the License, or (at your option) any later version.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ * GNU General Public License for more details.
+ */
+
+#include "virtual_cdev_eth_net.h"
+
+static struct edma_eth_dev_s g_eth_edmaprivate;
+static struct edma_packet_node_s g_edma_recv_packet_tmp = {0, NULL};
+static struct edma_cut_packet_node_s *g_edma_send_cut_packet;
+static unsigned int g_last_token = TK_START_END;
+static unsigned int g_device_opened = CDEV_CLOSED;
+static unsigned int g_last_number;
+static unsigned int g_peer_not_ready;
+static unsigned int g_read_pos;
+static unsigned int g_delay_ms;
+static int g_write_count;
+
+static int cdev_open(struct inode *inode_ptr, struct file *filp);
+static int cdev_release(struct inode *inode_ptr, struct file *filp);
+static unsigned int cdev_poll(struct file *file, poll_table *wait);
+static ssize_t cdev_read(struct file *filp, char __user *data,
+			 size_t count, loff_t *ppos);
+static ssize_t cdev_write(struct file *filp, const char __user *data,
+			  size_t count, loff_t *ppos);
+
+#define IS_CDEV_IN_OPEN_STATE() (g_device_opened != CDEV_CLOSED)
+#define SET_CDEV_OPEN_STATE(x) (g_device_opened = (x))
+
+int debug = DLOG_ERROR;
+module_param(debug, int, 0644);
+MODULE_PARM_DESC(debug, "Debug switch (0=close debug, 1=open debug)");
+
+#define GET_PRIVATE_DATA(f) (((struct cdev_dev_s *)((f)->private_data))->priv)
+
+const struct file_operations g_eth_edma_cdev_fops = {
+	.owner = THIS_MODULE,
+	.open = cdev_open,
+	.release = cdev_release,
+	.poll = cdev_poll,
+	.read = cdev_read,
+	.write = cdev_write,
+};
+
+void dump_global_info(void)
+{
+	struct edma_shmq_hd_s *pshmqhd_v = NULL;
+
+	if (!debug)
+		return;
+
+	LOG(DLOG_DEBUG, "\r\n=================VETH INFO=================\r\n");
+
+	pshmqhd_v = g_eth_edmaprivate.ptx_queue->pshmqhd_v;
+	LOG(DLOG_DEBUG, "TX head/tail: %u/%u ------------",
+	    pshmqhd_v->head, pshmqhd_v->tail);
+
+	pshmqhd_v = g_eth_edmaprivate.prx_queue->pshmqhd_v;
+	LOG(DLOG_DEBUG, "RX head/tail: %u/%u ------------",
+	    pshmqhd_v->head, pshmqhd_v->tail);
+}
+
+static inline int edma_is_queue_ready(struct edma_rxtx_q_s *prxtx_queue)
+{
+	if (!prxtx_queue)
+		return 0;
+
+	return (prxtx_queue->pshmqhd_v->init == BSPVETH_SHMQUEUE_INITOK_V2);
+}
+
+static inline void edma_veth_host_addr_init(void *priv)
+{
+	struct bma_priv_data_s *edma_priv = (struct bma_priv_data_s *)priv;
+
+	g_eth_edmaprivate.pshmpool_p =
+			(u8 *)edma_priv->specific.veth.veth_swap_phy_addr;
+	g_eth_edmaprivate.pshmpool_v =
+			(u8 *)edma_priv->specific.veth.veth_swap_addr;
+	g_eth_edmaprivate.shmpoolsize =
+			(u32)edma_priv->specific.veth.veth_swap_len;
+}
+
+void edma_veth_free_tx_resources(struct edma_rxtx_q_s *ptx_queue)
+{
+	struct edma_bd_info_s *pbdinfobase_v = NULL;
+
+	if (!ptx_queue || !ptx_queue->pbdinfobase_v)
+		return;
+
+	pbdinfobase_v = ptx_queue->pbdinfobase_v;
+	ptx_queue->pbdinfobase_v = NULL;
+	ptx_queue->pshmqhd_v = NULL;
+
+	vfree(pbdinfobase_v);
+
+	LOG(DLOG_DEBUG, "%s ok. count=%d", __func__, ptx_queue->count);
+}
+
+void edma_veth_free_all_tx_resources(struct edma_eth_dev_s *edma_eth)
+{
+	if (edma_eth && edma_eth->ptx_queue) {
+		edma_veth_free_tx_resources(edma_eth->ptx_queue);
+		kfree(edma_eth->ptx_queue);
+		edma_eth->ptx_queue = NULL;
+	}
+}
+
+int edma_veth_setup_tx_resources(struct edma_rxtx_q_s *ptx_queue)
+{
+	int size;
+
+	ptx_queue->count = MAX_QUEUE_BDNUM;
+	size = sizeof(struct edma_bd_info_s) * ptx_queue->count;
+
+	ptx_queue->pbdinfobase_v = vmalloc(size);
+	if (!ptx_queue->pbdinfobase_v) {
+		LOG(DLOG_ERROR, "Failed to alloc memory for the TX queue.");
+		return -ENOMEM;
+	}
+
+	memset(ptx_queue->pbdinfobase_v, 0, size);
+
+	/* round up to nearest 4K */
+	size = sizeof(struct edma_dma_shmbd_s) * ptx_queue->count;
+	ptx_queue->size = ALIGN(size, ALIGN_MASK);
+
+	ptx_queue->work_limit = BSPVETH_WORK_LIMIT;
+
+	return 0;
+}
+
+int edma_veth_setup_all_tx_resources(struct edma_eth_dev_s *edma_eth)
+{
+	int err;
+	u8 *shmq_head = NULL;
+	u8 *shmq_head_p = NULL;
+	struct edma_rxtx_q_s *tx_queue = NULL;
+
+	tx_queue = (struct edma_rxtx_q_s *)
+		   kmalloc(sizeof(struct edma_rxtx_q_s), GFP_KERNEL);
+	if (!tx_queue) {
+		LOG(DLOG_ERROR, "Failed to alloc TX queue.");
+		return -ENOMEM;
+	}
+
+	memset(tx_queue, 0, sizeof(struct edma_rxtx_q_s));
+
+	shmq_head = edma_eth->pshmpool_v + (MAX_SHAREQUEUE_SIZE * 0);
+	shmq_head_p = edma_eth->pshmpool_p + (MAX_SHAREQUEUE_SIZE * 0);
+
+	tx_queue->pshmqhd_v = (struct edma_shmq_hd_s *)shmq_head;
+	tx_queue->pshmqhd_p = shmq_head_p;
+
+	tx_queue->pshmbdbase_v = (struct edma_dma_shmbd_s *)
+				 (shmq_head + BSPVETH_SHMBDBASE_OFFSET);
+	tx_queue->pshmbdbase_p = shmq_head_p + BSPVETH_SHMBDBASE_OFFSET;
+
+	tx_queue->pdmalbase_v = (struct edma_dmal_s *)
+				(shmq_head + SHMDMAL_OFFSET);
+	tx_queue->pdmalbase_p = (u8 *)(VETH_SHAREPOOL_BASE_INBMC +
+				(MAX_SHAREQUEUE_SIZE * 0) + SHMDMAL_OFFSET);
+
+	memset(tx_queue->pdmalbase_v, 0, MAX_SHMDMAL_SIZE);
+
+	err = edma_veth_setup_tx_resources(tx_queue);
+	if (err) {
+		kfree(tx_queue);
+		return err;
+	}
+
+	edma_eth->ptx_queue = tx_queue;
+
+	return 0;
+}
+
+int edma_veth_setup_rx_resources(struct edma_rxtx_q_s *prx_queue)
+{
+	int size;
+
+	prx_queue->count = MAX_QUEUE_BDNUM;
+	size = sizeof(struct edma_bd_info_s) * prx_queue->count;
+
+	prx_queue->pbdinfobase_v = vmalloc(size);
+	if (!prx_queue->pbdinfobase_v) {
+		LOG(DLOG_ERROR, "Failed to alloc memory for the RX queue.");
+		return -ENOMEM;
+	}
+
+	memset(prx_queue->pbdinfobase_v, 0, size);
+
+	/* Round up to nearest 4K */
+	size = sizeof(struct edma_dma_shmbd_s) * prx_queue->count;
+	prx_queue->size = ALIGN(size, ALIGN_MASK);
+
+	prx_queue->work_limit = BSPVETH_WORK_LIMIT;
+
+	return 0;
+}
+
+int edma_veth_setup_all_rx_resources(struct edma_eth_dev_s *edma_eth)
+{
+	int err;
+	u8 *shmq_head = NULL;
+	u8 *shmq_head_p = NULL;
+	struct edma_rxtx_q_s *rx_queue = NULL;
+
+	rx_queue = (struct edma_rxtx_q_s *)
+		   kmalloc(sizeof(struct edma_rxtx_q_s), GFP_KERNEL);
+	if (!rx_queue) {
+		LOG(DLOG_ERROR, "Failed to alloc RX queue.");
+		return -ENOMEM;
+	}
+
+	memset(rx_queue, 0, sizeof(struct edma_rxtx_q_s));
+
+	shmq_head = edma_eth->pshmpool_v + MAX_SHAREQUEUE_SIZE;
+	shmq_head_p = edma_eth->pshmpool_p + MAX_SHAREQUEUE_SIZE;
+	rx_queue->pshmqhd_v = (struct edma_shmq_hd_s *)shmq_head;
+	rx_queue->pshmqhd_p = shmq_head_p;
+
+	rx_queue->pshmbdbase_v = (struct edma_dma_shmbd_s *)(shmq_head +
+				  BSPVETH_SHMBDBASE_OFFSET);
+	rx_queue->pshmbdbase_p = shmq_head_p + BSPVETH_SHMBDBASE_OFFSET;
+
+	/* DMA address list (only used in host). */
+	rx_queue->pdmalbase_v = (struct edma_dmal_s *)
+				(shmq_head + SHMDMAL_OFFSET);
+	rx_queue->pdmalbase_p = (u8 *)(VETH_SHAREPOOL_BASE_INBMC +
+				MAX_SHAREQUEUE_SIZE + SHMDMAL_OFFSET);
+	memset(rx_queue->pdmalbase_v, 0, MAX_SHMDMAL_SIZE);
+
+	err = edma_veth_setup_rx_resources(rx_queue);
+	if (err) {
+		kfree(rx_queue);
+		return err;
+	}
+
+	edma_eth->prx_queue = rx_queue;
+
+	return 0;
+}
+
+void edma_veth_free_rx_resources(struct edma_rxtx_q_s *prx_queue)
+{
+	struct edma_bd_info_s *pbdinfobase_v = NULL;
+
+	if (!prx_queue || !prx_queue->pbdinfobase_v)
+		return;
+
+	pbdinfobase_v = prx_queue->pbdinfobase_v;
+	prx_queue->pbdinfobase_v = NULL;
+	prx_queue->pshmqhd_v = NULL;
+
+	/* Free all the Rx ring pages */
+	vfree(pbdinfobase_v);
+
+	LOG(DLOG_DEBUG, "%s ok. count=%d", __func__, prx_queue->count);
+}
+
+void edma_veth_free_all_rx_resources(struct edma_eth_dev_s *edma_eth)
+{
+	if (edma_eth && edma_eth->prx_queue) {
+		edma_veth_free_rx_resources(edma_eth->prx_queue);
+		kfree(edma_eth->prx_queue);
+		edma_eth->prx_queue = NULL;
+	}
+}
+
+int edma_veth_setup_all_rxtx_queue(struct edma_eth_dev_s *edma_eth)
+{
+	void *buf = NULL;
+	unsigned int i;
+	unsigned int j;
+
+	dma_addr_t dmaaddr;
+	struct edma_bd_info_s *pbdinfobase_v = NULL;
+
+	struct edma_rxtx_q_s *ptx_queue = NULL;
+	struct edma_rxtx_q_s *prx_queue = NULL;
+
+	struct bma_priv_data_s *priv = NULL;
+	struct device *dev = NULL;
+
+	priv = (struct bma_priv_data_s *)edma_eth->edma_priv;
+	dev = &priv->specific.veth.pdev->dev;
+
+	ptx_queue = edma_eth->ptx_queue;
+	prx_queue = edma_eth->prx_queue;
+
+	edma_eth->pages_tx = 0;
+
+	pbdinfobase_v = ptx_queue->pbdinfobase_v;
+
+	for (i = 0; i < MAX_QUEUE_BDNUM; i++) {
+		buf = kmalloc(NODE_SIZE, GFP_KERNEL | GFP_DMA);
+		if (!buf) {
+			for (j = 0; j < i; j++)
+				kfree((void *)pbdinfobase_v[j].pdma_v);
+
+			LOG(DLOG_ERROR, "Fail to alloc tx buf.");
+			return -ENOMEM;
+		}
+
+		dmaaddr = dma_map_single(dev, buf, NODE_SIZE, DMA_TO_DEVICE);
+		if (dma_mapping_error(dev, dmaaddr)) {
+			LOG(DLOG_ERROR, "Failed to map tx DMA address.");
+			kfree(buf);
+			return -EIO;
+		}
+
+		memset(buf, 0xFF, NODE_SIZE);
+
+		pbdinfobase_v[i].pdma_v = (u8 *)(buf);
+		pbdinfobase_v[i].dma_p = dmaaddr;
+		pbdinfobase_v[i].len = NODE_SIZE;
+	}
+
+	LOG(DLOG_DEBUG, "set tx done.");
+
+	edma_eth->pages_rx = 0;
+
+	pbdinfobase_v = prx_queue->pbdinfobase_v;
+
+	for (i = 0; i < MAX_QUEUE_BDNUM; i++) {
+		buf = kmalloc(NODE_SIZE, GFP_KERNEL | GFP_DMA);
+		if (!buf) {
+			for (j = 0; j < i; j++)
+				kfree((void *)pbdinfobase_v[j].pdma_v);
+
+			LOG(DLOG_ERROR, "Fail to alloc rx buf.");
+			return -ENOMEM;
+		}
+
+		dmaaddr = dma_map_single(dev, buf, NODE_SIZE, DMA_FROM_DEVICE);
+		if (dma_mapping_error(dev, dmaaddr)) {
+			LOG(DLOG_ERROR, "Failed to map rx DMA address.");
+			kfree(buf);
+			return -EIO;
+		}
+
+		memset(buf, 0xFF, NODE_SIZE);
+
+		pbdinfobase_v[i].pdma_v = (u8 *)(buf);
+		pbdinfobase_v[i].dma_p = dmaaddr;
+		pbdinfobase_v[i].len = NODE_SIZE;
+	}
+
+	LOG(DLOG_DEBUG, "set rx done.");
+
+	return 0;
+}
+
+void edma_veth_dump(void)
+{
+	struct edma_eth_dev_s *edma_eth = &g_eth_edmaprivate;
+	struct edma_rxtx_q_s *ptx_queue = edma_eth->ptx_queue;
+	struct edma_rxtx_q_s *prx_queue = edma_eth->prx_queue;
+	struct edma_shmq_hd_s *pshmq_head = NULL;
+
+	if (!debug)
+		return;
+
+	pshmq_head = prx_queue->pshmqhd_v;
+
+	LOG(DLOG_DEBUG,
+	    "RX host_head:%u, host_tail:%u, shm_head:%u, shm_tail:%u, ",
+	    prx_queue->head, prx_queue->tail,
+	    pshmq_head->head, pshmq_head->tail);
+	LOG(DLOG_DEBUG, "count: %u, total: %u, init: %u.",
+	    pshmq_head->count, pshmq_head->total, pshmq_head->init);
+
+	pshmq_head = ptx_queue->pshmqhd_v;
+
+	LOG(DLOG_DEBUG,
+	    "TX host_head:%u, host_tail:%u, shm_head:%u, shm_tail:%u, ",
+	    ptx_queue->head, ptx_queue->tail,
+	    pshmq_head->head, pshmq_head->tail);
+	LOG(DLOG_DEBUG, "count: %u, total: %u, init: %u.",
+	    pshmq_head->count, pshmq_head->total, pshmq_head->init);
+}
+
+int edma_veth_setup_resource(struct edma_eth_dev_s *edma_eth)
+{
+	int err;
+
+	err = edma_veth_setup_all_rx_resources(edma_eth);
+	if (err < 0)
+		return err;
+
+	err = edma_veth_setup_all_tx_resources(edma_eth);
+	if (err < 0)
+		goto FREE_RX;
+
+	err = edma_veth_setup_all_rxtx_queue(edma_eth);
+	if (err < 0)
+		goto FREE_TX;
+
+	return 0;
+
+FREE_TX:
+	edma_veth_free_all_tx_resources(edma_eth);
+FREE_RX:
+	edma_veth_free_all_rx_resources(edma_eth);
+
+	return err;
+}
+
+int edma_veth_free_rxtx_queue(struct edma_eth_dev_s *edma_eth)
+{
+	int i;
+	struct edma_rxtx_q_s *ptx_queue = NULL;
+	struct edma_rxtx_q_s *prx_queue = NULL;
+
+	struct bma_priv_data_s *priv = NULL;
+	struct device *dev = NULL;
+
+	struct edma_bd_info_s *pbdinfobase_v = NULL;
+
+	if (!edma_eth || !edma_eth->edma_priv)
+		return 0;
+
+	priv = (struct bma_priv_data_s *)edma_eth->edma_priv;
+	dev = &priv->specific.veth.pdev->dev;
+
+	ptx_queue = edma_eth->ptx_queue;
+	prx_queue = edma_eth->prx_queue;
+
+	pbdinfobase_v = ptx_queue->pbdinfobase_v;
+
+	for (i = 0; i < MAX_QUEUE_BDNUM; i++) {
+		dma_unmap_single(dev, pbdinfobase_v[i].dma_p,
+				 NODE_SIZE, DMA_TO_DEVICE);
+		kfree(pbdinfobase_v[i].pdma_v);
+	}
+
+	pbdinfobase_v = prx_queue->pbdinfobase_v;
+
+	for (i = 0; i < MAX_QUEUE_BDNUM; i++) {
+		dma_unmap_single(dev, pbdinfobase_v[i].dma_p,
+				 NODE_SIZE, DMA_FROM_DEVICE);
+		kfree(pbdinfobase_v[i].pdma_v);
+	}
+
+	return 0;
+}
+
+void edma_veth_free_resource(struct edma_eth_dev_s *edma_eth)
+{
+	edma_veth_free_rxtx_queue(edma_eth);
+	LOG(DLOG_DEBUG, "edma_veth_free_rxtx_queue done.");
+
+	edma_veth_free_all_rx_resources(edma_eth);
+	LOG(DLOG_DEBUG, "edma_veth_free_all_rx_resources done.");
+
+	edma_veth_free_all_tx_resources(edma_eth);
+	LOG(DLOG_DEBUG, "edma_veth_free_all_tx_resources done.");
+}
+
+int edma_veth_send_one_pkt(struct edma_cut_packet_node_s *cut_packet_node)
+{
+	u32 head, tail;
+	struct edma_bd_info_s *pbdinfo_v = NULL;
+	struct edma_rxtx_q_s *ptx_queue = g_eth_edmaprivate.ptx_queue;
+	struct bma_priv_data_s *priv = NULL;
+	struct device *dev = NULL;
+
+	if (!cut_packet_node || !ptx_queue || !ptx_queue->pshmbdbase_v) {
+		LOG(DLOG_ERROR, "Invalid packet node.");
+		return -EFAULT;
+	}
+
+	priv = (struct bma_priv_data_s *)(g_eth_edmaprivate.edma_priv);
+	dev = &priv->specific.veth.pdev->dev;
+
+	if (!bma_intf_is_link_ok()) {
+		LOG(DLOG_ERROR, "EDMA link is not ready.");
+		return -EIO;
+	}
+
+	head = ptx_queue->head;
+	tail = ptx_queue->tail;
+
+	LOG(DLOG_DEBUG, "TX queue, before: head/tail: %u/%u", head, tail);
+
+	if (!JUDGE_RING_QUEUE_SPACE(head, tail, 1))
+		return -EBUSY;
+
+	ptx_queue->head = (head + 1) & BSPVETH_POINT_MASK;
+
+	pbdinfo_v = ptx_queue->pbdinfobase_v + head;
+
+	pbdinfo_v->len = NODE_TO_PACKET_SIZE(cut_packet_node);
+	(void)memcpy(pbdinfo_v->pdma_v, cut_packet_node, pbdinfo_v->len);
+
+	/* Force sync data from CPU to device. */
+	dma_sync_single_for_device(dev, pbdinfo_v->dma_p,
+				   pbdinfo_v->len, DMA_TO_DEVICE);
+
+	LOG(DLOG_DEBUG, "TX queue, after: head/tail: %u -> %u\n",
+	    ptx_queue->head, ptx_queue->tail);
+
+	return 0;
+}
+
+static inline unsigned int edma_veth_get_ring_buf_count(unsigned int head,
+							unsigned int tail,
+							unsigned int size)
+{
+	return (tail + size - head) % size;
+}
+
+static inline void edma_veth_flush_ring_node(struct edma_packet_node_s *node,
+					     unsigned int ring_len)
+{
+	unsigned int i;
+
+	for (i = 0; i < ring_len; i++) {
+		kfree(node[i].packet);
+		node[i].packet = NULL;
+	}
+}
+
+static int get_peer_queue_stress(struct edma_rxtx_q_s *queue)
+{
+	int stress;
+
+	if (++g_write_count < RL_MAX_PACKET) {
+		/* not enough packets, use the last delay. */
+		return -1;
+	}
+
+	g_write_count = 0;
+
+	/* check peer rx queue stress. */
+	if (!queue || queue->pshmqhd_v->total == 0) {
+		/* no rate limit allowed. */
+		return 0;
+	}
+
+	stress = (int)((queue->pshmqhd_v->count * STRESS_FACTOR) /
+			queue->pshmqhd_v->total);
+
+	return stress;
+}
+
+static void do_queue_rate_limit(struct edma_rxtx_q_s *queue)
+{
+	unsigned long delay_jiffies;
+	int stress = get_peer_queue_stress(queue);
+
+	LOG(DLOG_DEBUG, "count: %u, total: %u, stress: %d",
+	    queue->pshmqhd_v->count, queue->pshmqhd_v->total, stress);
+
+	if (stress >= RL_STRESS_HIGH)
+		g_delay_ms = RL_DELAY_MS_HIGH;
+	else if (stress >= RL_STRESS_LOW)
+		g_delay_ms = RL_DELAY_MS_LOW;
+	else if (stress >= 0)
+		g_delay_ms = 0;
+
+	if (g_delay_ms) {
+		delay_jiffies = msecs_to_jiffies(g_delay_ms);
+		schedule_timeout_killable(delay_jiffies);
+	}
+}
+
+static int edma_veth_cut_tx_packet_send(struct edma_eth_dev_s *eth_dev,
+					const char __user *data, size_t len)
+{
+	int ret = 0;
+	struct edma_cut_packet_node_s *tx_cut_pkt = g_edma_send_cut_packet;
+	unsigned int length = len;
+	unsigned int already_read_len = 0;
+	unsigned int count = 0;
+
+	if (!tx_cut_pkt)
+		return -EFAULT;
+
+	do_queue_rate_limit(eth_dev->ptx_queue);
+
+	while (length > 0) {
+		LOG(DLOG_DEBUG, "length: %u/%u", length, len);
+
+		if (length > BSPPACKET_MTU_MAX) {
+			/* fragment. */
+			if (copy_from_user(tx_cut_pkt->cut_packet,
+					   data + already_read_len,
+					   BSPPACKET_MTU_MAX)) {
+				LOG(DLOG_DEBUG, "Failed to copy user data.");
+				return -EFAULT;
+			}
+			tx_cut_pkt->number = count++;
+			length = length - BSPPACKET_MTU_MAX;
+
+			if (tx_cut_pkt->number == 0) {
+				tx_cut_pkt->token = TK_START_PACKET;
+				tx_cut_pkt->cut_packet_len = BSPPACKET_MTU_MAX;
+			} else {
+				tx_cut_pkt->token = TK_MIDDLE_PACKET;
+				tx_cut_pkt->cut_packet_len = BSPPACKET_MTU_MAX;
+			}
+		} else {
+			if (copy_from_user(tx_cut_pkt->cut_packet,
+					   data + already_read_len, length)) {
+				LOG(DLOG_DEBUG, "Failed to copy user data.");
+				return -EFAULT;
+			}
+			tx_cut_pkt->number = count++;
+			if (len > BSPPACKET_MTU_MAX)
+				tx_cut_pkt->token = TK_END_PACKET;
+			else
+				tx_cut_pkt->token = TK_START_END;
+
+			tx_cut_pkt->cut_packet_len = length;
+			length = 0;
+		}
+
+		already_read_len += tx_cut_pkt->cut_packet_len;
+		ret = edma_veth_send_one_pkt(tx_cut_pkt);
+		if (ret < 0) {
+			LOG(DLOG_DEBUG, "edma_veth_send_one_pkt failed, %d.",
+			    ret);
+			return ret;
+		}
+	}
+
+	LOG(DLOG_DEBUG, "send done, length: %u", length);
+
+	return 0;
+}
+
+static int edma_veth_copy_full_packet(struct edma_eth_dev_s *eth_dev,
+				      u8 *packet, u32 len)
+{
+	unsigned int count = 0;
+	u8 *ptr = NULL;
+
+	LOG(DLOG_DEBUG, "Recv full packet, len %u.", len);
+
+	ptr = kmalloc(len, GFP_ATOMIC);
+	if (ptr) {
+		/* lock the queue. */
+		spin_lock(&eth_dev->rx_queue_lock);
+
+		count = edma_veth_get_ring_buf_count(eth_dev->rx_packet_head,
+						     eth_dev->rx_packet_tail,
+						     MAX_RXTX_PACKET_LEN);
+		if (count >= (MAX_RXTX_PACKET_LEN - 1)) {
+			LOG(DLOG_DEBUG, "The rx queue is full.");
+			spin_unlock(&eth_dev->rx_queue_lock);
+			kfree(ptr);
+			return -EBUSY;
+		}
+
+		(void)memcpy(ptr, packet, len);
+		eth_dev->rx_packet[eth_dev->rx_packet_tail].packet = ptr;
+		eth_dev->rx_packet[eth_dev->rx_packet_tail].len = len;
+		eth_dev->rx_packet_tail = (eth_dev->rx_packet_tail + 1) %
+					   MAX_RXTX_PACKET_LEN;
+
+		spin_unlock(&eth_dev->rx_queue_lock);
+
+		return 0;
+	}
+
+	return -ENOMEM;
+}
+
+static int edma_veth_cut_rx_packet_recv(struct edma_eth_dev_s *eth_dev,
+					u8 *packet, u32 len)
+{
+	int ret = 0;
+	struct edma_cut_packet_node_s *node =
+	    (struct edma_cut_packet_node_s *)packet;
+	struct edma_packet_node_s *g_packet = &g_edma_recv_packet_tmp;
+	unsigned int copy_back = 0;
+
+	if (node->cut_packet_len && len > NODE_TO_PACKET_SIZE(node))
+		len = NODE_TO_PACKET_SIZE(node);
+
+	LOG(DLOG_DEBUG,
+	    "cut_packet_len: %u, token: %u/%u, number: %u, real length: %u.",
+	    node->cut_packet_len, node->token, g_last_token, node->number, len);
+
+	if (node->cut_packet_len > BSPPACKET_MTU_MAX ||
+	    ((g_packet->len + node->cut_packet_len) > MAX_PACKET_LEN)) {
+		LOG(DLOG_ERROR, "This packet is too long, packet length %u/%u",
+		    node->cut_packet_len, g_packet->len);
+		ret = -EINVAL;
+		goto fail;
+	}
+
+	if (g_last_token == TK_START_END || g_last_token == TK_END_PACKET) {
+		/* This should be a new packet. */
+		if (node->token == TK_START_PACKET ||
+		    node->token == TK_START_END) {
+			(void)memcpy(g_packet->packet, node->cut_packet,
+				     node->cut_packet_len);
+			g_packet->len = node->cut_packet_len;
+
+			if (node->token == TK_START_END) {
+				/* A full packet, increase tail. */
+				copy_back = 1;
+			} else {
+				LOG(DLOG_DEBUG,
+				    "Add middle packet with length %u",
+				    node->cut_packet_len);
+			}
+		} else {
+			LOG(DLOG_ERROR, "The rx packet is out-of-order");
+			LOG(DLOG_ERROR, "token: %d, len: %u, number: %u",
+			    node->token, node->cut_packet_len, node->number);
+			ret = -EINVAL;
+			goto fail;
+		}
+	} else {
+		/* Fragments, last token: TK_MIDDLE_PACKET/TK_START_PACKET. */
+		if (g_last_number != (node->number - 1)) {
+			LOG(DLOG_ERROR, "The number is not correct (%u/%u)",
+			    g_last_number, node->number);
+			ret = -EINVAL;
+			goto fail;
+		}
+
+		if (node->token == TK_MIDDLE_PACKET) {
+			(void)memcpy(g_packet->packet + g_packet->len,
+				     node->cut_packet, node->cut_packet_len);
+			g_packet->len = g_packet->len + node->cut_packet_len;
+			LOG(DLOG_DEBUG, "Add middle packet with length %u",
+			    node->cut_packet_len);
+		} else if (node->token == TK_END_PACKET) {
+			(void)memcpy(g_packet->packet + g_packet->len,
+				     node->cut_packet, node->cut_packet_len);
+			g_packet->len = g_packet->len + node->cut_packet_len;
+			copy_back = 1;
+		} else {
+			LOG(DLOG_ERROR, "Unexpected token: %u", node->token);
+			ret = -EINVAL;
+			goto fail;
+		}
+	}
+
+	if (copy_back) {
+		ret = edma_veth_copy_full_packet(eth_dev, g_packet->packet,
+						 g_packet->len);
+		g_packet->len = 0;
+	}
+
+	g_last_token = node->token;
+	g_last_number = node->number;
+
+	LOG(DLOG_DEBUG, "rx_packet_head:%u, rx_packet_tail: %u",
+	    eth_dev->rx_packet_head, eth_dev->rx_packet_tail);
+
+	return copy_back;
+
+fail:
+	g_last_token = TK_START_END;
+	g_last_number = 0;
+	memset(g_packet->packet, 0, MAX_PACKET_LEN);
+	g_packet->len = 0;
+
+	return ret;
+}
+
+int edma_veth_recv_pkt(struct edma_rxtx_q_s *prx_queue,
+		       struct bma_priv_data_s *priv)
+{
+	int ret = BSP_OK;
+
+	u32 i, work_limit;
+	u32 tail, head;
+
+	struct edma_bd_info_s *prx_bdinfo_v = NULL;
+	struct device *dev = NULL;
+
+	u8 *packet = NULL;
+	u32 len;
+	u32 off;
+
+	wait_queue_head_t *queue_head = NULL;
+	u8 do_wake_up = 0;
+
+	if (!priv)
+		return BSP_OK;
+
+	dev = &priv->specific.veth.pdev->dev;
+
+	work_limit = prx_queue->work_limit;
+	tail = prx_queue->tail;
+
+	for (i = 0; i < work_limit; i++) {
+		head = prx_queue->head;
+
+		if (tail == head)
+			break;
+
+		LOG(DLOG_DEBUG, "===== enter ===== [%u/%u] ======", head, tail);
+		prx_bdinfo_v = prx_queue->pbdinfobase_v + tail;
+
+		len = prx_bdinfo_v->len;
+		off = prx_bdinfo_v->off;
+		packet = prx_bdinfo_v->pdma_v;
+
+		LOG(DLOG_DEBUG, "off:%u, len: %u.", off, len);
+
+		if (!IS_CDEV_IN_OPEN_STATE()) {
+			LOG(DLOG_DEBUG,
+			    "Local char device is not opened, drop packet");
+			tail = BD_QUEUE_MASK(tail + 1);
+			continue;
+		}
+
+		dma_sync_single_for_cpu(dev, prx_bdinfo_v->dma_p,
+					len + off, DMA_FROM_DEVICE);
+
+		if (off)
+			packet += off;
+
+		ret = edma_veth_cut_rx_packet_recv(&g_eth_edmaprivate,
+						   packet, len);
+		if (ret < 0)
+			LOG(DLOG_DEBUG, "recv rx pkt fail, ret: %d", ret);
+		else if (ret != 0)
+			do_wake_up = 1;
+
+		tail = BD_QUEUE_MASK(tail + 1);
+	}
+
+	prx_queue->tail = tail;
+	head = prx_queue->head;
+
+	if (tail != head) {
+		/* check if more processing is needed. */
+		return BSP_ERR_AGAIN;
+	} else if (do_wake_up) {
+		queue_head = (wait_queue_head_t *)bma_cdev_get_wait_queue(priv);
+		/* finish reciving pkt, wake up the waiting process. */
+		if (queue_head && waitqueue_active(queue_head)) {
+			LOG(DLOG_DEBUG, "Wake up queue.");
+			wake_up(queue_head);
+		}
+	}
+
+	return BSP_OK;
+}
+
+void edma_task_do_packet_recv(unsigned long data)
+{
+	int ret = BSP_OK;
+	struct edma_rxtx_q_s *prx_queue = NULL;
+	struct bma_priv_data_s *priv = NULL;
+	struct tasklet_struct *t = (struct tasklet_struct *)data;
+
+	priv = (struct bma_priv_data_s *)g_eth_edmaprivate.edma_priv;
+	prx_queue = g_eth_edmaprivate.prx_queue;
+
+	if (prx_queue) {
+		g_eth_edmaprivate.run_skb_RX_task++;
+
+		ret = edma_veth_recv_pkt(prx_queue, priv);
+	}
+
+	if (ret == BSP_ERR_AGAIN)
+		tasklet_hi_schedule(t);
+}
+
+static inline void edma_veth_reset_dma(int type)
+{
+	bma_intf_reset_dma(GET_DMA_DIRECTION(type));
+}
+
+int __dmacmp_err_deal_2(struct edma_rxtx_q_s *prxtx_queue, u32 type)
+{
+	prxtx_queue->dmacmperr = 0;
+	prxtx_queue->start_dma = 0;
+
+	(void)edma_veth_reset_dma(type);
+
+	if (type == BSPVETH_RX) {
+		LOG(DLOG_DEBUG,
+		    "bmc->host dma time out, dma count:%d, work_limit:%d\n",
+		    prxtx_queue->dmal_cnt,
+		    prxtx_queue->work_limit);
+
+		prxtx_queue->s.dma_failed++;
+	} else {
+		LOG(DLOG_DEBUG,
+		    "host->bmc dma time out, dma count:%d, work_limit:%d\n",
+		    prxtx_queue->dmal_cnt,
+		    prxtx_queue->work_limit);
+
+		prxtx_queue->s.dma_failed++;
+	}
+
+	if (prxtx_queue->dmal_cnt > 1)
+		prxtx_queue->work_limit = (prxtx_queue->dmal_cnt >> 1);
+
+	prxtx_queue->dma_overtime++;
+	if (prxtx_queue->dma_overtime > BSPVETH_MAX_QUE_DEEP)
+		return BSPVETH_DMA_BUSY;
+
+	return BSP_OK;
+}
+
+int edma_veth_check_dma_status(struct edma_rxtx_q_s *prxtx_queue, u32 type)
+{
+	int i = 0;
+	enum dma_direction_e dir = GET_DMA_DIRECTION(type);
+
+	for (i = 0; i < BSPVETH_CHECK_DMA_STATUS_TIMES; i++) {
+		if (bma_intf_check_dma_status(dir) == BSPVETH_DMA_OK)
+			return BSP_OK;
+
+		cpu_relax();
+
+		if (i > DMA_STATUS_CHECK_DELAY_LIMIT)
+			udelay(DMA_STATUS_CHECK_DELAY_MS);
+	}
+
+	prxtx_queue->s.dma_busy++;
+	prxtx_queue->dmacmperr++;
+
+	return BSPVETH_DMA_BUSY;
+}
+
+int __check_dmacmp_H_2(struct edma_rxtx_q_s *prxtx_queue, u32 type)
+{
+	u16 start_dma;
+	u16 dmacmperr;
+	u32 cnt = 0;
+	u32 len = 0;
+	u32 host_head = 0;
+	u32 host_tail = 0;
+	u32 shm_head = 0;
+	u32 shm_tail = 0;
+	s32 ret;
+	struct edma_shmq_hd_s *pshmq_head = NULL;
+
+	if (!prxtx_queue || !prxtx_queue->pshmqhd_v)
+		return BSP_ERR_NULL_POINTER;
+
+	start_dma = prxtx_queue->start_dma;
+	if (!start_dma)
+		return BSP_OK;
+
+	pshmq_head = prxtx_queue->pshmqhd_v;
+	dmacmperr = prxtx_queue->dmacmperr;
+
+	if (dmacmperr > BSPVETH_WORK_LIMIT / DMACMP_ERR_FACTOR)
+		return __dmacmp_err_deal_2(prxtx_queue, type);
+
+	ret = edma_veth_check_dma_status(prxtx_queue, type);
+	if (ret != BSP_OK)
+		return ret;
+
+	prxtx_queue->start_dma = 0;
+	prxtx_queue->dma_overtime = 0;
+
+	if (type == BSPVETH_RX) {
+		cnt = prxtx_queue->dmal_cnt;
+		len = prxtx_queue->dmal_byte;
+
+		host_head = prxtx_queue->head;
+		shm_tail = pshmq_head->tail;
+
+		pshmq_head->tail = BD_QUEUE_MASK(shm_tail + cnt);
+		prxtx_queue->head = BD_QUEUE_MASK(host_head + cnt);
+
+		LOG(DLOG_DEBUG, "RX:host_head:%u, host_tail:%u, ",
+		    prxtx_queue->head, prxtx_queue->tail);
+
+		LOG(DLOG_DEBUG, "shm_head:%u, shm_tail:%u, inc: %u.",
+		    pshmq_head->head, pshmq_head->tail, cnt);
+
+		prxtx_queue->s.dmapkt += cnt;
+		prxtx_queue->s.dmapktbyte += len;
+	} else {
+		cnt = prxtx_queue->dmal_cnt;
+		len = prxtx_queue->dmal_byte;
+
+		host_tail = prxtx_queue->tail;
+		shm_head = pshmq_head->head;
+
+		prxtx_queue->tail = BD_QUEUE_MASK(host_tail + cnt);
+		pshmq_head->head = BD_QUEUE_MASK(shm_head + cnt);
+
+		LOG(DLOG_DEBUG, "TX:host_head:%u, host_tail:%u, ",
+		    prxtx_queue->head, prxtx_queue->tail);
+
+		LOG(DLOG_DEBUG, "shm_head:%u, shm_tail:%u, inc: %u.",
+		    pshmq_head->head, pshmq_head->tail, cnt);
+
+		prxtx_queue->s.dmapkt += cnt;
+		prxtx_queue->s.dmapktbyte += len;
+	}
+
+	tasklet_hi_schedule(&g_eth_edmaprivate.skb_task);
+
+	(void)bma_intf_int_to_bmc(g_eth_edmaprivate.edma_priv);
+
+	g_eth_edmaprivate.tobmc_int++;
+
+	return BSP_OK;
+}
+
+int __checkspace_H_2(struct edma_rxtx_q_s *prxtx_queue, u32 type, u32 *pcnt)
+{
+	u32 host_head, host_tail;
+	u32 shm_head, shm_tail;
+	u32 shm_cnt, host_cnt, cnt_tmp, cnt;
+	struct edma_shmq_hd_s *pshmq_head = NULL;
+
+	if (!prxtx_queue || !prxtx_queue->pshmqhd_v)
+		return -EFAULT;
+
+	pshmq_head = prxtx_queue->pshmqhd_v;
+
+	host_head = prxtx_queue->head;
+	host_tail = prxtx_queue->tail;
+	shm_head = pshmq_head->head;
+	shm_tail = pshmq_head->tail;
+
+	LOG(DLOG_DEBUG, "host_head:%u, host_tail:%u, shm_head:%u, shm_tail:%u.",
+	    host_head, host_tail, shm_head, shm_tail);
+
+	switch (type) {
+	case BSPVETH_RX:
+		if (shm_head == shm_tail) {
+			prxtx_queue->s.shm_empty++;
+			return BSP_ERR_NOT_TO_HANDLE;
+		}
+
+		if (!JUDGE_RING_QUEUE_SPACE(host_head, host_tail, 1))
+			return -EFAULT;
+
+		shm_cnt = GET_BD_RING_QUEUE_COUNT(shm_head, shm_tail);
+		cnt_tmp = min(shm_cnt, prxtx_queue->work_limit);
+
+		host_cnt = GET_BD_RING_QUEUE_SPACE(host_tail, host_head);
+		cnt = min(cnt_tmp, host_cnt);
+
+		LOG(DLOG_DEBUG,
+		    "RX, host_cnt: %u, shm_cnt: %u, cnt_tmp: %u, cnt: %u",
+		    host_cnt, shm_cnt, cnt_tmp, cnt);
+
+		break;
+
+	case BSPVETH_TX:
+		if (host_tail == host_head) {
+			prxtx_queue->s.q_empty++;
+			return BSP_ERR_NOT_TO_HANDLE;
+		}
+
+		host_cnt = GET_BD_RING_QUEUE_COUNT(host_head, host_tail);
+		cnt_tmp = min(host_cnt, prxtx_queue->work_limit);
+
+		shm_cnt = GET_BD_RING_QUEUE_SPACE(shm_head, shm_tail);
+		cnt = min(cnt_tmp, shm_cnt);
+
+		LOG(DLOG_DEBUG,
+		    "TX, host_cnt: %u, shm_cnt: %u, cnt_tmp: %u, cnt: %u",
+		    host_cnt, shm_cnt, cnt_tmp, cnt);
+
+		break;
+
+	default:
+		prxtx_queue->s.type_err++;
+		return -EFAULT;
+	}
+
+	if (cnt > ((BSPVETH_DMABURST_MAX * DMABURST_FACTOR) /
+				(DMABURST_FACTOR + 1)))
+		prxtx_queue->s.dma_burst++;
+
+	*pcnt = cnt;
+
+	return BSP_OK;
+}
+
+int __make_dmalistbd_h2b_H_2(struct edma_rxtx_q_s *prxtx_queue, u32 cnt)
+{
+	u32 i = 0;
+	u32 len = 0;
+	u32 off = 0;
+	struct edma_dmal_s *pdmalbase_v = NULL;
+	struct edma_shmq_hd_s *pshmq_head = NULL;
+	struct edma_bd_info_s *pbdinfobase_v = NULL;
+	struct edma_dma_shmbd_s *pshmbdbase_v = NULL;
+
+	unsigned long addr;
+
+	u32 host_tail;
+	u32 shm_head;
+
+	if (!prxtx_queue)
+		return -EFAULT;
+
+	if (cnt == 0)
+		return 0;
+
+	pdmalbase_v = prxtx_queue->pdmalbase_v;
+	pbdinfobase_v = prxtx_queue->pbdinfobase_v;
+	pshmbdbase_v = prxtx_queue->pshmbdbase_v;
+
+	pshmq_head = prxtx_queue->pshmqhd_v;
+
+	host_tail = prxtx_queue->tail;
+	shm_head = pshmq_head->head;
+
+	for (i = 0; i < cnt; i++) {
+		LOG(DLOG_DEBUG, "TX DMA, HOST: %u -> BMC: %u",
+		    host_tail, shm_head);
+
+		pdmalbase_v[i].chl = 0x1;
+
+		addr = EDMA_ADDR_ALIGNED(pbdinfobase_v[host_tail].dma_p);
+		off = EDMA_ADDR_OFFSET(addr);
+
+		/* src: veth_send_one_pkt. */
+		pdmalbase_v[i].slow = lower_32_bits(addr);
+		pdmalbase_v[i].shi = upper_32_bits(addr);
+
+		/* dst: bmc dma, in shared memory. */
+		pdmalbase_v[i].dlow =
+		    lower_32_bits(pshmbdbase_v[shm_head].dma_p);
+		pdmalbase_v[i].dhi = 0;
+
+		/* len: len + offset caused by alignment */
+		pdmalbase_v[i].len = pbdinfobase_v[host_tail].len + off;
+
+		LOG(DLOG_DEBUG,
+		    "TX DMA %08x%08x -> %08x%08x, off: %u, len: %u.",
+		    pdmalbase_v[i].shi, pdmalbase_v[i].slow,
+		    pdmalbase_v[i].dhi, pdmalbase_v[i].dlow,
+		    off, pbdinfobase_v[host_tail].len);
+
+		pshmbdbase_v[shm_head].len = pbdinfobase_v[host_tail].len;
+		pshmbdbase_v[shm_head].off = off;
+
+		len += pdmalbase_v[i].len;
+
+		/* ready for the next round. */
+		host_tail = BD_QUEUE_MASK(host_tail + 1);
+		shm_head = BD_QUEUE_MASK(shm_head + 1);
+	}
+
+	pdmalbase_v[i - 1].chl = 0x9;
+
+	pdmalbase_v[i].chl = 0x7;
+	pdmalbase_v[i].len = 0x0;
+	pdmalbase_v[i].slow = lower_32_bits((u64)prxtx_queue->pdmalbase_p);
+	pdmalbase_v[i].shi = upper_32_bits((u64)prxtx_queue->pdmalbase_p);
+	pdmalbase_v[i].dlow = 0;
+	pdmalbase_v[i].dhi = 0;
+
+	prxtx_queue->dmal_cnt = cnt;
+	prxtx_queue->dmal_byte = len;
+
+	return 0;
+}
+
+int __make_dmalistbd_b2h_H_2(struct edma_rxtx_q_s *prxtx_queue, u32 cnt)
+{
+	u32 i;
+	u32 len = 0;
+
+	struct edma_dmal_s *pdmalbase_v = NULL;
+	struct edma_shmq_hd_s *pshmq_head = NULL;
+	struct edma_bd_info_s *pbdinfobase_v = NULL;
+	struct edma_dma_shmbd_s *pshmbdbase_v = NULL;
+
+	u32 host_head;
+	u32 shm_tail;
+
+	if (!prxtx_queue)
+		return -EFAULT;
+
+	if (cnt == 0)
+		return -EFAULT;
+
+	pdmalbase_v = prxtx_queue->pdmalbase_v;
+	pbdinfobase_v = prxtx_queue->pbdinfobase_v;
+	pshmbdbase_v = prxtx_queue->pshmbdbase_v;
+
+	pshmq_head = prxtx_queue->pshmqhd_v;
+
+	host_head = prxtx_queue->head;
+	shm_tail = pshmq_head->tail;
+
+	for (i = 0; i < cnt; i++) {
+		LOG(DLOG_DEBUG, "RX DMA, BMC: %u -> HOST: %u",
+		    shm_tail, host_head);
+
+		pbdinfobase_v[host_head].off = pshmbdbase_v[shm_tail].off;
+		pbdinfobase_v[host_head].len = pshmbdbase_v[shm_tail].len;
+
+		pdmalbase_v[i].chl = 0x1;
+
+		/* src: bmc set in shared memory. */
+		pdmalbase_v[i].slow =
+		    lower_32_bits(pshmbdbase_v[shm_tail].dma_p);
+		pdmalbase_v[i].shi = 0;
+
+		/* dst: edma_veth_setup_all_rxtx_queue. */
+		pdmalbase_v[i].dlow =
+		    lower_32_bits(pbdinfobase_v[host_head].dma_p);
+		pdmalbase_v[i].dhi =
+		    upper_32_bits(pbdinfobase_v[host_head].dma_p);
+
+		pdmalbase_v[i].len = pshmbdbase_v[shm_tail].len +
+				     pshmbdbase_v[shm_tail].off;
+
+		LOG(DLOG_DEBUG,
+		    "RX DMA %08x%08x -> %08x%08x, off: %u, len: %u, total: %u.",
+		    pdmalbase_v[i].shi, pdmalbase_v[i].slow,
+		    pdmalbase_v[i].dhi, pdmalbase_v[i].dlow,
+		    pshmbdbase_v[shm_tail].off, pshmbdbase_v[shm_tail].len,
+		    pdmalbase_v[i].len);
+
+		len += pdmalbase_v[i].len;
+
+		/* ready for the next round. */
+		host_head = BD_QUEUE_MASK(host_head + 1);
+		shm_tail = BD_QUEUE_MASK(shm_tail + 1);
+	}
+
+	pdmalbase_v[i - 1].chl = 0x9;
+
+	pdmalbase_v[i].chl = 0x7;
+	pdmalbase_v[i].len = 0x0;
+	pdmalbase_v[i].slow = lower_32_bits((u64)prxtx_queue->pdmalbase_p);
+	pdmalbase_v[i].shi = upper_32_bits((u64)prxtx_queue->pdmalbase_p);
+	pdmalbase_v[i].dlow = 0;
+	pdmalbase_v[i].dhi = 0;
+
+	prxtx_queue->dmal_cnt = cnt;
+	prxtx_queue->dmal_byte = len;
+
+	return 0;
+}
+
+int __start_dmalist_H_2(struct edma_rxtx_q_s *prxtx_queue, u32 type, u32 cnt)
+{
+	int ret = BSP_OK;
+	struct bma_dma_transfer_s dma_transfer = { 0 };
+	struct edma_shmq_hd_s *pshmq_head = NULL;
+
+	if (!prxtx_queue)
+		return -1;
+
+	pshmq_head = prxtx_queue->pshmqhd_v;
+
+	LOG(DLOG_DEBUG, "before -> %u/%u/%u/%u.",
+	    prxtx_queue->head, prxtx_queue->tail,
+	    pshmq_head->head, pshmq_head->tail);
+
+	if (type == BSPVETH_RX) {
+		dma_transfer.dir = BMC_TO_HOST;
+		ret = __make_dmalistbd_b2h_H_2(prxtx_queue, cnt);
+	} else {
+		dma_transfer.dir = HOST_TO_BMC;
+		ret = __make_dmalistbd_h2b_H_2(prxtx_queue, cnt);
+	}
+
+	if (ret < 0)
+		return ret;
+
+	dma_transfer.type = DMA_LIST;
+	dma_transfer.transfer.list.dma_addr =
+	    (dma_addr_t)prxtx_queue->pdmalbase_p;
+
+	ret = bma_intf_start_dma(g_eth_edmaprivate.edma_priv, &dma_transfer);
+	LOG(DLOG_DEBUG, "after -> %u/%u/%u/%u, ret: %d",
+	    prxtx_queue->head, prxtx_queue->tail,
+	    pshmq_head->head, pshmq_head->tail,
+	    ret);
+
+	if (ret < 0)
+		return ret;
+
+	prxtx_queue->start_dma = 1;
+
+	return BSP_OK;
+}
+
+int check_dma_queue_fault_2(struct edma_rxtx_q_s *prxtx_queue,
+			    u32 type, u32 *pcnt)
+{
+	int ret;
+	u32 cnt = 0;
+
+	if (prxtx_queue->dma_overtime > BSPVETH_MAX_QUE_DEEP)
+		return -EFAULT;
+
+	ret = __check_dmacmp_H_2(prxtx_queue, type);
+	if (ret != BSP_OK)
+		return -EFAULT;
+
+	ret = __checkspace_H_2(prxtx_queue, type, &cnt);
+	if (ret != BSP_OK)
+		return -EFAULT;
+
+	if (CHECK_DMA_RXQ_FAULT(prxtx_queue, type, cnt)) {
+		udelay(DMA_RXQ_FAULT_DELAY);
+
+		prxtx_queue->dmal_cnt--;
+
+		return -EFAULT;
+	}
+
+	*pcnt = cnt;
+
+	return BSP_OK;
+}
+
+int __dma_rxtx_H_2(struct edma_rxtx_q_s *prxtx_queue, u32 type)
+{
+	int ret;
+	u32 cnt = 0;
+
+	if (!prxtx_queue || !prxtx_queue->pshmqhd_v)
+		return -EFAULT;
+
+	if (CHECK_DMA_QUEUE_EMPTY(type, prxtx_queue)) {
+		LOG(DLOG_DEBUG, "Queue (type: %u) is empty.", type);
+		return BSP_OK;
+	}
+
+	ret = check_dma_queue_fault_2(prxtx_queue, type, &cnt);
+	if (ret != BSP_OK) {
+		LOG(DLOG_DEBUG, "check_dma_queue_fault_2 (ret: %d).", ret);
+		return -EFAULT;
+	}
+
+	if (cnt == 0)
+		return BSP_OK;
+
+	ret = __start_dmalist_H_2(prxtx_queue, type, cnt);
+	if (ret != BSP_OK) {
+		LOG(DLOG_DEBUG, "__start_dmalist_H_2 returns %d", ret);
+		return -EFAULT;
+	}
+
+	if (cnt <= DMA_QUEUE_FAULT_LIMIT) {
+		ret = __check_dmacmp_H_2(prxtx_queue, type);
+		if (ret != BSP_OK) {
+			LOG(DLOG_DEBUG, "__check_dmacmp_H_2 returns %d", ret);
+			return -EFAULT;
+		}
+	}
+
+	return BSP_OK;
+}
+
+inline int veth_dma_task_H_2(u32 type)
+{
+	struct edma_rxtx_q_s *prxtx_queue = NULL;
+
+	if (type == BSPVETH_RX) {
+		g_eth_edmaprivate.run_dma_RX_task++;
+		prxtx_queue = g_eth_edmaprivate.prx_queue;
+	} else {
+		g_eth_edmaprivate.run_dma_TX_task++;
+		prxtx_queue = g_eth_edmaprivate.ptx_queue;
+	}
+
+	if (prxtx_queue) {
+		if (!edma_is_queue_ready(prxtx_queue)) {
+			LOG(DLOG_DEBUG, "queue is not ready, init flag: %u.",
+			    prxtx_queue->pshmqhd_v->init);
+			return BSP_OK;
+		}
+
+		(void)__dma_rxtx_H_2(prxtx_queue, type);
+
+		if (!CHECK_DMA_QUEUE_EMPTY(type, prxtx_queue))
+			return BSP_ERR_AGAIN;
+	}
+
+	return BSP_OK;
+}
+
+void edma_task_do_data_transmit(unsigned long data)
+{
+	struct tasklet_struct *t = (struct tasklet_struct *)data;
+	int txret, rxret;
+
+	LOG(DLOG_DEBUG, "host_head/host_tail/shm_head/shm_tail - ");
+	LOG(DLOG_DEBUG, "rx:%u/%u/%u/%u, tx:%u/%u/%u/%u.",
+	    g_eth_edmaprivate.prx_queue->head,
+	    g_eth_edmaprivate.prx_queue->tail,
+	    g_eth_edmaprivate.prx_queue->pshmqhd_v->head,
+	    g_eth_edmaprivate.prx_queue->pshmqhd_v->tail,
+	    g_eth_edmaprivate.ptx_queue->head,
+	    g_eth_edmaprivate.ptx_queue->tail,
+	    g_eth_edmaprivate.ptx_queue->pshmqhd_v->head,
+	    g_eth_edmaprivate.ptx_queue->pshmqhd_v->tail);
+
+	txret = veth_dma_task_H_2(BSPVETH_TX);
+
+	rxret = veth_dma_task_H_2(BSPVETH_RX);
+
+	LOG(DLOG_DEBUG, "host_head/host_tail/shm_head/shm_tail - ");
+	LOG(DLOG_DEBUG, "rx:%u/%u/%u/%u, tx:%u/%u/%u/%u.\n",
+	    g_eth_edmaprivate.prx_queue->head,
+	    g_eth_edmaprivate.prx_queue->tail,
+	    g_eth_edmaprivate.prx_queue->pshmqhd_v->head,
+	    g_eth_edmaprivate.prx_queue->pshmqhd_v->tail,
+	    g_eth_edmaprivate.ptx_queue->head,
+	    g_eth_edmaprivate.ptx_queue->tail,
+	    g_eth_edmaprivate.ptx_queue->pshmqhd_v->head,
+	    g_eth_edmaprivate.ptx_queue->pshmqhd_v->tail);
+
+	if (txret == BSP_ERR_AGAIN || rxret == BSP_ERR_AGAIN) {
+		/* restart transmission. */
+		tasklet_hi_schedule(t);
+	}
+}
+
+int edma_tasklet_setup(struct edma_eth_dev_s *dev, u8 **rx_buf,
+		       struct edma_cut_packet_node_s **tx_cut_pkt_buf)
+{
+	u8 *rx_pkt_buf;
+	struct edma_packet_node_s *rx_packet = NULL;
+	struct edma_cut_packet_node_s *tx_cut_buf = NULL;
+	size_t rx_size =
+	    sizeof(struct edma_packet_node_s) * MAX_RXTX_PACKET_LEN;
+
+	rx_pkt_buf = kmalloc(MAX_PACKET_LEN, GFP_KERNEL);
+	if (!rx_pkt_buf)
+		return -ENOMEM;
+
+	tx_cut_buf = (struct edma_cut_packet_node_s *)
+	    kmalloc(sizeof(*tx_cut_buf), GFP_KERNEL);
+	if (!tx_cut_buf) {
+		kfree(rx_pkt_buf);
+		return -ENOMEM;
+	}
+
+	rx_packet = kmalloc(rx_size, GFP_KERNEL);
+	if (!rx_packet) {
+		kfree(rx_pkt_buf);
+		kfree(tx_cut_buf);
+		return -ENOMEM;
+	}
+
+	memset(rx_pkt_buf, 0, MAX_PACKET_LEN);
+	memset(tx_cut_buf, 0, sizeof(*tx_cut_buf));
+	memset(rx_packet, 0, rx_size);
+
+	*rx_buf = rx_pkt_buf;
+	*tx_cut_pkt_buf = tx_cut_buf;
+	dev->rx_packet = rx_packet;
+
+	spin_lock_init(&dev->rx_queue_lock);
+
+	tasklet_init(&dev->skb_task,
+		     edma_task_do_packet_recv,
+		     (unsigned long)&dev->skb_task);
+
+	tasklet_init(&dev->dma_task,
+		     edma_task_do_data_transmit,
+		     (unsigned long)&dev->dma_task);
+
+	return 0;
+}
+
+void edma_tasklet_free(struct edma_eth_dev_s *dev, u8 **rx_buf,
+		       struct edma_cut_packet_node_s **tx_cut_pkt_buf)
+{
+	if (!*rx_buf)
+		return;
+
+	/* stop task before releasing resource. */
+	tasklet_kill(&dev->dma_task);
+	tasklet_kill(&dev->skb_task);
+
+	kfree(*rx_buf);
+	kfree(*tx_cut_pkt_buf);
+
+	/* flush the ring buf. */
+	edma_veth_flush_ring_node(dev->rx_packet, MAX_RXTX_PACKET_LEN);
+	kfree(dev->rx_packet);
+
+	*rx_buf = NULL;
+	*tx_cut_pkt_buf = NULL;
+	dev->rx_packet = NULL;
+}
+
+static int edma_veth_int_handler(struct notifier_block *nb,
+				 unsigned long ev, void *unuse)
+{
+	g_eth_edmaprivate.recv_int++;
+
+	if (g_eth_edmaprivate.dma_task.func)
+		tasklet_hi_schedule(&g_eth_edmaprivate.dma_task);
+
+	return IRQ_HANDLED;
+}
+
+static struct notifier_block g_edma_veth_int_nb = {
+	.notifier_call = edma_veth_int_handler,
+};
+
+static int comm_init_dev(struct edma_eth_dev_s *edma,
+			 const struct file_operations *fops)
+{
+	struct cdev_dev_s *dev = &edma->cdev;
+	int ret;
+
+	dev->priv = edma->edma_priv;
+	dev->dev.minor = MISC_DYNAMIC_MINOR;
+	dev->dev.name = CDEV_VETH_NAME;
+	dev->dev.fops = fops;
+
+	ret = misc_register(&dev->dev);
+	if (ret < 0) {
+		LOG(DLOG_ERROR, "Failed to alloc major number, %d", ret);
+		return ret;
+	}
+
+	return 0;
+}
+
+static inline void comm_cleanup_dev(struct edma_eth_dev_s *edma)
+{
+	struct cdev_dev_s *dev = &edma->cdev;
+
+	misc_deregister(&dev->dev);
+}
+
+static int __init edma_cdev_init(void)
+{
+	int ret;
+
+	g_write_count = 0;
+	g_delay_ms = 0;
+	g_last_number = 0;
+	g_peer_not_ready = 0;
+
+	LOG(DLOG_DEBUG, "Module init.");
+
+	if (!bma_intf_check_edma_supported())
+		return -ENXIO;
+
+	(void)memset(&g_eth_edmaprivate, 0, sizeof(g_eth_edmaprivate));
+
+	/* register EDMA sub-subyem. */
+	ret = bma_intf_register_type(TYPE_VETH, 0, INTR_ENABLE,
+				     &g_eth_edmaprivate.edma_priv);
+	if (ret < 0) {
+		LOG(DLOG_ERROR, "Failed to register EDMA interface.");
+		goto failed;
+	}
+
+	/* initialize host DMA address. */
+	edma_veth_host_addr_init(g_eth_edmaprivate.edma_priv);
+
+	/* setup TX/RX resource */
+	ret = edma_veth_setup_resource(&g_eth_edmaprivate);
+	if (ret < 0) {
+		LOG(DLOG_ERROR, "Failed to setup resource.");
+		goto failed1;
+	}
+
+	/* setup resource for user packets. */
+	ret = edma_tasklet_setup(&g_eth_edmaprivate,
+				 &g_edma_recv_packet_tmp.packet,
+				 &g_edma_send_cut_packet);
+	if (ret < 0)
+		goto failed2;
+
+	/* register char device. */
+	ret = comm_init_dev(&g_eth_edmaprivate, &g_eth_edma_cdev_fops);
+	if (ret != 0) {
+		LOG(DLOG_ERROR, "Failed to register cdev device.");
+		goto failed3;
+	}
+
+	/* register EDMA INT notifier. */
+	ret = bma_intf_register_int_notifier(&g_edma_veth_int_nb);
+	if (ret < 0) {
+		LOG(DLOG_ERROR, "Failed to register INT notifier.");
+		goto failed4;
+	}
+
+	dump_global_info();
+
+	GET_SYS_SECONDS(g_eth_edmaprivate.init_time);
+
+	return 0;
+
+failed4:
+	comm_cleanup_dev(&g_eth_edmaprivate);
+failed3:
+	edma_tasklet_free(&g_eth_edmaprivate,
+			  &g_edma_recv_packet_tmp.packet,
+			  &g_edma_send_cut_packet);
+failed2:
+	edma_veth_free_resource(&g_eth_edmaprivate);
+failed1:
+	(void)bma_intf_unregister_type(&g_eth_edmaprivate.edma_priv);
+failed:
+	return ret;
+}
+
+static void __exit edma_cdev_exit(void)
+{
+	LOG(DLOG_DEBUG, "Module exit.");
+
+	bma_intf_unregister_int_notifier(&g_edma_veth_int_nb);
+
+	comm_cleanup_dev(&g_eth_edmaprivate);
+
+	edma_tasklet_free(&g_eth_edmaprivate,
+			  &g_edma_recv_packet_tmp.packet,
+			  &g_edma_send_cut_packet);
+
+	edma_veth_free_resource(&g_eth_edmaprivate);
+
+	bma_intf_unregister_type(&g_eth_edmaprivate.edma_priv);
+}
+
+static inline int cdev_check_ring_recv(void)
+{
+	unsigned int count;
+
+	count = edma_veth_get_ring_buf_count(g_eth_edmaprivate.rx_packet_head,
+					     g_eth_edmaprivate.rx_packet_tail,
+					     MAX_RXTX_PACKET_LEN);
+	return (count > 0 ? 1 : 0);
+}
+
+static ssize_t cdev_copy_packet_to_user(struct edma_eth_dev_s *dev,
+					char __user *data, size_t count)
+{
+	unsigned char *packet = NULL;
+	unsigned char *start = NULL;
+	unsigned int free_packet = 0;
+	ssize_t length = (ssize_t)count;
+	ssize_t left;
+
+	LOG(DLOG_DEBUG, "rx_packet_head:%u, rx_packet_tail: %u",
+	    dev->rx_packet_head, dev->rx_packet_tail);
+
+	spin_lock(&dev->rx_queue_lock);
+
+	if (!cdev_check_ring_recv()) {
+		spin_unlock(&dev->rx_queue_lock);
+		return -EAGAIN;
+	}
+
+	left = (ssize_t)(dev->rx_packet[dev->rx_packet_head].len) - g_read_pos;
+	start = dev->rx_packet[dev->rx_packet_head].packet + g_read_pos;
+
+	LOG(DLOG_DEBUG,
+	    "User needs %ld bytes, pos: %ld, total len: %u, left: %ld.",
+	    count, g_read_pos, dev->rx_packet[dev->rx_packet_head].len, left);
+	if (left <= 0) {
+		/* No more data in this message, retry. */
+		length = -EAGAIN;
+		free_packet = 1;
+	} else if (length > left) {
+		/* A full message is returned. */
+		length = left;
+		free_packet = 1;
+	} else {
+		/* Update pos. */
+		g_read_pos += length;
+	}
+
+	if (free_packet) {
+		g_read_pos = 0;
+		packet = dev->rx_packet[dev->rx_packet_head].packet;
+		dev->rx_packet[dev->rx_packet_head].packet = NULL;
+		dev->rx_packet_head = (dev->rx_packet_head + 1) %
+				      MAX_RXTX_PACKET_LEN;
+	}
+
+	spin_unlock(&dev->rx_queue_lock);
+
+	if (length > 0 && copy_to_user(data, start, length)) {
+		LOG(DLOG_DEBUG, "Failed to copy to user, skip this message.");
+		length = -EFAULT;
+		g_read_pos = 0;
+	}
+
+	LOG(DLOG_DEBUG,
+	    "Copied bytes: %ld, pos: %ld, buf len: %lu, free_packet: %d.",
+	    length, g_read_pos, count, free_packet);
+
+	if (packet) {
+		/* Free the packet as needed. */
+		kfree(packet);
+	}
+
+	return length;
+}
+
+int cdev_open(struct inode *inode_ptr, struct file *filp)
+{
+	struct cdev_dev_s *dev = &g_eth_edmaprivate.cdev;
+
+	LOG(DLOG_DEBUG, "Open device.");
+
+	if (!inode_ptr || !filp)
+		return -EFAULT;
+
+	/* only one instance is allowed. */
+	if (IS_CDEV_IN_OPEN_STATE())
+		return -EBUSY;
+
+	LOG(DLOG_DEBUG, "Init flag, rx: %d, tx:%d",
+	    g_eth_edmaprivate.prx_queue->pshmqhd_v->init,
+	    g_eth_edmaprivate.ptx_queue->pshmqhd_v->init);
+
+	/* save to private data. */
+	filp->private_data = dev;
+	SET_CDEV_OPEN_STATE(CDEV_OPENED);
+	g_read_pos = 0;
+
+	return 0;
+}
+
+int cdev_release(struct inode *inode_ptr, struct file *filp)
+{
+	LOG(DLOG_DEBUG, "Close device.");
+
+	if (!filp)
+		return 0;
+
+	filp->private_data = NULL;
+
+	SET_CDEV_OPEN_STATE(CDEV_CLOSED);
+
+	return 0;
+}
+
+unsigned int cdev_poll(struct file *filp, poll_table *wait)
+{
+	unsigned int mask = 0;
+	wait_queue_head_t *queue_head = NULL;
+
+	if (!filp)
+		return 0;
+
+	edma_veth_dump();
+
+	queue_head = (wait_queue_head_t *)
+	    bma_cdev_get_wait_queue(GET_PRIVATE_DATA(filp));
+	if (!queue_head)
+		return 0;
+
+	/* check or add to wait queue. */
+	poll_wait(filp, queue_head, wait);
+
+	if (!edma_is_queue_ready(g_eth_edmaprivate.prx_queue))
+		return 0;
+
+	if (cdev_check_ring_recv() > 0)
+		mask = (POLLIN | POLLRDNORM);
+
+	return mask;
+}
+
+ssize_t cdev_read(struct file *filp, char __user *data,
+		  size_t count, loff_t *ppos)
+{
+	struct edma_eth_dev_s *dev = &g_eth_edmaprivate;
+	ssize_t length = 0;
+
+	if (!data || count >= MAX_PACKET_LEN)
+		return -EFAULT;
+
+	LOG(DLOG_DEBUG, "read begin, count: %ld, pos: %u.", count, g_read_pos);
+
+	length = cdev_copy_packet_to_user(dev, data, count);
+
+	LOG(DLOG_DEBUG, "read done, length: %ld, pos: %u.", length, g_read_pos);
+
+	return length;
+}
+
+ssize_t cdev_write(struct file *filp, const char __user *data,
+		   size_t count, loff_t *ppos)
+{
+	int ret = 0;
+	struct edma_eth_dev_s *pdev = &g_eth_edmaprivate;
+
+	if (!data || count <= 0 || count > MAX_PACKET_LEN)
+		return -EINVAL;
+
+	if (!edma_is_queue_ready(pdev->ptx_queue)) {
+		if (g_peer_not_ready == 0 && pdev->ptx_queue) {
+			LOG(DLOG_ERROR, "Peer rx queue is not ready (%u).",
+			    pdev->ptx_queue->pshmqhd_v->init);
+			g_peer_not_ready = 1;
+		}
+		return -EPERM;
+	} else if (g_peer_not_ready) {
+		LOG(DLOG_ERROR, "Peer rx queue becomes ready.");
+		g_peer_not_ready = 0;
+	}
+
+	LOG(DLOG_DEBUG, "data length is %lu, pos: %u (%u/%u)",
+	    count, g_read_pos,
+	    pdev->ptx_queue->pshmqhd_v->count,
+	    pdev->ptx_queue->pshmqhd_v->total);
+
+	ret = edma_veth_cut_tx_packet_send(pdev, data, count);
+	if (ret < 0) {
+		LOG(DLOG_ERROR, "Failed to send packet, return code: %d.", ret);
+	} else {
+		tasklet_hi_schedule(&g_eth_edmaprivate.dma_task);
+		ret = count;
+	}
+
+	return ret;
+}
+
+MODULE_VERSION(MICRO_TO_STR(CDEV_VETH_VERSION));
+MODULE_AUTHOR("HUAWEI TECHNOLOGIES CO., LTD.");
+MODULE_DESCRIPTION("HUAWEI CDEV DRIVER");
+MODULE_LICENSE("GPL");
+
+module_init(edma_cdev_init);
+module_exit(edma_cdev_exit);
+
diff --git a/drivers/net/ethernet/huawei/bma/cdev_veth_drv/virtual_cdev_eth_net.h b/drivers/net/ethernet/huawei/bma/cdev_veth_drv/virtual_cdev_eth_net.h
new file mode 100644
index 000000000000..23ac6ce489cc
--- /dev/null
+++ b/drivers/net/ethernet/huawei/bma/cdev_veth_drv/virtual_cdev_eth_net.h
@@ -0,0 +1,300 @@
+/* SPDX-License-Identifier: GPL-2.0 */
+/* Huawei iBMA driver.
+ * Copyright (c) 2019, Huawei Technologies Co., Ltd.
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public License
+ * as published by the Free Software Foundation; either version 2
+ * of the License, or (at your option) any later version.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ * GNU General Public License for more details.
+ */
+
+#ifndef _VETH_CDEV_NET_H_
+#define _VETH_CDEV_NET_H_
+
+#include <linux/kernel.h>
+#include <linux/module.h>
+#include <linux/types.h>
+#include <linux/slab.h>
+#include <linux/vmalloc.h>
+#include <linux/spinlock.h>
+#include <linux/dma-mapping.h>
+#include <linux/miscdevice.h>
+#include <linux/netdevice.h>
+#include <linux/poll.h>
+#include <linux/delay.h>
+
+#include "../edma_drv/bma_include.h"
+#include "../include/bma_ker_intf.h"
+
+#define BSP_OK                         (0)
+#define BSP_ERR                        (0xFFFFFFFF)
+#define BSP_NETDEV_TX_BUSY             (1)
+#define BSP_ERR_INIT_ERR               (BSP_NETDEV_TX_BUSY)
+#define BSP_ETH_ERR_BASE               (0x0FFFF000)
+#define BSP_ERR_OUT_OF_MEM             (BSP_ETH_ERR_BASE + 1)
+#define BSP_ERR_NULL_POINTER           (BSP_ETH_ERR_BASE + 2)
+#define BSP_ERR_INVALID_STR            (BSP_ETH_ERR_BASE + 3)
+#define BSP_ERR_INVALID_PARAM          (BSP_ETH_ERR_BASE + 4)
+#define BSP_ERR_INVALID_DATA           (BSP_ETH_ERR_BASE + 5)
+#define BSP_ERR_OUT_OF_RANGE           (BSP_ETH_ERR_BASE + 6)
+#define BSP_ERR_INVALID_CARD           (BSP_ETH_ERR_BASE + 7)
+#define BSP_ERR_INVALID_GRP            (BSP_ETH_ERR_BASE + 8)
+#define BSP_ERR_INVALID_ETH            (BSP_ETH_ERR_BASE + 9)
+#define BSP_ERR_SEND_ERR               (BSP_ETH_ERR_BASE + 10)
+#define BSP_ERR_DMA_ERR                (BSP_ETH_ERR_BASE + 11)
+#define BSP_ERR_RECV_ERR               (BSP_ETH_ERR_BASE + 12)
+#define BSP_ERR_SKB_ERR                (BSP_ETH_ERR_BASE + 13)
+#define BSP_ERR_DMA_ADDR_ERR           (BSP_ETH_ERR_BASE + 14)
+#define BSP_ERR_IOREMAP_ERR            (BSP_ETH_ERR_BASE + 15)
+#define BSP_ERR_LEN_ERR                (BSP_ETH_ERR_BASE + 16)
+#define BSP_ERR_STAT_ERR               (BSP_ETH_ERR_BASE + 17)
+#define BSP_ERR_AGAIN                  (BSP_ETH_ERR_BASE + 18)
+#define BSP_ERR_NOT_TO_HANDLE          (BSP_ETH_ERR_BASE + 19)
+
+#define VETH_SHAREPOOL_BASE_INBMC  (0x84820000)
+#define VETH_SHAREPOOL_SIZE        (0xdf000)
+#define VETH_SHAREPOOL_OFFSET      (0x10000)
+#define MAX_SHAREQUEUE_SIZE        (0x20000)
+
+#define BSPVETH_DMABURST_MAX       (64)
+#define BSPVETH_SHMBDBASE_OFFSET   (0x80)
+#define SHMDMAL_OFFSET             (0x10000)
+#define MAX_SHMDMAL_SIZE           (BSPVETH_DMABURST_MAX * 32)
+#define MAX_QUEUE_NUM              (1)
+#define MAX_QUEUE_BDNUM            (128)
+#define BSPVETH_MAX_QUE_DEEP       (MAX_QUEUE_BDNUM)
+#define BSPVETH_POINT_MASK         (MAX_QUEUE_BDNUM - 1)
+#define BSPVETH_WORK_LIMIT         (64)
+#define BSPVETH_CHECK_DMA_STATUS_TIMES (512)
+
+#define BSPPACKET_MTU_MAX           (1500)
+
+#define BSPVETH_DMA_OK              (1)
+#define BSPVETH_DMA_BUSY            (0)
+#define BSPVETH_RX                  (2)
+#define BSPVETH_TX                  (3)
+#define BSPVETH_SHMQUEUE_INITOK     (0x12)
+#define BSPVETH_SHMQUEUE_INITOK_V2  (0x16)
+
+#define MAX_PACKET_LEN              (128 * BSPPACKET_MTU_MAX)
+#define MAX_RXTX_PACKET_LEN         64
+#define RESERVE_SPACE               24
+
+/* device name. */
+#define CDEV_VETH_NAME              "net_cdev"
+#define CDEV_OPENED					(1)
+#define CDEV_CLOSED					(0)
+
+#ifndef GET_SYS_SECONDS
+#define GET_SYS_SECONDS(t) do { \
+	struct timespec _uptime; \
+	get_monotonic_boottime(&_uptime); \
+	t = _uptime.tv_sec; \
+} while (0)
+#endif
+
+struct edma_packet_node_s {
+	u32 len;
+	u8 *packet;
+};
+
+struct edma_cut_packet_node_s {
+	u32 token;
+	u32 number;
+	u32 cut_packet_len;
+	u8 cut_packet[BSPPACKET_MTU_MAX];
+	u8 resv[RESERVE_SPACE];
+};
+
+#define TK_MIDDLE_PACKET 0
+#define TK_START_PACKET 1
+#define TK_END_PACKET 2
+#define TK_START_END 3
+
+/* EDMA transfer requires an alignment of 4. */
+#define EDMA_ADDR_ALIGNMENT         (4UL)
+#define EDMA_ADDR_ALIGN_MASK        (EDMA_ADDR_ALIGNMENT - 1)
+#define EDMA_ADDR_ALIGNED(dma_p)    (((unsigned long)(dma_p)) & \
+				    (~(EDMA_ADDR_ALIGN_MASK)))
+#define EDMA_ADDR_OFFSET(dma_p)     (((unsigned long)(dma_p)) & \
+				    (EDMA_ADDR_ALIGN_MASK))
+
+#define NODE_SIZE                   (sizeof(struct edma_cut_packet_node_s))
+#define NODE_TO_PACKET_SIZE(n)      (n->cut_packet_len + (3 * sizeof(u32)))
+#define NODE_PER_PAGE               (PAGE_SIZE / (NODE_SIZE))
+
+#define ALIGN_MASK 4096
+#define STRESS_FACTOR 100
+#define DMA_STATUS_CHECK_DELAY_LIMIT 20
+#define DMA_STATUS_CHECK_DELAY_MS 5
+#define DMA_RXQ_FAULT_DELAY 50
+#define DMA_QUEUE_FAULT_LIMIT 16
+#define DMACMP_ERR_FACTOR 4
+#define DMABURST_FACTOR 7
+
+struct cdev_dev_s {
+	struct miscdevice dev;
+	void *priv;
+};
+
+struct edma_rxtx_statistics {
+	u64 dmapkt;
+	u64 dmapktbyte;
+
+	u32 q_empty;
+	u32 shm_empty;
+	u32 dma_busy;
+	u32 type_err;
+
+	u32 dma_need_offset;
+	u32 dma_failed;
+	u32 dma_burst;
+};
+
+struct edma_bd_info_s {
+	u8 *pdma_v;
+	dma_addr_t dma_p;
+	u32 len;
+	u32 off;
+};
+
+struct edma_dma_shmbd_s {
+	u32 dma_p;
+	u32 len;
+	u32 off;
+};
+
+struct edma_shmq_hd_s {
+	u32 count;
+	u32 total;
+	u32 next_to_fill;
+	u32 next_to_free;
+	u32 resv1;
+	u32 resv2;
+	u32 init;
+	u32 head;
+	u32 tail;
+};
+
+struct edma_dmal_s {
+	u32 chl;
+	u32 len;
+	u32 slow;
+	u32 shi;
+	u32 dlow;
+	u32 dhi;
+};
+
+struct edma_rxtx_q_s {
+	struct edma_bd_info_s *pbdinfobase_v;
+
+	struct edma_shmq_hd_s *pshmqhd_v;
+	u8 *pshmqhd_p;
+
+	struct edma_dma_shmbd_s *pshmbdbase_v;
+	u8 *pshmbdbase_p;
+
+	struct edma_dmal_s *pdmalbase_v;
+	u8 *pdmalbase_p;
+
+	u32 dmal_cnt;
+	u32 dmal_byte;
+
+	u32 count;
+	u32 size;
+
+	u32 head;
+	u32 tail;
+
+	u16 start_dma;
+	u16 dmacmperr;
+	u16 dma_overtime;
+
+	u32 work_limit;
+
+	struct edma_rxtx_statistics s;
+};
+
+struct edma_eth_dev_s {
+	struct edma_rxtx_q_s *ptx_queue;
+	struct edma_rxtx_q_s *prx_queue;
+
+	struct edma_packet_node_s *rx_packet;
+	spinlock_t rx_queue_lock; /* spinlock for rx queue */
+
+	u32 rx_packet_head;
+	u32 rx_packet_tail;
+
+	unsigned long pages_tx;
+	unsigned long pages_rx;
+
+	u8 *pshmpool_p;
+	u8 *pshmpool_v;
+	u32 shmpoolsize;
+
+	u32 recv_int;
+	u32 tobmc_int;
+	u32 run_dma_TX_task;
+	u32 run_dma_RX_task;
+	u32 run_skb_RX_task;
+
+	struct tasklet_struct skb_task;
+	struct tasklet_struct dma_task;
+
+	struct cdev_dev_s cdev;
+	__kernel_time_t init_time;
+
+	void *edma_priv;
+};
+
+#ifndef LOG
+#define LOG(level, fmt, ...) do {\
+	if (debug >= (level)) {\
+		netdev_err(0, "[%s,%d] -> " fmt "\n", \
+			   __func__, __LINE__, ##__VA_ARGS__); \
+	} \
+} while (0)
+#endif
+
+#define BD_QUEUE_MASK(p) ((p) & (BSPVETH_POINT_MASK))
+
+#define GET_BD_RING_QUEUE_COUNT(head, tail) \
+	((BSPVETH_MAX_QUE_DEEP + (head) - (tail)) & BSPVETH_POINT_MASK)
+#define GET_BD_RING_QUEUE_SPACE(head, tail) \
+	((BSPVETH_MAX_QUE_DEEP - 1 + (tail) - (head)) & BSPVETH_POINT_MASK)
+#define JUDGE_RING_QUEUE_SPACE(head, tail, len) \
+	(GET_BD_RING_QUEUE_SPACE(head, tail) >= (len))
+
+#define CHECK_DMA_QUEUE_EMPTY(type, queue) \
+	(((type) == BSPVETH_RX && \
+	 (queue)->pshmqhd_v->head == (queue)->pshmqhd_v->tail) || \
+	 ((type) == BSPVETH_TX && (queue)->head == (queue)->tail))
+
+#define CHECK_DMA_RXQ_FAULT(queue, type, cnt) \
+	((type) == BSPVETH_RX && (queue)->dmal_cnt > 1 && \
+	 (cnt) < ((queue)->work_limit / 2))
+
+#define GET_DMA_DIRECTION(type) \
+	(((type) == BSPVETH_RX) ? BMC_TO_HOST : HOST_TO_BMC)
+
+/******* rate limit *********/
+#define RL_MAX_PACKET 10
+#define RL_STRESS_LOW 50
+#define RL_STRESS_HIGH 80
+#define RL_DELAY_MS_LOW 20
+#define RL_DELAY_MS_HIGH 100
+
+void veth_dma_task_H(u32 type);
+void veth_skbtimer_close(void);
+int veth_skbtimer_init(void);
+int veth_dmatimer_close_H(void);
+int veth_dmatimer_init_H(void);
+int veth_skb_tr_task(unsigned long data);
+
+#endif
+
-- 
2.26.2.windows.1


Powered by blists - more mailing lists

Powered by Openwall GNU/*/Linux Powered by OpenVZ