| 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
| ||
|
Message-ID: <20080604082514.241e7aad@mailhost.serverengines.com>
Date: Wed, 04 Jun 2008 01:25:14 -0700
From: "Subbu Seetharaman" <subbus@...verengines.com>
To: "Stephen Hemminger" <shemminger@...tta.com>
Cc: netdev@...r.kernel.org
Subject: Re: [PATCH 1/12] BE NIC driver - Header files and initialization
functions
Thanks for you comments Stephen.
We did not make MSIx the default since most testing
was done with INTx. Since this version of the driver does not use multiple
interrupt vectors, there is no performance advantage in using MSIx.
If MSIx is still the preferred default, we can change it.
NAPI makes tasklet redundant in the RX completion processing; but
we felt deffering TX completion processing to a tasklet is a better
idea for 10Gb data rates than doing it in hard interrupt.
I will get back soon with other comments taken care of.
Subbu
----- Original Message -----
From: Stephen Hemminger [mailto:shemminger@...tta.com]
To: Subbu Seetharaman [mailto:subbus@...verengines.com]
Cc: netdev@...r.kernel.org
Sent: Tue, 03 Jun 2008 10:16:56 -0700
Subject: Re: [PATCH 1/12] BE NIC driver - Header files and initialization functions
Overall this part looks pretty good.
> Thanks to everyone who reviewed the previous two submissions of the NIC driver
> for BladeEngine (ServerEngines' 10Gb NIC) driver. I am submitting the driver
> with changes suggested in the last review.
>
> BladeEngine is a dual function device with network and storage functions.
> This patch includes the network driver and beclib.
> beclib is a library of function that the driver uses to access the
> hardware. It is common to both storage and network drivers and
> hence organized under the directory drivers/message/beclib. The
> storage driver is not part of this patch and will be submitted after
> this review.
>
> This patch is made against the current git tree.
>
> Thank you.
> Signed-off-by: Subbu Seetharaman <subbus@...verengines.com>
> ---
> drivers/net/benet/be_init.c | 1135 +++++++++++++++++++++++++++++++++++++++++++
> drivers/net/benet/benet.h | 301 ++++++++++++
> drivers/net/benet/bni.h | 327 +++++++++++++
> 3 files changed, 1763 insertions(+), 0 deletions(-)
> create mode 100644 drivers/net/benet/be_init.c
> create mode 100644 drivers/net/benet/benet.h
> create mode 100644 drivers/net/benet/bni.h
>
> diff --git a/drivers/net/benet/be_init.c b/drivers/net/benet/be_init.c
> new file mode 100644
> index 0000000..2eb9ce7
> --- /dev/null
> +++ b/drivers/net/benet/be_init.c
> @@ -0,0 +1,1135 @@
> +/*
> + * Copyright (C) 2005 - 2008 ServerEngines
> + * All rights reserved.
> + *
> + * This program is free software; you can redistribute it and/or
> + * modify it under the terms of the GNU General Public License version 2
> + * as published by the Free Software Foundation. The full GNU General
> + * Public License is included in this distribution in the file called COPYING.
> + *
> + * Contact Information:
> + * linux-drivers@...verengines.com
> + *
> + * ServerEngines
> + * 209 N. Fair Oaks Ave
> + * Sunnyvale, CA 94085
> + */
> +
> +#include <linux/pci.h>
> +#include <linux/etherdevice.h>
> +
> +#include "benet.h"
> +
> +#define DRVR_VERSION "1.0.688"
> +
> +static struct pci_device_id be_device_id_table[] = {
> + {PCI_DEVICE(0x19a2, 0x0201)},
> + {0}
> +};
> +
> +MODULE_DEVICE_TABLE(pci, be_device_id_table);
> +
> +MODULE_VERSION(DRVR_VERSION);
> +
> +#define DRV_DESCRIPTION "ServerEngines BladeEngine Network Driver Version "
> +
> +MODULE_DESCRIPTION(DRV_DESCRIPTION DRVR_VERSION);
> +MODULE_AUTHOR("ServerEngines");
> +MODULE_LICENSE("GPL");
> +
> +static unsigned int msix; /* default - msix disabled */
> +module_param(msix, uint, S_IRUGO);
> +MODULE_PARM_DESC(msix, "Use MSI-x interrupts");
Why is msix off by default? hardware problem?
> +static unsigned int rxbuf_size = 2048; /* Size of RX buffers */
> +module_param(rxbuf_size, uint, S_IRUGO);
> +MODULE_PARM_DESC(rxbuf_size, "Size of buffers to hold Rx data");
> +
> +const char be_drvr_ver[] = DRVR_VERSION;
> +char be_fw_ver[32]; /* F/W version filled in by be_probe */
> +char be_driver_name[] = "benet";
> +
> +/*
> + * Number of entries in each queue.
> + */
> +#define EVENT_Q_LEN 1024
> +#define ETH_TXQ_LEN 2048
> +#define ETH_TXCQ_LEN 1024
> +#define ETH_RXQ_LEN 1024 /* Does not support any other value */
> +#define ETH_UC_RXCQ_LEN 1024
> +#define ETH_BC_RXCQ_LEN 256
> +#define MCC_Q_LEN 64 /* total size not to exceed 8 pages */
> +#define MCC_CQ_LEN 256
> +
> +/*
> + * Intialize and register a network device for the pnob. Initialize to No Link.
> + * Link will be enabled during benet_open() or when physical Link is up
> + */
> +static int
> +be_netdev_init(struct be_adapter *adapter, struct bni_net_object *pnob)
> +{
> + struct net_device *netdev = OSM_NOB(pnob)->netdev;
> + int ret = 0;
> +
> + bni_get_uc_mac_adrr(pnob, 0, 0, OSM_NOB(pnob)->devno,
> + (u8 *)netdev->dev_addr, NULL, NULL);
> +
> + netdev->init = &benet_probe;
> + netif_carrier_off(netdev);
> + netif_stop_queue(netdev);
> +
> + SET_NETDEV_DEV(netdev, &(adapter->pdev->dev));
> + ret = register_netdev(netdev);
> + return ret;
> +}
Why not?
return register_netdev(netdev);
> +/* Initialize the pci_info structure for this function */
> +static int
> +init_pci_be_function(struct be_adapter *adapter, struct pci_dev *pdev)
> +{
> + adapter->num_bars = 3;
> + /* CSR */
> + adapter->pci_bars[0].base_pa = pci_resource_start(pdev, 2);
> + adapter->pci_bars[0].base_va =
> + ioremap_nocache(adapter->pci_bars[0].base_pa,
> + pci_resource_len(pdev, 2));
> + if (adapter->pci_bars[0].base_va == NULL)
> + return -ENOMEM;
> + adapter->pci_bars[0].length = sizeof(struct BLADE_ENGINE_CSRMAP_AMAP);
> +
> + /* Door Bell */
> + adapter->pci_bars[1].base_pa = pci_resource_start(pdev, 4);
> + adapter->pci_bars[1].base_va =
> + ioremap_nocache(adapter->pci_bars[1].base_pa, (128 * 1024));
> + if (adapter->pci_bars[1].base_va == NULL) {
> + iounmap(adapter->pci_bars[0].base_va);
> + return -ENOMEM;
> + }
> + adapter->pci_bars[1].length =
> + sizeof(struct PROTECTION_DOMAIN_DBMAP_AMAP);
> +
> + /* PCI */
> + adapter->pci_bars[2].base_pa = pci_resource_start(pdev, 1);
> + adapter->pci_bars[2].length = pci_resource_len(pdev, 1);
> + adapter->pci_bars[2].base_va =
> + ioremap_nocache(adapter->pci_bars[2].base_pa,
> + adapter->pci_bars[2].length);
> + if (adapter->pci_bars[2].base_va == NULL) {
> + iounmap(adapter->pci_bars[0].base_va);
> + iounmap(adapter->pci_bars[1].base_va);
> + return -ENOMEM;
> + }
> +
> +
> + return 0;
> +}
> +
> +/*
> + * Enable MSIx and return 1 if successful. Else return 0
> + */
> +static int be_enable_msix(struct be_adapter *adapter)
> +{
> + unsigned int i, ret;
> +
> + if (!msix)
> + return 0;
> +
> + adapter->msix_enabled = 1;
> +
> + for (i = 0; i < BE_MAX_REQ_MSIX_VECTORS; i++)
> + adapter->msix_entries[i].entry = i;
> +
> + ret = pci_enable_msix(adapter->pdev,
> + adapter->msix_entries, BE_MAX_REQ_MSIX_VECTORS);
> +
> + if (ret) {
> + adapter->msix_enabled = 0;
> + return 0;
> + }
> +
> + return 1;
> +}
> +
> +/*
> + * Registers ISR for BE. Uses MSIx interrupt if configured and requested.
> + * If not, uses INTx interrupt. Returns 0 for success and -1 for filure.
> + */
> +static int
> +be_register_isr(struct be_adapter *adapter, struct bni_net_object *pnob)
> +{
> + int msix_intr, r;
> + struct net_device *netdev = OSM_NOB(pnob)->netdev;
> + u32 msix_ret = 0;
> +
> + netdev->irq = adapter->pdev->irq;
> +
> + msix_intr = 0;
> + msix_ret = be_enable_msix(adapter);
> + if (msix_ret) {
> + r = request_irq(adapter->msix_entries[0].vector,
> + be_int, IRQF_SHARED, netdev->name, netdev);
> + if (r) {
> + printk(KERN_WARNING
> + "MSIX Request IRQ failed - Errno %d\n", r);
> + } else {
> + msix_intr = 1;
> + }
> + }
> + if (msix_intr == 0) {
> + /* request legacy INTx interrupt */
> + r = request_irq(netdev->irq, be_int, IRQF_SHARED,
> + netdev->name, netdev);
> + if (r) {
> + printk(KERN_ERR
> + "INTx Request IRQ failed - Errno %d\n", r);
> + return (-1);
> + }
> + }
> + return (0);
> +}
> +
> +/*
> + * free all resources associated with a pnob
> + * Called at the time of module cleanup as well a any error during
> + * module init. Some resources may be partially allocated in a NetObj.
> + */
> +static void netobject_cleanup(struct be_adapter *adapter,
> + struct bni_net_object *pnob)
> +{
> + struct net_device *netdev;
> + struct sk_buff *skb;
> + int i;
> +
> + netdev = adapter->netdevp;
> +
> + if (netif_running(netdev)) {
> + netif_stop_queue(netdev);
> + be_wait_nic_tx_cmplx_cmpl(pnob);
> + bni_disable_eq_intr(pnob);
> + }
> +
> + if (adapter->isr_registered && adapter->msix_enabled)
> + free_irq(adapter->msix_entries[0].vector, netdev);
> + else if (adapter->isr_registered && !adapter->msix_enabled)
> + free_irq(netdev->irq, netdev);
> +
> + adapter->isr_registered = 0;
> + if (adapter->msix_enabled) {
> + pci_disable_msix(adapter->pdev);
> + adapter->msix_enabled = 0;
> + }
> + if (adapter->tasklet_started) {
> + tasklet_kill(&(adapter->sts_handler));
> + adapter->tasklet_started = 0;
> + }
> + if (pnob->fn_obj_created)
> + bni_disable_intr(pnob);
> +
> + /* In cases of partial initialization, it's OK to call unregister
> + * even if netdev is not registered: handled in unregister_netdev()
> + */
> + unregister_netdev(netdev);
> +
> + if (pnob->fn_obj_created)
> + bni_destroy_netobj(pnob, &adapter->sa_device);
> +
> + adapter->net_obj = NULL;
> + adapter->netdevp = NULL;
> +
> + if (pnob->mcc_q)
> + pci_free_consistent(adapter->pdev, pnob->mcc_q_size,
> + pnob->mcc_q, pnob->mcc_q_bus);
> +
> + if (pnob->mcc_wrb_ctxt)
> + free_pages((unsigned long)pnob->mcc_wrb_ctxt,
> + get_order(pnob->mcc_wrb_ctxt_size));
> +
> + if (pnob->mcc_cq)
> + pci_free_consistent(adapter->pdev, pnob->mcc_cq_size,
> + pnob->mcc_cq, pnob->mcc_cq_bus);
> +
> + if (pnob->event_q)
> + pci_free_consistent(adapter->pdev, pnob->event_q_size,
> + pnob->event_q, pnob->event_q_bus);
> +
> + if (pnob->tx_cq)
> + pci_free_consistent(adapter->pdev, pnob->tx_cq_size,
> + pnob->tx_cq, pnob->tx_cq_bus);
> +
> + if (pnob->tx_q)
> + pci_free_consistent(adapter->pdev, pnob->tx_q_size,
> + pnob->tx_q, pnob->tx_q_bus);
> +
> + if (pnob->bcrx_cq)
> + pci_free_consistent(adapter->pdev, pnob->bcrx_cq_size,
> + pnob->bcrx_cq, pnob->bcrx_cq_bus);
> +
> + if (pnob->rx_q)
> + pci_free_consistent(adapter->pdev, pnob->rx_q_size,
> + pnob->rx_q, pnob->rx_q_bus);
> +
> + if (pnob->ucrx_cq)
> + pci_free_consistent(adapter->pdev, pnob->ucrx_cq_size,
> + pnob->ucrx_cq, pnob->ucrx_cq_bus);
> +
> + if (pnob->rx_ctxt) {
> + struct be_rx_page_info *rx_page_info;
> + for (i = 0; i < pnob->rx_q_len; i++) {
> + rx_page_info = &(OSM_NOB(pnob)->rx_page_info[i]);
> + if ((OSM_NOB(pnob)->rx_pg_shared == FALSE) ||
> + (rx_page_info->page_offset)) {
> + pci_unmap_page(adapter->pdev,
> + pci_unmap_addr(rx_page_info,
> + bus),
> + pnob->rx_buf_size,
> + PCI_DMA_FROMDEVICE);
> + }
> + if (rx_page_info->page)
> + put_page(rx_page_info->page);
> + memset(rx_page_info, 0, sizeof(struct be_rx_page_info));
> + }
> + OSM_NOB(pnob)->rx_pg_info_hd = 0;
> + kfree(OSM_NOB(pnob)->rx_page_info);
> + kfree(pnob->rx_ctxt);
> + }
> +
> + if (pnob->tx_ctxt) {
> + for (i = 0; i < pnob->tx_q_len; i++) {
> + skb = (struct sk_buff *)pnob->tx_ctxt[i];
> + if (skb)
> + kfree_skb(skb);
> + }
> + kfree(pnob->tx_ctxt);
> + }
> +
> + if (pnob->mb_ptr)
> + pci_free_consistent(adapter->pdev, pnob->mb_size, pnob->mb_ptr,
> + pnob->mb_bus);
> +
> + if (OSM_NOB(pnob))
> + kfree(OSM_NOB(pnob));
Check for null not needed, kfree() handles it.
> + free_netdev(netdev);
> +}
> +
> +static int be_nob_ring_create(struct be_adapter *adapter)
> +{
> + struct bni_net_object *pnob = NULL;
> + u32 size;
> + struct net_device *netdev;
> +
> + /* Allocate nob as a part of netdev */
> + netdev = alloc_etherdev(sizeof(struct bni_net_object));
> + if (netdev == NULL)
> + return -ENOMEM;
> + pnob = netdev->priv;
> + memset(pnob, 0, sizeof(struct bni_net_object));
This memset is unnecessary, alloc_etherdev returns zeroed memory.
> + adapter->net_obj = pnob;
> + adapter->netdevp = netdev;
> +
> + pnob->osm_netobj = (struct linux_net_object *)
> + kzalloc(sizeof(struct linux_net_object), GFP_KERNEL);
> + if (pnob->osm_netobj == NULL)
> + return -1;
> + OSM_NOB(pnob)->devno = 0;
> + OSM_NOB(pnob)->adapter = adapter;
> + OSM_NOB(pnob)->netdev = netdev;
> +
> + /* Mail box sgl; mailbox pointer needs to be 16 byte aligned */
> + pnob->mb_size = sizeof(struct MCC_MAILBOX_AMAP) + 16;
> + pnob->mb_ptr = pci_alloc_consistent(adapter->pdev, pnob->mb_size,
> + &pnob->mb_bus);
> + if (!pnob->mb_bus)
> + return -1;
> + memset(pnob->mb_ptr, 0, pnob->mb_size);
> + sa_sgl_create_contiguous(PTR_ALIGN(pnob->mb_ptr, 16),
> + PTR_ALIGN(pnob->mb_bus, 16), sizeof(struct MCC_MAILBOX_AMAP),
> + &pnob->mb_sgl);
> +
> + /*
> + * Event queue
> + */
> + pnob->event_q_len = EVENT_Q_LEN;
> + pnob->event_q_size = pnob->event_q_len * sizeof(struct EQ_ENTRY_AMAP);
> + pnob->event_q = pci_alloc_consistent(adapter->pdev, pnob->event_q_size,
> + &pnob->event_q_bus);
> + if (!pnob->event_q_bus)
> + return -1;
> + /*
> + * Eth TX queue
> + */
> + pnob->tx_q_len = ETH_TXQ_LEN;
> + pnob->tx_q_port = 0;
> + pnob->tx_q_size = pnob->tx_q_len * sizeof(struct ETH_WRB_AMAP);
> + pnob->tx_q = pci_alloc_consistent(adapter->pdev, pnob->tx_q_size,
> + &pnob->tx_q_bus);
> + if (!pnob->tx_q_bus)
> + return -1;
> + /*
> + * Eth TX Compl queue
> + */
> + pnob->txcq_len = ETH_TXCQ_LEN;
> + pnob->tx_cq_size = pnob->txcq_len * sizeof(struct ETH_TX_COMPL_AMAP);
> + pnob->tx_cq = pci_alloc_consistent(adapter->pdev, pnob->tx_cq_size,
> + &pnob->tx_cq_bus);
> + if (!pnob->tx_cq_bus)
> + return -1;
> + /*
> + * Eth RX queue
> + */
> + pnob->rx_q_len = ETH_RXQ_LEN;
> + pnob->rx_q_size = pnob->rx_q_len * sizeof(struct ETH_RX_D_AMAP);
> + pnob->rx_q = pci_alloc_consistent(adapter->pdev, pnob->rx_q_size,
> + &pnob->rx_q_bus);
> + if (!pnob->rx_q_bus)
> + return -1;
> + /*
> + * Eth Unicast RX Compl queue
> + */
> + pnob->ucrx_cq_len = ETH_UC_RXCQ_LEN;
> + pnob->ucrx_cq_size = pnob->ucrx_cq_len *
> + sizeof(struct ETH_RX_COMPL_AMAP);
> + pnob->ucrx_cq = pci_alloc_consistent(adapter->pdev, pnob->ucrx_cq_size,
> + &pnob->ucrx_cq_bus);
> + if (!pnob->ucrx_cq_bus)
> + return -1;
> + /*
> + * Eth Broadcast RX Compl queue
> + */
> + pnob->bcrx_cq_len = ETH_BC_RXCQ_LEN;
> + pnob->bcrx_cq_size = pnob->bcrx_cq_len *
> + sizeof(struct ETH_RX_COMPL_AMAP);
> + pnob->bcrx_cq = pci_alloc_consistent(adapter->pdev, pnob->bcrx_cq_size,
> + &pnob->bcrx_cq_bus);
> + if (!pnob->bcrx_cq_bus)
> + return -1;
> +
> + /* TX resources */
> + size = pnob->tx_q_len * sizeof(void **);
> + pnob->tx_ctxt = kmalloc(size, GFP_KERNEL);
> + if (pnob->tx_ctxt == NULL)
> + return -1;
> +
> + /* RX resources */
> + size = pnob->rx_q_len * sizeof(void *);
> + pnob->rx_ctxt = kmalloc(size, GFP_KERNEL);
> + if (pnob->rx_ctxt == NULL)
> + return -1;
> +
> + size = (pnob->rx_q_len * sizeof(struct be_rx_page_info));
> + OSM_NOB(pnob)->rx_page_info = kzalloc(size, GFP_KERNEL);
> + if (OSM_NOB(pnob)->rx_page_info == NULL)
> + return -1;
> +
> + adapter->eth_statsp = (struct FWCMD_ETH_GET_STATISTICS *)
> + kmalloc(sizeof(struct FWCMD_ETH_GET_STATISTICS), GFP_KERNEL);
> + if (adapter->eth_statsp == NULL)
> + return -1;
> + pnob->rx_buf_size = rxbuf_size;
> + adapter->dev_state = BE_DEV_STATE_NONE;
> + return 0;
> +}
> +
> +static int be_nob_ring_init(struct be_adapter *adapter,
> + struct bni_net_object *pnob)
> +{
> + struct sa_dev_bar_locations pci_bars[3];
> + int status;
> +
> + memset(pnob->event_q, 0, pnob->event_q_size);
> + pnob->event_q_tl = 0;
> +
> + memset(pnob->tx_q, 0, pnob->tx_q_size);
> + pnob->tx_q_hd = 0;
> + pnob->tx_q_tl = 0;
> +
> + memset(pnob->tx_cq, 0, pnob->tx_cq_size);
> + pnob->tx_cq_tl = 0;
> +
> + memset(pnob->rx_q, 0, pnob->rx_q_size);
> +
> + memset(pnob->ucrx_cq, 0, pnob->ucrx_cq_size);
> + pnob->ucrx_cq_tl = 0;
> +
> + memset(pnob->bcrx_cq, 0, pnob->bcrx_cq_size);
> + pnob->bcrx_cq_tl = 0;
> +
> + memset(pnob->tx_ctxt, 0, pnob->tx_q_len * sizeof(void **));
> + memset(pnob->rx_ctxt, 0, pnob->rx_q_len * sizeof(void *));
> + memset(OSM_NOB(pnob)->rx_page_info, 0,
> + pnob->rx_q_len * sizeof(struct be_rx_page_info));
> + OSM_NOB(pnob)->rx_pg_info_hd = 0;
> + pnob->rx_q_hd = 0;
> + atomic_set(&pnob->rx_q_posted, 0);
> +
> + memcpy(pci_bars, adapter->pci_bars, sizeof(adapter->pci_bars));
> + status = bni_create_netobj(pnob, pci_bars, adapter->num_bars,
> + &adapter->sa_device, &adapter->chip_object);
> + if (status != BE_SUCCESS)
> + return -1;
> +
> + be_post_eth_rx_buffs(pnob);
> + return 0;
> +}
> +
> +/* This function handles async callback for link status */
> +static void
> +be_link_status_async_callback(void *context, u32 event_code, void *event)
> +{
> + struct ASYNC_EVENT_LINK_STATE_AMAP *link_status = event;
> + struct be_adapter *adapter = context;
> + bool link_enable = FALSE;
> + struct bni_net_object *pnob;
> + struct ASYNC_EVENT_TRAILER_AMAP *async_trailer;
> + struct net_device *netdev;
> + u32 async_event_code, async_event_type, active_port;
> + u32 port0_link_status, port1_link_status, port0_duplex, port1_duplex;
> + u32 port0_speed, port1_speed;
> +
> + if (event_code != ASYNC_EVENT_CODE_LINK_STATE) {
> + /* Not our event to handle */
> + return;
> + }
> + async_trailer = (struct ASYNC_EVENT_TRAILER_AMAP *)
> + ((u8 *) event + sizeof(struct MCC_CQ_ENTRY_AMAP) -
> + sizeof(struct ASYNC_EVENT_TRAILER_AMAP));
> +
> + async_event_code = AMAP_GET_BITS_PTR(ASYNC_EVENT_TRAILER, event_code,
> + async_trailer);
> + BUG_ON(async_event_code != ASYNC_EVENT_CODE_LINK_STATE);
> +
> + pnob = adapter->net_obj;
> + netdev = OSM_NOB(pnob)->netdev;
> +
> + /* Determine if this event is a switch VLD or a physical link event */
> + async_event_type = AMAP_GET_BITS_PTR(ASYNC_EVENT_TRAILER, event_type,
> + async_trailer);
> + active_port = AMAP_GET_BITS_PTR(ASYNC_EVENT_LINK_STATE,
> + active_port, link_status);
> + port0_link_status = AMAP_GET_BITS_PTR(ASYNC_EVENT_LINK_STATE,
> + port0_link_status, link_status);
> + port1_link_status = AMAP_GET_BITS_PTR(ASYNC_EVENT_LINK_STATE,
> + port1_link_status, link_status);
> + port0_duplex = AMAP_GET_BITS_PTR(ASYNC_EVENT_LINK_STATE,
> + port0_duplex, link_status);
> + port1_duplex = AMAP_GET_BITS_PTR(ASYNC_EVENT_LINK_STATE,
> + port1_duplex, link_status);
> + port0_speed = AMAP_GET_BITS_PTR(ASYNC_EVENT_LINK_STATE,
> + port0_speed, link_status);
> + port1_speed = AMAP_GET_BITS_PTR(ASYNC_EVENT_LINK_STATE,
> + port1_speed, link_status);
> + if (async_event_type == NTWK_LINK_TYPE_VIRTUAL) {
> + adapter->be_stat.bes_link_change_virtual++;
> + if (adapter->be_link_sts->active_port != active_port) {
> + dev_notice(&netdev->dev,
> + "Active port changed due to VLD on switch\n");
> + } else {
> + dev_notice(&netdev->dev, "Link status update\n");
> + }
> +
> + } else {
> + adapter->be_stat.bes_link_change_physical++;
> + if (adapter->be_link_sts->active_port != active_port) {
> + dev_notice(&netdev->dev,
> + "Active port changed due to port link"
> + " status change\n");
> + } else {
> + dev_notice(&netdev->dev, "Link status update\n");
> + }
> + }
> +
> + memset(adapter->be_link_sts, 0, sizeof(adapter->be_link_sts));
> +
> + if ((port0_link_status == ASYNC_EVENT_LINK_UP) ||
> + (port1_link_status == ASYNC_EVENT_LINK_UP)) {
> + if ((adapter->port0_link_sts == BE_PORT_LINK_DOWN) &&
> + (adapter->port1_link_sts == BE_PORT_LINK_DOWN)) {
> + /* Earlier both the ports are down So link is up */
> + link_enable = TRUE;
> + }
> +
> + if (port0_link_status == ASYNC_EVENT_LINK_UP) {
> + adapter->port0_link_sts = BE_PORT_LINK_UP;
> + adapter->be_link_sts->mac0_duplex = port0_duplex;
> + adapter->be_link_sts->mac0_speed = port0_speed;
> + if (active_port == NTWK_PORT_A)
> + adapter->be_link_sts->active_port = 0;
> + } else
> + adapter->port0_link_sts = BE_PORT_LINK_DOWN;
> +
> + if (port1_link_status == ASYNC_EVENT_LINK_UP) {
> + adapter->port1_link_sts = BE_PORT_LINK_UP;
> + adapter->be_link_sts->mac1_duplex = port1_duplex;
> + adapter->be_link_sts->mac1_speed = port1_speed;
> + if (active_port == NTWK_PORT_B)
> + adapter->be_link_sts->active_port = 1;
> + } else
> + adapter->port1_link_sts = BE_PORT_LINK_DOWN;
> +
> + printk(KERN_INFO "Link Properties for %s:\n", netdev->name);
> + be_print_link_info(adapter->be_link_sts);
> +
> + if (!link_enable)
> + return;
> + /*
> + * Both ports were down previously, but atleast one of
> + * them has come up if this netdevice's carrier is not up,
> + * then indicate to stack
> + */
> + if (!netif_carrier_ok(netdev)) {
> + netif_start_queue(netdev);
> + netif_carrier_on(netdev);
> + }
> + return;
> + }
> +
> + /* Now both the ports are down. Tell the stack about it */
> + dev_info(&netdev->dev, "Both ports are down\n");
> + adapter->port0_link_sts = BE_PORT_LINK_DOWN;
> + adapter->port1_link_sts = BE_PORT_LINK_DOWN;
> + if (netif_carrier_ok(netdev)) {
> + netif_carrier_off(netdev);
> + netif_stop_queue(netdev);
> + }
> + return;
> +}
> +
> +static int be_mcc_create(struct be_adapter *adapter)
> +{
> + struct bni_net_object *pnob;
> +
> + pnob = adapter->net_obj;
> + /*
> + * Create the MCC ring so that all further communication with
> + * MCC can go thru the ring. we do this at the end since
> + * we do not want to be dealing with interrupts until the
> + * initialization is complete.
> + */
> + pnob->mcc_q_len = MCC_Q_LEN;
> + pnob->mcc_q_size = pnob->mcc_q_len * sizeof(struct MCC_WRB_AMAP);
> + pnob->mcc_q = pci_alloc_consistent(adapter->pdev, pnob->mcc_q_size,
> + &pnob->mcc_q_bus);
> + if (!pnob->mcc_q_bus)
> + return -1;
> + /*
> + * space for MCC WRB context
> + */
> + pnob->mcc_wrb_ctxtLen = MCC_Q_LEN;
> + pnob->mcc_wrb_ctxt_size = pnob->mcc_wrb_ctxtLen *
> + sizeof(struct be_mcc_wrb_context);
> + pnob->mcc_wrb_ctxt = (void *)__get_free_pages(GFP_KERNEL,
> + get_order(pnob->mcc_wrb_ctxt_size));
> + if (pnob->mcc_wrb_ctxt == NULL)
> + return -1;
> + /*
> + * Space for MCC compl. ring
> + */
> + pnob->mcc_cq_len = MCC_CQ_LEN;
> + pnob->mcc_cq_size = pnob->mcc_cq_len * sizeof(struct MCC_CQ_ENTRY_AMAP);
> + pnob->mcc_cq = pci_alloc_consistent(adapter->pdev, pnob->mcc_cq_size,
> + &pnob->mcc_cq_bus);
> + if (!pnob->mcc_cq_bus)
> + return -1;
> + return 0;
> +}
> +
> +static int be_mcc_init(struct be_adapter *adapter)
> +{
> + u32 r;
> + struct bni_net_object *pnob;
> +
> + pnob = adapter->net_obj;
> + memset(pnob->mcc_q, 0, pnob->mcc_q_size);
> + pnob->mcc_q_hd = 0;
> +
> + memset(pnob->mcc_wrb_ctxt, 0, pnob->mcc_wrb_ctxt_size);
> +
> + memset(pnob->mcc_cq, 0, pnob->mcc_cq_size);
> + pnob->mcc_cq_tl = 0;
> +
> + r = bni_create_mcc_rings(adapter->net_obj);
> + if (r != BE_SUCCESS)
> + return -1;
> +
> + return 0;
> +}
> +
> +static void be_remove(struct pci_dev *pdev)
> +{
> + struct bni_net_object *pnob;
> + struct be_adapter *adapter;
> + int i;
> +
> + adapter = pci_get_drvdata(pdev);
> + if (!adapter)
> + return;
> +
> + pci_set_drvdata(pdev, 0);
> + pnob = (struct bni_net_object *)adapter->net_obj;
> +
> + flush_scheduled_work();
> +
> + if (pnob) {
> + /* Unregister async callback function for link status updates */
> + if (pnob->mcc_q_created)
> + be_mcc_add_async_event_callback(
> + &pnob->mcc_q_obj, NULL, NULL);
> +
> + netobject_cleanup(adapter, pnob);
> + }
> +
> + bni_cleanup(&adapter->chip_object);
> +
> + for (i = 0; i < adapter->num_bars; i++) {
> + if (adapter->pci_bars[i].base_va)
> + iounmap(adapter->pci_bars[i].base_va);
> + }
> + pci_release_regions(adapter->pdev);
> + pci_disable_device(adapter->pdev);
> +
> + if (adapter->be_link_sts)
> + kfree(adapter->be_link_sts);
Another case where if() check is unneeded for kfree.
> + if (adapter->eth_statsp)
> + kfree(adapter->eth_statsp);
> +
> + if (adapter->timer_ctxt.get_stats_timer.function)
> + del_timer_sync(&adapter->timer_ctxt.get_stats_timer);
> + kfree(adapter);
> +}
> +
> +/*
> + * This function is called by the PCI sub-system when it finds a PCI
> + * device with dev/vendor IDs that match with one of our devices.
> + * All of the driver initialization is done in this function.
> + */
> +static int be_probe(struct pci_dev *pdev, const struct pci_device_id *pdev_id)
> +{
> + int status = 0;
> + struct be_adapter *adapter = NULL;
> + struct FWCMD_COMMON_GET_FW_VERSION_RESPONSE_PAYLOAD get_fwv;
> + struct bni_net_object *pnob = NULL;
> + struct linux_net_object *lno;
> +
> + status = pci_enable_device(pdev);
> + if (status) {
> + dev_err(&pdev->dev, "pci_enable_device() failed");
> + goto error;
> + }
> +
> + status = pci_request_regions(pdev, be_driver_name);
> + if (status) {
> + pci_disable_device(pdev);
> + goto error;
> + }
> +
> + pci_set_master(pdev);
> + adapter = kzalloc(sizeof(struct be_adapter), GFP_KERNEL);
> + if (adapter == NULL) {
> + pci_release_regions(pdev);
> + pci_disable_device(pdev);
> + status = -ENOMEM;
> + goto error;
> + }
> +
> + pci_set_drvdata(pdev, adapter);
> + adapter->pdev = pdev;
> +
> + /* Adapative interrupt coalescing limits in usecs.
> + * should be a multiple of 8.
> + */
> + adapter->enable_aic = 1;
> + adapter->max_eqd = MAX_EQD;
> + adapter->min_eqd = 0;
> + adapter->cur_eqd = 0;
> + status = pci_set_dma_mask(pdev, DMA_64BIT_MASK);
> + if (!status) {
> + /* Device is DAC Capable. */
> + adapter->dma_64bit_cap = TRUE;
> + } else {
> + adapter->dma_64bit_cap = FALSE;
> + status = pci_set_dma_mask(pdev, DMA_32BIT_MASK);
> + if (status != 0) {
> + printk(KERN_ERR "Could not set PCI DMA Mask\n");
> + goto cleanup;
> + }
> + }
> +
> + status = init_pci_be_function(adapter, pdev);
> + if (status != 0) {
> + printk(KERN_ERR "Failed to map PCI BARS\n");
> + status = -ENOMEM;
> + goto cleanup;
> + }
> +
> + sa_trace_set_level(DL_ALWAYS | DL_ERR);
> +
> + status = bni_init(&adapter->chip_object);
> + if (status != 0) {
> + printk(KERN_ERR "bni_init() failed - Error %d\n", status);
> + goto cleanup;
> + }
> +
> + adapter->be_link_sts = (struct BE_LINK_STATUS *)
> + kmalloc(sizeof(struct BE_LINK_STATUS), GFP_KERNEL);
> + if (adapter->be_link_sts == NULL) {
> + printk(KERN_ERR "Memory allocation for link status "
> + "buffer failed\n");
> + goto cleanup;
> + }
> + spin_lock_init(&adapter->txq_lock);
> +
> + status = be_nob_ring_create(adapter);
> + if (status != 0)
> + goto cleanup;
> + pnob = adapter->net_obj;
> + lno = OSM_NOB(pnob);
> +
> + status = be_nob_ring_init(adapter, pnob);
> + if (status != 0)
> + goto cleanup;
> +
> + status = be_netdev_init(adapter, pnob);
> + if (status != 0)
> + goto cleanup;
> +
> +#ifdef CONFIG_BENET_NAPI
Just do NAPI, having a configuration option means more possible
test cases and distributions only ship one binary anyway.
> + netif_napi_add(lno->netdev, &lno->napi, be_poll, 64);
> + lno->rx_sched = FALSE;
> + spin_lock_init(&lno->rx_lock);
You should not need this lock, if:
1) you use NAPI properly,
2) you get rid of the unnecessary state flags like rx_sched etc..
> +#endif
> +
> + /* if the rx_frag size if 2K, one page is shared as two RX frags */
> + lno->rx_pg_shared = (pnob->rx_buf_size <= PAGE_SIZE / 2)? TRUE : FALSE;
> + if (pnob->rx_buf_size != rxbuf_size) {
> + printk(KERN_WARNING
> + "Could not set Rx buffer size to %d. Using %d\n",
> + rxbuf_size, pnob->rx_buf_size);
> + rxbuf_size = pnob->rx_buf_size;
> + }
> +
> + tasklet_init(&(adapter->sts_handler), be_process_intr,
> + (unsigned long)adapter);
> + adapter->tasklet_started = 1;
Do you need a tasklet if using NAPI? That would
> + spin_lock_init(&(adapter->int_lock));
> +
> + status = be_register_isr(adapter, pnob);
> + if (status != 0)
> + goto cleanup;
> +
> + adapter->isr_registered = 1;
> + adapter->rx_csum = 1;
> + adapter->max_rx_coal = BE_LRO_MAX_PKTS;
> +
> + memset(&get_fwv, 0,
> + sizeof(struct FWCMD_COMMON_GET_FW_VERSION_RESPONSE_PAYLOAD));
> + printk(KERN_INFO "BladeEngine Driver version:%s. "
> + "Copyright ServerEngines, Corporation 2005 - 2008\n",
> + be_drvr_ver);
> + status = be_function_get_fw_version(&pnob->fn_obj, &get_fwv, NULL,
> + NULL);
> + if (status == BE_SUCCESS) {
> + strncpy(be_fw_ver, get_fwv.firmware_version_string, 32);
> + printk(KERN_INFO "BladeEngine Firmware Version:%s\n",
> + get_fwv.firmware_version_string);
> + } else {
> + printk(KERN_WARNING "Unable to get BE Firmware Version\n");
> + }
> +
> + sema_init(&adapter->get_eth_stat_sem, 0);
> +
> + init_timer(&adapter->timer_ctxt.get_stats_timer);
> + atomic_set(&adapter->timer_ctxt.get_stat_flag, 0);
> + adapter->timer_ctxt.get_stats_timer.function =
> + &be_get_stats_timer_handler;
> +
> + status = be_mcc_create(adapter);
> + if (status < 0)
> + goto cleanup;
> + status = be_mcc_init(adapter);
> + if (status < 0)
> + goto cleanup;
> +
> + be_update_link_status(adapter);
> +
> + /* Register async call back function to handle link status updates */
> + status = be_mcc_add_async_event_callback(&adapter->net_obj->mcc_q_obj,
> + be_link_status_async_callback,
> + (void *)adapter);
> + if (status != BE_SUCCESS) {
> + printk(KERN_WARNING "add_async_event_callback failed");
> + printk(KERN_WARNING
> + "Link status changes may not be reflected\n");
> + }
> +
> + bni_enable_intr(adapter->net_obj);
> + bni_enable_eq_intr(adapter->net_obj);
> + adapter->dev_state = BE_DEV_STATE_INIT;
> + return 0;
> +
> +cleanup:
> + be_remove(pdev);
> +
> +error:
> + printk(KERN_ERR "BladeEngine initalization failed\n");
> + return status;
> +}
> +
> +/*
> + * Get the current link status and print the status on console
> + */
> +void be_update_link_status(struct be_adapter *adapter)
> +{
> + int status;
> + struct bni_net_object *pnob = adapter->net_obj;
> +
> + status = bni_get_link_sts(pnob, adapter->be_link_sts, NULL, NULL);
> +
> + if (status == BE_SUCCESS) {
> + if (adapter->be_link_sts->mac0_speed &&
> + adapter->be_link_sts->mac0_duplex)
> + adapter->port0_link_sts = BE_PORT_LINK_UP;
> + else
> + adapter->port0_link_sts = BE_PORT_LINK_DOWN;
> +
> + if (adapter->be_link_sts->mac1_speed &&
> + adapter->be_link_sts->mac1_duplex)
> + adapter->port1_link_sts = BE_PORT_LINK_UP;
> + else
> + adapter->port1_link_sts = BE_PORT_LINK_DOWN;
> +
> + printk(KERN_INFO "Link Properties for %s:\n",
> + OSM_NOB(pnob)->netdev->name);
> + be_print_link_info(adapter->be_link_sts);
> + return;
> + }
> + printk(KERN_WARNING "Could not get link status for %s\n",
> + OSM_NOB(pnob)->netdev->name);
> + return;
> +}
> +
> +#ifdef CONFIG_PM
> +static void
> +be_pm_cleanup(struct be_adapter *adapter,
> + struct bni_net_object *pnob, struct net_device *netdev)
> +{
> + u32 i;
> +
> + netif_carrier_off(netdev);
> + netif_stop_queue(netdev);
> +
> + be_wait_nic_tx_cmplx_cmpl(pnob);
> + bni_disable_eq_intr(pnob);
> +
> + if (adapter->tasklet_started) {
> + tasklet_kill(&adapter->sts_handler);
> + adapter->tasklet_started = 0;
> + }
> +
> + if (adapter->msix_enabled) {
> + if (adapter->isr_registered) {
> + free_irq(adapter->msix_entries[0].vector, netdev);
> + adapter->tasklet_started = 0;
> + adapter->isr_registered = 0;
> + }
> + }
> +
> + if (adapter->isr_registered) {
> + /* This is an INTX Interrupt */
> + free_irq(netdev->irq, netdev);
> + adapter->isr_registered = 0;
> + }
> +
> + bni_disable_intr(pnob);
> + bni_destroy_netobj(pnob, &adapter->sa_device);
> +
> + if (pnob->rx_ctxt) {
> + struct be_rx_page_info *rx_page_info;
> +
> + /*
> + * go through RX context array and free
> + * data buffs
> + */
> + for (i = 0; i < pnob->rx_q_len; i++) {
> + rx_page_info = &(OSM_NOB(pnob)->rx_page_info[i]);
> + if ((OSM_NOB(pnob)->rx_pg_shared == FALSE) ||
> + (rx_page_info->page_offset))
> + pci_unmap_page(adapter->pdev,
> + pci_unmap_addr(rx_page_info,
> + bus),
> + pnob->rx_buf_size,
> + PCI_DMA_FROMDEVICE);
> + if (rx_page_info->page)
> + put_page(rx_page_info->page);
> + memset(rx_page_info, 0, sizeof(struct be_rx_page_info));
> + }
> + OSM_NOB(pnob)->rx_pg_info_hd = 0;
> + }
> +
> +}
> +static int be_suspend(struct pci_dev *pdev, pm_message_t state)
> +{
> + struct be_adapter *adapter = pci_get_drvdata(pdev);
> + struct net_device *netdev = adapter->netdevp;
> + struct bni_net_object *pnob = (struct bni_net_object *)netdev->priv;
> +
> + adapter->dev_pm_state = adapter->dev_state;
> + adapter->dev_state = BE_DEV_STATE_SUSPEND;
> +
> + netif_device_detach(netdev);
> + netif_device_detach(netdev);
> + if (netif_running(netdev))
> + be_pm_cleanup(adapter, pnob, netdev);
> +
> + pci_enable_wake(pdev, 3, 1);
> + pci_enable_wake(pdev, 4, 1); /* D3 Cold = 4 */
> + pci_save_state(pdev);
> + pci_disable_device(pdev);
> + pci_set_power_state(pdev, pci_choose_state(pdev, state));
> + return 0;
> +}
> +
> +static void be_up(struct be_adapter *adapter)
> +{
> + struct bni_net_object *pnob = adapter->net_obj;
> +
> + if (OSM_NOB(pnob)->num_vlans != 0)
> + bni_config_vlan(pnob, OSM_NOB(pnob)->vlan_tag,
> + OSM_NOB(pnob)->num_vlans, NULL, NULL, 0);
> +
> +}
> +static int be_resume(struct pci_dev *pdev)
> +{
> + int status = 0;
> + struct be_adapter *adapter = pci_get_drvdata(pdev);
> + struct net_device *netdev = adapter->netdevp;
> + struct bni_net_object *pnob = (struct bni_net_object *)netdev->priv;
> +
> + netif_device_detach(netdev);
> +
> + status = pci_enable_device(pdev);
> + if (status)
> + return status;
> +
> + pci_set_power_state(pdev, 0);
> + pci_restore_state(pdev);
> + pci_enable_wake(pdev, 3, 0);
> + pci_enable_wake(pdev, 4, 0); /* 4 is D3 cold */
> +
> + netif_carrier_on(netdev);
> + netif_start_queue(netdev);
> +
> + if (netif_running(netdev)) {
> + status = be_nob_ring_init(adapter, pnob);
> + if (status < 0)
> + return (status);
> +
> + bni_set_uc_mac_adr(pnob, 0, 0, 0,
> + (u8 *)netdev->dev_addr, NULL, NULL);
> +
> + tasklet_init(&(adapter->sts_handler), be_process_intr,
> + (unsigned long)adapter);
> + adapter->tasklet_started = 1; /* indication to cleanup */
> +
> + if (be_register_isr(adapter, pnob) != 0) {
> + printk(KERN_ERR "be_register_isr failed\n");
> + return (status);
> + }
> +
> + adapter->isr_registered = 1;
> +
> + status = be_mcc_init(adapter);
> + if (status < 0) {
> + printk(KERN_ERR "be_mcc_init failed\n");
> + return (status);
> + }
> + be_update_link_status(adapter);
> + /*
> + * Register async call back function to handle link
> + * status updates
> + */
> + status = be_mcc_add_async_event_callback(
> + &adapter->net_obj->mcc_q_obj,
> + be_link_status_async_callback, (void *)adapter);
> + if (status != BE_SUCCESS) {
> + printk(KERN_WARNING "add_async_event_callback failed");
> + printk(KERN_WARNING
> + "Link status changes may not be reflected\n");
> + }
> + bni_enable_intr(pnob);
> + bni_enable_eq_intr(pnob);
> + be_up(adapter);
> + }
> + netif_device_attach(netdev);
> + adapter->dev_state = adapter->dev_pm_state;
> + return 0;
> +
> +}
> +
> +#endif
> +
> +/* Wait until no more pending transmits */
> +void be_wait_nic_tx_cmplx_cmpl(struct bni_net_object *pnob)
> +{
> + int i;
> +
> + /* Wait for 20us * 50000 (= 1s) and no more */
> + i = 0;
> + while ((pnob->tx_q_tl != pnob->tx_q_hd) && (i < 50000)) {
> + ++i;
> + udelay(20);
> + }
> +
> + /* Check for no more pending transmits */
> + if (i >= 50000) {
> + printk(KERN_WARNING
> + "Did not receive completions for all TX requests\n");
> + }
> +}
> +
> +static struct pci_driver be_driver = {
> + .name = be_driver_name,
> + .id_table = be_device_id_table,
> + .probe = be_probe,
> +#ifdef CONFIG_PM
> + .suspend = be_suspend,
> + .resume = be_resume,
> +#endif
> + .remove = be_remove
> +};
> +
> +/*
> + * Module init entry point. Registers our our device and return.
> + * Our probe will be called if the device is found.
> + */
> +
> +static int __init be_init_module(void)
> +{
> + int ret;
> +
> + if ((rxbuf_size != 8192) && (rxbuf_size != 4096)
> + && (rxbuf_size != 2048)) {
> + printk(KERN_WARNING
> + "Unsupported receive buffer size (%d) requested\n",
> + rxbuf_size);
> + printk(KERN_WARNING
> + "Must be 2048, 4096 or 8192. Defaulting to 2048\n");
> + rxbuf_size = 2048;
> + }
> +
> + ret = pci_register_driver(&be_driver);
> +
> + return ret;
> +}
> +
> +module_init(be_init_module);
> +
> +/*
> + * be_exit_module - Driver Exit Cleanup Routine
> + */
> +static void __exit be_exit_module(void)
> +{
> + pci_unregister_driver(&be_driver);
> +}
> +
> +module_exit(be_exit_module);
> diff --git a/drivers/net/benet/benet.h b/drivers/net/benet/benet.h
> new file mode 100644
> index 0000000..2174dfa
> --- /dev/null
> +++ b/drivers/net/benet/benet.h
> @@ -0,0 +1,301 @@
> +/*
> + * Copyright (C) 2005 - 2008 ServerEngines
> + * All rights reserved.
> + *
> + * This program is free software; you can redistribute it and/or
> + * modify it under the terms of the GNU General Public License version 2
> + * as published by the Free Software Foundation. The full GNU General
> + * Public License is included in this distribution in the file called COPYING.
> + *
> + * Contact Information:
> + * linux-drivers@...verengines.com
> + *
> + * ServerEngines
> + * 209 N. Fair Oaks Ave
> + * Sunnyvale, CA 94085
> + */
> +#ifndef _BENET_H_
> +#define _BENET_H_
> +
> +#include <linux/netdevice.h>
> +#include <linux/inet_lro.h>
> +#include "bni.h"
> +
> +#define BE_MAX_MTU 8974
> +
> +#define BE_MAX_LRO_DESCRIPTORS 8
> +#define BE_LRO_MAX_PKTS 64
> +#define BE_MAX_FRAGS_PER_FRAME 6
> +
> +extern const char be_drvr_ver[];
> +extern char be_fw_ver[];
> +extern char be_driver_name[];
> +
> +extern struct ethtool_ops be_ethtool_ops;
> +
> +
> +#define BE_DEV_STATE_NONE 0
> +#define BE_DEV_STATE_INIT 1
> +#define BE_DEV_STATE_OPEN 2
> +#define BE_DEV_STATE_SUSPEND 3
> +
> +/*
> + * BE driver statistics.
> + */
> +struct be_drvr_stat {
> + u32 bes_tx_reqs; /* number of TX requests initiated */
> + u32 bes_tx_fails; /* number of TX requests that failed */
> + u32 bes_fwd_reqs; /* number of send reqs through forwarding i/f */
> + u32 bes_tx_wrbs; /* number of tx WRBs used */
> +
> + u32 bes_ints; /* number of interrupts */
> + u32 bes_polls; /* number of times NAPI called poll function */
> + u32 bes_events; /* total evet entries processed */
> + u32 bes_tx_events; /* number of tx completion events */
> + u32 bes_ucrx_events; /* number of ucast rx completion events */
> + u32 bes_bcrx_events; /* number of bcast rx completion events */
> + u32 bes_tx_compl; /* number of tx completion entries processed */
> + u32 bes_ucrx_compl; /* number of ucrx completion entries
> + processed */
> + u32 bes_bcrx_compl; /* number of bcrx completion entries
> + processed */
> + u32 bes_ethrx_post_fail; /* number of ethrx buffer alloc
> + failures */
> + /*
> + * number of non ether type II frames dropped where
> + * frame len > length field of Mac Hdr
> + */
> + u32 bes_802_3_dropped_frames;
> + /*
> + * number of non ether type II frames malformed where
> + * in frame len < length field of Mac Hdr
> + */
> + u32 bes_802_3_malformed_frames;
> + u32 bes_ips; /* interrupts / sec */
> + u32 bes_prev_ints; /* bes_ints at last IPS calculation */
> + u16 bes_eth_tx_rate; /* ETH TX rate - Mb/sec */
> + u16 bes_eth_rx_rate; /* ETH RX rate - Mb/sec */
> + u32 bes_rx_coal; /* Num pkts coalasced */
> + u32 bes_rx_flush; /* Num times coalasced */
> + u32 bes_link_change_physical; /*Num of times physical link changed */
> + u32 bes_link_change_virtual; /*Num of times virtual link changed */
> + u32 bes_rx_misc_pkts; /* Misc pkts received */
> +};
> +
> +/* Maximum interrupt delay (in microseconds) allowed */
> +#define MAX_EQD 120
> +
> +/*
> + * timer to prevent system shutdown hang for ever if h/w stops responding
> + */
> +struct be_timer_ctxt {
> + atomic_t get_stat_flag;
> + struct timer_list get_stats_timer;
> + unsigned long get_stat_sem_addr;
> +} ;
> +
> +/* This structure is the main BladeEngine driver context. */
> +struct be_adapter {
> + struct net_device *netdevp;
> + struct be_drvr_stat be_stat;
> + struct net_device_stats benet_stats;
> + u32 num_bars;
> + struct sa_dev_bar_locations pci_bars[3]; /* PCI BAR details */
> + struct sa_dev sa_device; /* device object owned by beclib */
> + struct be_chip_object chip_object; /* BEClib chip object */
> +
> + struct tasklet_struct sts_handler;
> + struct timer_list cq_timer;
> + spinlock_t int_lock;
> +
> + struct FWCMD_ETH_GET_STATISTICS *eth_statsp;
> + /*
> + * This will enable the use of ethtool to enable or disable
> + * Checksum on Rx pkts to be obeyed or disobeyed.
> + * If this is TRUE = 1, then whatever is the checksum on the
> + * Received pkt as per BE, it will be given to the stack.
> + * Else the stack will re calculate it.
> + */
> + bool rx_csum;
> + /*
> + * This will enable the use of ethtool to enable or disable
> + * Coalese on Rx pkts to be obeyed or disobeyed.
> + * If this is grater than 0 and less than 16 then coalascing
> + * is enabled else it is disabled
> + */
> + u32 max_rx_coal;
> + struct pci_dev *pdev; /* Pointer to OS's PCI dvice */
> +
> + spinlock_t txq_lock;
> +
> + u32 isr; /* copy of Intr status reg. */
> +
> + u32 port0_link_sts; /* Port 0 link status */
> + u32 port1_link_sts; /* port 1 list status */
> + struct BE_LINK_STATUS *be_link_sts;
> +
> + /* pointer to the first netobject of this adapter */
> + struct bni_net_object *net_obj;
> +
> + /* Flags to indicate what to clean up */
> + bool tasklet_started;
> + bool isr_registered;
> + /*
> + * adaptive interrupt coalescing (AIC) related
> + */
> + bool enable_aic; /* 1 if AIC is enabled */
> + u16 min_eqd; /* minimum EQ delay in usec */
> + u16 max_eqd; /* minimum EQ delay in usec */
> + u16 cur_eqd; /* current EQ delay in usec */
> + /*
> + * book keeping for interrupt / sec and TX/RX rate calculation
> + */
> + ulong ips_jiffies; /* jiffies at last IPS calc */
> + u32 eth_tx_bytes;
> + ulong eth_tx_jiffies;
> + u32 eth_rx_bytes;
> + ulong eth_rx_jiffies;
> +
> + struct semaphore get_eth_stat_sem;
> +
> + /* timer ctxt to prevent shutdown hanging due to un-responsive BE */
> + struct be_timer_ctxt timer_ctxt;
> +
> +#define BE_MAX_MSIX_VECTORS 32
> +#define BE_MAX_REQ_MSIX_VECTORS 1 /* only one EQ in Linux driver */
> + struct msix_entry msix_entries[BE_MAX_MSIX_VECTORS];
> + bool msix_enabled;
> + bool dma_64bit_cap; /* the Device DAC capable or not */
> + u8 dev_state; /* The current state of the device */
> + u8 dev_pm_state; /* The State of device before going to suspend */
> +};
> +
> +
> +struct be_rx_page_info {
> + struct page *page;
> + dma_addr_t bus;
> + u16 page_offset;
> +} ;
> +
> +/*
> + * linux_net_object is an extension to BNI's NetObject structure.
> + * NetObject has a pointer to this structure
> + */
> +struct linux_net_object {
> + struct net_device *netdev;
> + struct bni_recv_buffer eth_rx_bufs[256]; /* to pass Rx buffer
> + addresses */
> + struct be_adapter *adapter; /* Pointer to OSM adapter */
> + u32 devno; /* OSM, network dev no. */
> + u32 use_port; /* Current active port */
> + struct be_rx_page_info *rx_page_info; /* Array of Rx buf pages */
> + u32 rx_pg_info_hd; /* Head of queue */
> + int rxbuf_post_fail; /* RxBuff posting fail count */
> + bool rx_pg_shared; /* Is an allocsted page shared as two frags ? */
> + struct vlan_group *vlan_grp;
> + u32 num_vlans; /* Number of vlans in BE's filter */
> + u16 vlan_tag[BE_NUM_VLAN_SUPPORTED]; /* vlans currently configured */
> +#ifdef CONFIG_BENET_NAPI
> + struct napi_struct napi;
> + u32 work_quota; /* Max RX packets to process */
> + bool rx_sched;
> + spinlock_t rx_lock;
> +#endif
> + struct net_lro_mgr lro_mgr;
> + struct net_lro_desc lro_desc[BE_MAX_LRO_DESCRIPTORS];
> +} ;
> +
> +/* functions to update RX/TX rates */
> +static inline void
> +update_rx_rate(struct be_adapter *adapter)
> +{
> + /* update the rate once in two seconds */
> + if ((jiffies - adapter->eth_rx_jiffies) > 2*(HZ)) {
> + u32 r;
> + r = adapter->eth_rx_bytes /
> + ((jiffies-adapter->eth_rx_jiffies)/(HZ));
> + r = (r / 1000000); /* MB/Sec */
> + adapter->be_stat.bes_eth_rx_rate = (r * 8); /* Mega Bits/Sec */
> + adapter->eth_rx_jiffies = jiffies;
> + adapter->eth_rx_bytes = 0;
> + }
> +}
> +
> +static inline void
> +update_tx_rate(struct be_adapter *adapter)
> +{
> + /* update the rate once in two seconds */
> + if ((jiffies - adapter->eth_tx_jiffies) > 2*(HZ)) {
> + u32 r;
> + r = adapter->eth_tx_bytes /
> + ((jiffies-adapter->eth_tx_jiffies)/(HZ));
> + r = (r / 1000000); /* MB/Sec */
> + adapter->be_stat.bes_eth_tx_rate = (r * 8); /* Mega Bits/Sec */
> + adapter->eth_tx_jiffies = jiffies;
> + adapter->eth_tx_bytes = 0;
> + }
> +}
> +/*
> + * Every second we look at the ints/sec and adjust eq_delay
> + * between adapter->min_eqd and adapter->max_eqd to keep the ints/sec between
> + * IPS_HI_WM and IPS_LO_WM.
> + */
> +#define IPS_HI_WM 18000
> +#define IPS_LO_WM 8000
> +
> +static inline void
> +update_eqd(struct be_adapter *adapter, struct bni_net_object *pnob)
> +{
> + /* update once a second */
> + if ((jiffies - adapter->ips_jiffies) > 1*(HZ)) {
> + /* One second elapsed since last update */
> + u32 r, new_eqd = -1;
> + r = adapter->be_stat.bes_ints -
> + adapter->be_stat.bes_prev_ints;
> + r = r / ((jiffies - adapter->ips_jiffies)/(HZ));
> + adapter->be_stat.bes_ips = r;
> + adapter->ips_jiffies = jiffies;
> + adapter->be_stat.bes_prev_ints = adapter->be_stat.bes_ints;
> + if (r > IPS_HI_WM && adapter->cur_eqd < adapter->max_eqd) {
> + /* increase eqdelay by a notch */
> + new_eqd = (adapter->cur_eqd + 8);
> + }
> + if (r < IPS_LO_WM && adapter->cur_eqd > adapter->min_eqd) {
> + /* decrease eqdelay by a notch */
> + new_eqd = (adapter->cur_eqd - 8);
> + }
> + if (adapter->enable_aic && new_eqd != -1) {
> + /* program new delay */
> + if (bni_change_eqd(pnob, new_eqd) == BE_SUCCESS)
> + adapter->cur_eqd = new_eqd;
> + }
> + }
> +}
> +/* convenience macro to access members in Linux extension of NetObject */
> +#define OSM_NOB(x) ((struct linux_net_object *)((x)->osm_netobj))
> +
> +/* proto declarations */
> +
> +int benet_probe(struct net_device *);
> +int be_ethtool_ioctl(struct net_device *, struct ifreq *);
> +struct net_device_stats *benet_get_stats(struct net_device *);
> +
> +void be_process_intr(unsigned long context);
> +irqreturn_t be_int(int irq, void *dev);
> +
> +void be_post_eth_rx_buffs(struct bni_net_object *);
> +void be_get_stat_cb(void *, BESTATUS, struct MCC_WRB_AMAP *);
> +
> +void be_get_stats_timer_handler(unsigned long);
> +
> +void be_wait_nic_tx_cmplx_cmpl(struct bni_net_object *);
> +void be_print_link_info(struct BE_LINK_STATUS *);
> +void be_update_link_status(struct be_adapter *);
> +
> +void be_init_procfs(struct be_adapter *);
> +void be_cleanup_procfs(struct be_adapter *);
> +
> +#ifdef CONFIG_BENET_NAPI
> +int be_poll(struct napi_struct *, int);
> +#endif
> +#endif /* _BENET_H_ */
> diff --git a/drivers/net/benet/bni.h b/drivers/net/benet/bni.h
> new file mode 100644
> index 0000000..fc9cd86
> --- /dev/null
> +++ b/drivers/net/benet/bni.h
> @@ -0,0 +1,327 @@
> +/*
> + * Copyright (C) 2005 - 2008 ServerEngines
> + * All rights reserved.
> + *
> + * This program is free software; you can redistribute it and/or
> + * modify it under the terms of the GNU General Public License version 2
> + * as published by the Free Software Foundation. The full GNU General
> + * Public License is included in this distribution in the file called COPYING.
> + *
> + * Contact Information:
> + * linux-drivers@...verengines.com
> + *
> + * ServerEngines
> + * 209 N. Fair Oaks Ave
> + * Sunnyvale, CA 94085
> + */
> +/*
> +
> +@...e
> + bni.h
> +
> +@...ef
> + Definitions and macros that are required for all .c files
> + that use the BNI API and implement the BNI API functions
> +*/
> +#ifndef _BNI_H
> +#define _BNI_H
> +
> +#define _SA_MODULE_NAME "net-driver"
> +#include "beclib_ll.h"
> +
> +#define VLAN_VALID_BIT 0x8000
> +#define BE_NUM_VLAN_SUPPORTED 32
> +#define BE_PORT_LINK_DOWN 0000
> +#define BE_PORT_LINK_UP 0001
> +
> +
> +/*
> +@...ef
> + This structure is used by the OSM driver to give BNI
> + physical fragments to use for DMAing data from NIC.
> +*/
> +struct bni_recv_buffer {
> + struct list_head rxb_list; /* for maintaining a linked list */
> + void *rxb_va; /* buffer virtual address */
> + u32 rxb_pa_lo; /* low part of physical address */
> + u32 rxb_pa_hi; /* high part of physical address */
> + u32 rxb_len; /* length of recv buffer */
> + void *rxb_ctxt; /* context for OSM driver to use */
> +};
> +
> +/*
> + * fragment list to describe scattered data.
> + */
> +struct bni_tx_frag_list {
> + u32 txb_len; /* Size of this fragment */
> + u32 txb_pa_lo; /* Lower 32 bits of 64 bit physical addr */
> + u32 txb_pa_hi; /* Higher 32 bits of 64 bit physical addr */
> +};
> +/*
> + * maximum fragements in a TX request
> + */
> +#define BE_MAX_TX_FRAG_COUNT (30)
> +
> +/*
> + * Flag bits for send operation
> + */
> +#define IPCS (1 << 0) /* Enable IP checksum offload */
> +#define UDPCS (1 << 1) /* Enable UDP checksum offload */
> +#define TCPCS (1 << 2) /* Enable TCP checksum offload */
> +#define LSO (1 << 3) /* Enable Large Segment offload */
> +#define ETHVLAN (1 << 4) /* Enable VLAN insert */
> +#define ETHEVENT (1 << 5) /* Generate event on completion */
> +#define ETHCOMPLETE (1 << 6) /* Generate completion when done */
> +#define IPSEC (1 << 7) /* Enable IPSEC */
> +#define FORWARD (1 << 8) /* Send the packet in forwarding path */
> +#define FIN (1 << 9) /* Issue FIN segment */
> +
> +/* @brief
> + * This structure is the main tracking structure for a NIC interface.
> + * This data structure contains OS agnostic data members for processing
> + * intialization, sends, receives, and asynchronous events from the
> + * BladeEngine network function. The OSM driver makes
> + * calls into functions defined at this layer for initialization,
> + * eumeration and population of physical fragments with per-packet
> + * control flags for send and receive operations, population of
> + * receive buffers for NIC , and handling asynchronous
> + * events (such as link status change, packet pattern recognition etc.).
> + */
> +struct bni_net_object {
> +
> + /*
> + * MCC Ring - used to send ioctl cmds to embedded ARM processor
> + */
> + struct MCC_WRB_AMAP *mcc_q; /* VA of the start of the ring */
> + u32 mcc_q_len; /* # of WRB entries in this ring */
> + u32 mcc_q_size;
> + u32 mcc_q_hd; /* MCC ring head */
> + u8 mcc_q_created; /* flag to help cleanup */
> + struct be_mcc_object mcc_q_obj; /* BECLIB's MCC ring Object */
> + dma_addr_t mcc_q_bus; /* DMA'ble bus address */
> + /*
> + * MCC Completion Ring - ARM's responses to ioctls sent from MCC ring
> + */
> + struct MCC_CQ_ENTRY_AMAP *mcc_cq; /* VA of the start of the ring */
> + u32 mcc_cq_len; /* # of compl. entries in this ring */
> + u32 mcc_cq_size;
> + u32 mcc_cq_tl; /* compl. ring tail */
> + u8 mcc_cq_created; /* flag to help cleanup */
> + struct be_cq_object mcc_cq_obj; /* BECLIB's MCC compl. ring object */
> + u32 mcc_cq_id; /* MCC ring ID */
> + dma_addr_t mcc_cq_bus; /* DMA'ble bus address */
> + /*
> + * BEClib uses an array of context objects to track outstanding
> + * requests to the MCC. We need allocate the same number of
> + * conext entries as the number of entries in the MCC WRB ring
> + */
> + u32 mcc_wrb_ctxt_size;
> + void *mcc_wrb_ctxt; /* pointer to the context area */
> + u32 mcc_wrb_ctxtLen; /* Number of entries in the context */
> + /*
> + * NIC send request ring - used for xmitting raw ether frames.
> + */
> + struct ETH_WRB_AMAP *tx_q; /* VA of the start of the ring */
> + u32 tx_q_len; /* # if entries in the send ring */
> + u32 tx_q_size;
> + u32 tx_q_hd; /* Head index. Next req. goes here */
> + u32 tx_q_tl; /* Tail indx. oldest outstanding req. */
> + u8 tx_q_created; /* flag to help cleanup */
> + struct be_ethsq_object tx_q_obj;/* BECLIB's send Q handle */
> + dma_addr_t tx_q_bus; /* DMA'ble bus address */
> + u32 tx_q_id; /* send queue ring ID */
> + u32 tx_q_port; /* 0 no binding, 1 port A, 2 port B */
> + atomic_t tx_q_used; /* # of WRBs used */
> + /* ptr to an array in which we store context info for each send req. */
> + void **tx_ctxt;
> + /*
> + * NIC Send compl. ring - completion status for all NIC frames xmitted.
> + */
> + struct ETH_TX_COMPL_AMAP *tx_cq;/* VA of start of the ring */
> + u32 txcq_len; /* # of entries in the ring */
> + u32 tx_cq_size;
> + /*
> + * index into compl ring where the host expects next completion entry
> + */
> + u32 tx_cq_tl;
> + u32 tx_cq_id; /* completion queue id */
> + u8 tx_cq_created; /* flag to help cleanup */
> + struct be_cq_object tx_cq_obj;
> + dma_addr_t tx_cq_bus; /* DMA'ble bus address */
> + /*
> + * Event Queue - all completion entries post events here.
> + */
> + struct EQ_ENTRY_AMAP *event_q; /* VA of start of event queue */
> + u32 event_q_len; /* # of entries */
> + u32 event_q_size;
> + u32 event_q_tl; /* Tail of the event queue */
> + u32 event_q_id; /* Event queue ID */
> + u8 event_q_created; /* flag to help cleanup */
> + struct be_eq_object event_q_obj; /* Queue handle */
> + dma_addr_t event_q_bus; /* DMA'ble bus address */
> + /*
> + * NIC receive queue - Data buffers to be used for receiving unicast,
> + * broadcast and multi-cast frames are posted here.
> + */
> + struct ETH_RX_D_AMAP *rx_q; /* VA of start of the queue */
> + u32 rx_q_len; /* # of entries */
> + u32 rx_q_size;
> + u32 rx_q_hd; /* Head of the queue */
> + atomic_t rx_q_posted; /* number of posted buffers */
> + u32 rx_q_id; /* queue ID */
> + u8 rx_q_created; /* flag to help cleanup */
> + struct be_ethrq_object rx_q_obj; /* NIC RX queue handle */
> + dma_addr_t rx_q_bus; /* DMA'ble bus address */
> + /*
> + * Pointer to an array of opaque context object for use by OSM driver
> + */
> + void **rx_ctxt;
> + /*
> + * NIC unicast RX completion queue - all unicast ether frame completion
> + * statuses from BE come here.
> + */
> + struct ETH_RX_COMPL_AMAP *ucrx_cq; /* VA of start of the queue */
> + u32 ucrx_cq_len; /* # of entries */
> + u32 ucrx_cq_size;
> + u32 ucrx_cq_tl; /* Tail of the queue */
> + u32 ucrx_cq_id; /* queue ID */
> + u8 ucrx_cq_created; /* flag to help cleanup */
> + struct be_cq_object ucrx_cq_obj; /* queue handle */
> + dma_addr_t ucrx_cq_bus; /* DMA'ble bus address */
> + /*
> + * Broadcast RX completion queue - all broadcast and multicast ether
> + * completion statues from BE come here.
> + */
> + struct ETH_RX_COMPL_AMAP *bcrx_cq; /* VA of start of queue */
> + u32 bcrx_cq_len; /* # of entries */
> + u32 bcrx_cq_size;
> + u32 bcrx_cq_tl; /* Tail of the queue */
> + u32 bcrx_cq_id; /* Queue ID */
> + u8 bcrx_cq_created; /* flag to help cleanup */
> + struct be_cq_object bcrx_cq_obj; /* queue handle */
> + dma_addr_t bcrx_cq_bus; /* DMA'ble bus address */
> +
> + struct be_function_object fn_obj; /* function object */
> + bool fn_obj_created;
> + u32 rx_buf_size; /* Size of the RX buffers */
> + /*
> + * OSM handle. OSM drivers can use this pointer to extend NetObject.
> + */
> + void *osm_netobj;
> + struct sa_sgl mb_sgl; /* SGL for MCC_MAIL_BOX */
> + void *mb_ptr; /* mailbox ptr to be freed */
> + dma_addr_t mb_bus; /* DMA'ble bus address */
> + u32 mb_size;
> +};
> +
> +/*
> + * convenience macros to access some NetObject members
> + */
> +#define NET_FH(np) (&(np)->fn_obj)
> +
> +static inline void index_advance(u32 *index, u32 limit)
> +{
> + BUG_ON(limit & (limit-1));
> + *index = (*index + 1) & (limit - 1);
> +}
> +
> +/*
> + * Functions to advance the head and tail in various rings.
> + */
> +static inline void bni_adv_eq_tl(struct bni_net_object *pnob)
> +{
> + index_advance(&pnob->event_q_tl, pnob->event_q_len);
> +}
> +
> +static inline void bni_adv_txq_hd(struct bni_net_object *pnob)
> +{
> + index_advance(&pnob->tx_q_hd, pnob->tx_q_len);
> +}
> +
> +static inline void bni_adv_txq_tl(struct bni_net_object *pnob)
> +{
> + index_advance(&pnob->tx_q_tl, pnob->tx_q_len);
> +}
> +
> +static inline void bni_adv_txcq_tl(struct bni_net_object *pnob)
> +{
> + index_advance(&pnob->tx_cq_tl, pnob->txcq_len);
> +}
> +
> +static inline void bni_adv_rxq_hd(struct bni_net_object *pnob)
> +{
> + index_advance(&pnob->rx_q_hd, pnob->rx_q_len);
> +}
> +
> +static inline void bni_adv_ucrxcq_tl(struct bni_net_object *pnob)
> +{
> + index_advance(&pnob->ucrx_cq_tl, pnob->ucrx_cq_len);
> +}
> +
> +static inline void bni_adv_bcrxcq_tl(struct bni_net_object *pnob)
> +{
> + index_advance(&pnob->bcrx_cq_tl, pnob->bcrx_cq_len);
> +}
> +
> +static inline BESTATUS bni_process_mcc_cmpl(struct be_mcc_object *pmccob)
> +{
> + return (be_mcc_process_cq(pmccob, 1));
> +}
> +
> +/* forward declarations of function prototypes */
> +BESTATUS bni_init(struct be_chip_object *);
> +BESTATUS bni_create_mcc_rings(struct bni_net_object *);
> +extern void bni_destroy_netobj(struct bni_net_object *, struct sa_dev *);
> +void bni_cleanup(struct be_chip_object *);
> +
> +BESTATUS bni_create_netobj(struct bni_net_object *,
> + struct sa_dev_bar_locations *, u32,
> + struct sa_dev *, struct be_chip_object *);
> +
> +BESTATUS bni_tx_pkt(struct bni_net_object *, struct bni_tx_frag_list *, u32,
> + u32, u32, void *, u32);
> +void bni_start_tx(struct bni_net_object *, u32);
> +
> +u32 bni_post_rx_buffs(struct bni_net_object *, struct list_head *);
> +BESTATUS bni_change_eqd(struct bni_net_object *, u32);
> +
> +struct ETH_TX_COMPL_AMAP *bni_get_tx_cmpl(struct bni_net_object *);
> +struct ETH_RX_COMPL_AMAP *bni_get_ucrx_cmpl(struct bni_net_object *);
> +struct ETH_RX_COMPL_AMAP *bni_get_bcrx_cmpl(struct bni_net_object *);
> +void bni_notify_cmpl(struct bni_net_object *, int, int, int);
> +
> +void bni_enable_intr(struct bni_net_object *);
> +void bni_enable_eq_intr(struct bni_net_object *);
> +void bni_disable_intr(struct bni_net_object *);
> +void bni_disable_eq_intr(struct bni_net_object *);
> +
> +u32 bni_get_isr(struct bni_net_object *);
> +
> +struct EQ_ENTRY_AMAP *bni_get_event(struct bni_net_object *);
> +void bni_notify_event(struct bni_net_object *, int, int);
> +
> +BESTATUS bni_get_uc_mac_adrr(struct bni_net_object *, u8, u8, u8,
> + u8 *, MCC_WRB_CQE_CALLBACK, void *);
> +
> +BESTATUS bni_set_uc_mac_adr(struct bni_net_object *, u8, u8, u8,
> + u8 *, MCC_WRB_CQE_CALLBACK, void *);
> +
> +BESTATUS bni_set_mc_filter(struct bni_net_object *, u32,
> + bool, u8 *, MCC_WRB_CQE_CALLBACK, void *);
> +
> +void bni_set_promisc(struct bni_net_object *);
> +void bni_reset_promisc(struct bni_net_object *);
> +BESTATUS bni_config_vlan(struct bni_net_object *, u16 *,
> + u32, MCC_WRB_CQE_CALLBACK, void *, bool);
> +
> +BESTATUS bni_get_stats(struct bni_net_object *,
> + struct FWCMD_ETH_GET_STATISTICS *,
> + u64, MCC_WRB_CQE_CALLBACK, void *);
> +
> +BESTATUS bni_get_link_sts(struct bni_net_object *, struct BE_LINK_STATUS *,
> + MCC_WRB_CQE_CALLBACK, void *);
> +BESTATUS bni_set_flow_ctll(struct be_function_object *, bool, bool);
> +BESTATUS bni_get_flow_ctl(struct be_function_object *pFnObj, bool *, bool *);
> +u32 bni_process_rx_flush_cmpl(struct bni_net_object *);
> +
> +#endif /* #ifndef _BNI_H_ */
> --
> 1.5.5
--
To unsubscribe from this list: send the line "unsubscribe netdev" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
___________________________________________________________________________________
This message, together with any attachment(s), contains confidential and proprietary information of
ServerEngines Corporation and is intended only for the designated recipient(s) named above. Any unauthorized
review, printing, retention, copying, disclosure or distribution is strictly prohibited. If you are not the
intended recipient of this message, please immediately advise the sender by reply email message and
delete all copies of this message and any attachment(s). Thank you.
--
To unsubscribe from this list: send the line "unsubscribe netdev" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Powered by blists - more mailing lists