diff options
Diffstat (limited to 'drivers/net/cxgb4vf/cxgb4vf_main.c')
| -rw-r--r-- | drivers/net/cxgb4vf/cxgb4vf_main.c | 2888 |
1 files changed, 2888 insertions, 0 deletions
diff --git a/drivers/net/cxgb4vf/cxgb4vf_main.c b/drivers/net/cxgb4vf/cxgb4vf_main.c new file mode 100644 index 000000000000..a16563219ac9 --- /dev/null +++ b/drivers/net/cxgb4vf/cxgb4vf_main.c @@ -0,0 +1,2888 @@ +/* + * This file is part of the Chelsio T4 PCI-E SR-IOV Virtual Function Ethernet + * driver for Linux. + * + * Copyright (c) 2009-2010 Chelsio Communications, Inc. All rights reserved. + * + * This software is available to you under a choice of one of two + * licenses. You may choose to be licensed under the terms of the GNU + * General Public License (GPL) Version 2, available from the file + * COPYING in the main directory of this source tree, or the + * OpenIB.org BSD license below: + * + * Redistribution and use in source and binary forms, with or + * without modification, are permitted provided that the following + * conditions are met: + * + * - Redistributions of source code must retain the above + * copyright notice, this list of conditions and the following + * disclaimer. + * + * - Redistributions in binary form must reproduce the above + * copyright notice, this list of conditions and the following + * disclaimer in the documentation and/or other materials + * provided with the distribution. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND + * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS + * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN + * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN + * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE + * SOFTWARE. + */ + +#include <linux/version.h> +#include <linux/module.h> +#include <linux/moduleparam.h> +#include <linux/init.h> +#include <linux/pci.h> +#include <linux/dma-mapping.h> +#include <linux/netdevice.h> +#include <linux/etherdevice.h> +#include <linux/debugfs.h> +#include <linux/ethtool.h> + +#include "t4vf_common.h" +#include "t4vf_defs.h" + +#include "../cxgb4/t4_regs.h" +#include "../cxgb4/t4_msg.h" + +/* + * Generic information about the driver. + */ +#define DRV_VERSION "1.0.0" +#define DRV_DESC "Chelsio T4 Virtual Function (VF) Network Driver" + +/* + * Module Parameters. + * ================== + */ + +/* + * Default ethtool "message level" for adapters. + */ +#define DFLT_MSG_ENABLE (NETIF_MSG_DRV | NETIF_MSG_PROBE | NETIF_MSG_LINK | \ + NETIF_MSG_TIMER | NETIF_MSG_IFDOWN | NETIF_MSG_IFUP |\ + NETIF_MSG_RX_ERR | NETIF_MSG_TX_ERR) + +static int dflt_msg_enable = DFLT_MSG_ENABLE; + +module_param(dflt_msg_enable, int, 0644); +MODULE_PARM_DESC(dflt_msg_enable, + "default adapter ethtool message level bitmap"); + +/* + * The driver uses the best interrupt scheme available on a platform in the + * order MSI-X then MSI. This parameter determines which of these schemes the + * driver may consider as follows: + * + * msi = 2: choose from among MSI-X and MSI + * msi = 1: only consider MSI interrupts + * + * Note that unlike the Physical Function driver, this Virtual Function driver + * does _not_ support legacy INTx interrupts (this limitation is mandated by + * the PCI-E SR-IOV standard). + */ +#define MSI_MSIX 2 +#define MSI_MSI 1 +#define MSI_DEFAULT MSI_MSIX + +static int msi = MSI_DEFAULT; + +module_param(msi, int, 0644); +MODULE_PARM_DESC(msi, "whether to use MSI-X or MSI"); + +/* + * Fundamental constants. + * ====================== + */ + +enum { + MAX_TXQ_ENTRIES = 16384, + MAX_RSPQ_ENTRIES = 16384, + MAX_RX_BUFFERS = 16384, + + MIN_TXQ_ENTRIES = 32, + MIN_RSPQ_ENTRIES = 128, + MIN_FL_ENTRIES = 16, + + /* + * For purposes of manipulating the Free List size we need to + * recognize that Free Lists are actually Egress Queues (the host + * produces free buffers which the hardware consumes), Egress Queues + * indices are all in units of Egress Context Units bytes, and free + * list entries are 64-bit PCI DMA addresses. And since the state of + * the Producer Index == the Consumer Index implies an EMPTY list, we + * always have at least one Egress Unit's worth of Free List entries + * unused. See sge.c for more details ... + */ + EQ_UNIT = SGE_EQ_IDXSIZE, + FL_PER_EQ_UNIT = EQ_UNIT / sizeof(__be64), + MIN_FL_RESID = FL_PER_EQ_UNIT, +}; + +/* + * Global driver state. + * ==================== + */ + +static struct dentry *cxgb4vf_debugfs_root; + +/* + * OS "Callback" functions. + * ======================== + */ + +/* + * The link status has changed on the indicated "port" (Virtual Interface). + */ +void t4vf_os_link_changed(struct adapter *adapter, int pidx, int link_ok) +{ + struct net_device *dev = adapter->port[pidx]; + + /* + * If the port is disabled or the current recorded "link up" + * status matches the new status, just return. + */ + if (!netif_running(dev) || link_ok == netif_carrier_ok(dev)) + return; + + /* + * Tell the OS that the link status has changed and print a short + * informative message on the console about the event. + */ + if (link_ok) { + const char *s; + const char *fc; + const struct port_info *pi = netdev_priv(dev); + + netif_carrier_on(dev); + + switch (pi->link_cfg.speed) { + case SPEED_10000: + s = "10Gbps"; + break; + + case SPEED_1000: + s = "1000Mbps"; + break; + + case SPEED_100: + s = "100Mbps"; + break; + + default: + s = "unknown"; + break; + } + + switch (pi->link_cfg.fc) { + case PAUSE_RX: + fc = "RX"; + break; + + case PAUSE_TX: + fc = "TX"; + break; + + case PAUSE_RX|PAUSE_TX: + fc = "RX/TX"; + break; + + default: + fc = "no"; + break; + } + + printk(KERN_INFO "%s: link up, %s, full-duplex, %s PAUSE\n", + dev->name, s, fc); + } else { + netif_carrier_off(dev); + printk(KERN_INFO "%s: link down\n", dev->name); + } +} + +/* + * Net device operations. + * ====================== + */ + +/* + * Record our new VLAN Group and enable/disable hardware VLAN Tag extraction + * based on whether the specified VLAN Group pointer is NULL or not. + */ +static void cxgb4vf_vlan_rx_register(struct net_device *dev, + struct vlan_group *grp) +{ + struct port_info *pi = netdev_priv(dev); + + pi->vlan_grp = grp; + t4vf_set_rxmode(pi->adapter, pi->viid, -1, -1, -1, -1, grp != NULL, 0); +} + +/* + * Perform the MAC and PHY actions needed to enable a "port" (Virtual + * Interface). + */ +static int link_start(struct net_device *dev) +{ + int ret; + struct port_info *pi = netdev_priv(dev); + + /* + * We do not set address filters and promiscuity here, the stack does + * that step explicitly. + */ + ret = t4vf_set_rxmode(pi->adapter, pi->viid, dev->mtu, -1, -1, -1, -1, + true); + if (ret == 0) { + ret = t4vf_change_mac(pi->adapter, pi->viid, + pi->xact_addr_filt, dev->dev_addr, true); + if (ret >= 0) { + pi->xact_addr_filt = ret; + ret = 0; + } + } + + /* + * We don't need to actually "start the link" itself since the + * firmware will do that for us when the first Virtual Interface + * is enabled on a port. + */ + if (ret == 0) + ret = t4vf_enable_vi(pi->adapter, pi->viid, true, true); + return ret; +} + +/* + * Name the MSI-X interrupts. + */ +static void name_msix_vecs(struct adapter *adapter) +{ + int namelen = sizeof(adapter->msix_info[0].desc) - 1; + int pidx; + + /* + * Firmware events. + */ + snprintf(adapter->msix_info[MSIX_FW].desc, namelen, + "%s-FWeventq", adapter->name); + adapter->msix_info[MSIX_FW].desc[namelen] = 0; + + /* + * Ethernet queues. + */ + for_each_port(adapter, pidx) { + struct net_device *dev = adapter->port[pidx]; + const struct port_info *pi = netdev_priv(dev); + int qs, msi; + + for (qs = 0, msi = MSIX_NIQFLINT; + qs < pi->nqsets; + qs++, msi++) { + snprintf(adapter->msix_info[msi].desc, namelen, + "%s-%d", dev->name, qs); + adapter->msix_info[msi].desc[namelen] = 0; + } + } +} + +/* + * Request all of our MSI-X resources. + */ +static int request_msix_queue_irqs(struct adapter *adapter) +{ + struct sge *s = &adapter->sge; + int rxq, msi, err; + + /* + * Firmware events. + */ + err = request_irq(adapter->msix_info[MSIX_FW].vec, t4vf_sge_intr_msix, + 0, adapter->msix_info[MSIX_FW].desc, &s->fw_evtq); + if (err) + return err; + + /* + * Ethernet queues. + */ + msi = MSIX_NIQFLINT; + for_each_ethrxq(s, rxq) { + err = request_irq(adapter->msix_info[msi].vec, + t4vf_sge_intr_msix, 0, + adapter->msix_info[msi].desc, + &s->ethrxq[rxq].rspq); + if (err) + goto err_free_irqs; + msi++; + } + return 0; + +err_free_irqs: + while (--rxq >= 0) + free_irq(adapter->msix_info[--msi].vec, &s->ethrxq[rxq].rspq); + free_irq(adapter->msix_info[MSIX_FW].vec, &s->fw_evtq); + return err; +} + +/* + * Free our MSI-X resources. + */ +static void free_msix_queue_irqs(struct adapter *adapter) +{ + struct sge *s = &adapter->sge; + int rxq, msi; + + free_irq(adapter->msix_info[MSIX_FW].vec, &s->fw_evtq); + msi = MSIX_NIQFLINT; + for_each_ethrxq(s, rxq) + free_irq(adapter->msix_info[msi++].vec, + &s->ethrxq[rxq].rspq); +} + +/* + * Turn on NAPI and start up interrupts on a response queue. + */ +static void qenable(struct sge_rspq *rspq) +{ + napi_enable(&rspq->napi); + + /* + * 0-increment the Going To Sleep register to start the timer and + * enable interrupts. + */ + t4_write_reg(rspq->adapter, T4VF_SGE_BASE_ADDR + SGE_VF_GTS, + CIDXINC(0) | + SEINTARM(rspq->intr_params) | + INGRESSQID(rspq->cntxt_id)); +} + +/* + * Enable NAPI scheduling and interrupt generation for all Receive Queues. + */ +static void enable_rx(struct adapter *adapter) +{ + int rxq; + struct sge *s = &adapter->sge; + + for_each_ethrxq(s, rxq) + qenable(&s->ethrxq[rxq].rspq); + qenable(&s->fw_evtq); + + /* + * The interrupt queue doesn't use NAPI so we do the 0-increment of + * its Going To Sleep register here to get it started. + */ + if (adapter->flags & USING_MSI) + t4_write_reg(adapter, T4VF_SGE_BASE_ADDR + SGE_VF_GTS, + CIDXINC(0) | + SEINTARM(s->intrq.intr_params) | + INGRESSQID(s->intrq.cntxt_id)); + +} + +/* + * Wait until all NAPI handlers are descheduled. + */ +static void quiesce_rx(struct adapter *adapter) +{ + struct sge *s = &adapter->sge; + int rxq; + + for_each_ethrxq(s, rxq) + napi_disable(&s->ethrxq[rxq].rspq.napi); + napi_disable(&s->fw_evtq.napi); +} + +/* + * Response queue handler for the firmware event queue. + */ +static int fwevtq_handler(struct sge_rspq *rspq, const __be64 *rsp, + const struct pkt_gl *gl) +{ + /* + * Extract response opcode and get pointer to CPL message body. + */ + struct adapter *adapter = rspq->adapter; + u8 opcode = ((const struct rss_header *)rsp)->opcode; + void *cpl = (void *)(rsp + 1); + + switch (opcode) { + case CPL_FW6_MSG: { + /* + * We've received an asynchronous message from the firmware. + */ + const struct cpl_fw6_msg *fw_msg = cpl; + if (fw_msg->type == FW6_TYPE_CMD_RPL) + t4vf_handle_fw_rpl(adapter, fw_msg->data); + break; + } + + case CPL_SGE_EGR_UPDATE: { + /* + * We've received an Egress Queue Status Update message. We + * get these, if the SGE is configured to send these when the + * firmware passes certain points in processing our TX + * Ethernet Queue or if we make an explicit request for one. + * We use these updates to determine when we may need to + * restart a TX Ethernet Queue which was stopped for lack of + * free TX Queue Descriptors ... + */ + const struct cpl_sge_egr_update *p = (void *)cpl; + unsigned int qid = EGR_QID(be32_to_cpu(p->opcode_qid)); + struct sge *s = &adapter->sge; + struct sge_txq *tq; + struct sge_eth_txq *txq; + unsigned int eq_idx; + + /* + * Perform sanity checking on the Queue ID to make sure it + * really refers to one of our TX Ethernet Egress Queues which + * is active and matches the queue's ID. None of these error + * conditions should ever happen so we may want to either make + * them fatal and/or conditionalized under DEBUG. + */ + eq_idx = EQ_IDX(s, qid); + if (unlikely(eq_idx >= MAX_EGRQ)) { + dev_err(adapter->pdev_dev, + "Egress Update QID %d out of range\n", qid); + break; + } + tq = s->egr_map[eq_idx]; + if (unlikely(tq == NULL)) { + dev_err(adapter->pdev_dev, + "Egress Update QID %d TXQ=NULL\n", qid); + break; + } + txq = container_of(tq, struct sge_eth_txq, q); + if (unlikely(tq->abs_id != qid)) { + dev_err(adapter->pdev_dev, + "Egress Update QID %d refers to TXQ %d\n", + qid, tq->abs_id); + break; + } + + /* + * Restart a stopped TX Queue which has less than half of its + * TX ring in use ... + */ + txq->q.restarts++; + netif_tx_wake_queue(txq->txq); + break; + } + + default: + dev_err(adapter->pdev_dev, + "unexpected CPL %#x on FW event queue\n", opcode); + } + + return 0; +} + +/* + * Allocate SGE TX/RX response queues. Determine how many sets of SGE queues + * to use and initializes them. We support multiple "Queue Sets" per port if + * we have MSI-X, otherwise just one queue set per port. + */ +static int setup_sge_queues(struct adapter *adapter) +{ + struct sge *s = &adapter->sge; + int err, pidx, msix; + + /* + * Clear "Queue Set" Free List Starving and TX Queue Mapping Error + * state. + */ + bitmap_zero(s->starving_fl, MAX_EGRQ); + + /* + * If we're using MSI interrupt mode we need to set up a "forwarded + * interrupt" queue which we'll set up with our MSI vector. The rest + * of the ingress queues will be set up to forward their interrupts to + * this queue ... This must be first since t4vf_sge_alloc_rxq() uses + * the intrq's queue ID as the interrupt forwarding queue for the + * subsequent calls ... + */ + if (adapter->flags & USING_MSI) { + err = t4vf_sge_alloc_rxq(adapter, &s->intrq, false, + adapter->port[0], 0, NULL, NULL); + if (err) + goto err_free_queues; + } + + /* + * Allocate our ingress queue for asynchronous firmware messages. + */ + err = t4vf_sge_alloc_rxq(adapter, &s->fw_evtq, true, adapter->port[0], + MSIX_FW, NULL, fwevtq_handler); + if (err) + goto err_free_queues; + + /* + * Allocate each "port"'s initial Queue Sets. These can be changed + * later on ... up to the point where any interface on the adapter is + * brought up at which point lots of things get nailed down + * permanently ... + */ + msix = MSIX_NIQFLINT; + for_each_port(adapter, pidx) { + struct net_device *dev = adapter->port[pidx]; + struct port_info *pi = netdev_priv(dev); + struct sge_eth_rxq *rxq = &s->ethrxq[pi->first_qset]; + struct sge_eth_txq *txq = &s->ethtxq[pi->first_qset]; + int qs; + + for (qs = 0; qs < pi->nqsets; qs++, rxq++, txq++) { + err = t4vf_sge_alloc_rxq(adapter, &rxq->rspq, false, + dev, msix++, + &rxq->fl, t4vf_ethrx_handler); + if (err) + goto err_free_queues; + + err = t4vf_sge_alloc_eth_txq(adapter, txq, dev, + netdev_get_tx_queue(dev, qs), + s->fw_evtq.cntxt_id); + if (err) + goto err_free_queues; + + rxq->rspq.idx = qs; + memset(&rxq->stats, 0, sizeof(rxq->stats)); + } + } + + /* + * Create the reverse mappings for the queues. + */ + s->egr_base = s->ethtxq[0].q.abs_id - s->ethtxq[0].q.cntxt_id; + s->ingr_base = s->ethrxq[0].rspq.abs_id - s->ethrxq[0].rspq.cntxt_id; + IQ_MAP(s, s->fw_evtq.abs_id) = &s->fw_evtq; + for_each_port(adapter, pidx) { + struct net_device *dev = adapter->port[pidx]; + struct port_info *pi = netdev_priv(dev); + struct sge_eth_rxq *rxq = &s->ethrxq[pi->first_qset]; + struct sge_eth_txq *txq = &s->ethtxq[pi->first_qset]; + int qs; + + for (qs = 0; qs < pi->nqsets; qs++, rxq++, txq++) { + IQ_MAP(s, rxq->rspq.abs_id) = &rxq->rspq; + EQ_MAP(s, txq->q.abs_id) = &txq->q; + + /* + * The FW_IQ_CMD doesn't return the Absolute Queue IDs + * for Free Lists but since all of the Egress Queues + * (including Free Lists) have Relative Queue IDs + * which are computed as Absolute - Base Queue ID, we + * can synthesize the Absolute Queue IDs for the Free + * Lists. This is useful for debugging purposes when + * we want to dump Queue Contexts via the PF Driver. + */ + rxq->fl.abs_id = rxq->fl.cntxt_id + s->egr_base; + EQ_MAP(s, rxq->fl.abs_id) = &rxq->fl; + } + } + return 0; + +err_free_queues: + t4vf_free_sge_resources(adapter); + return err; +} + +/* + * Set up Receive Side Scaling (RSS) to distribute packets to multiple receive + * queues. We configure the RSS CPU lookup table to distribute to the number + * of HW receive queues, and the response queue lookup table to narrow that + * down to the response queues actually configured for each "port" (Virtual + * Interface). We always configure the RSS mapping for all ports since the + * mapping table has plenty of entries. + */ +static int setup_rss(struct adapter *adapter) +{ + int pidx; + + for_each_port(adapter, pidx) { + struct port_info *pi = adap2pinfo(adapter, pidx); + struct sge_eth_rxq *rxq = &adapter->sge.ethrxq[pi->first_qset]; + u16 rss[MAX_PORT_QSETS]; + int qs, err; + + for (qs = 0; qs < pi->nqsets; qs++) + rss[qs] = rxq[qs].rspq.abs_id; + + err = t4vf_config_rss_range(adapter, pi->viid, + 0, pi->rss_size, rss, pi->nqsets); + if (err) + return err; + + /* + * Perform Global RSS Mode-specific initialization. + */ + switch (adapter->params.rss.mode) { + case FW_RSS_GLB_CONFIG_CMD_MODE_BASICVIRTUAL: + /* + * If Tunnel All Lookup isn't specified in the global + * RSS Configuration, then we need to specify a + * default Ingress Queue for any ingress packets which + * aren't hashed. We'll use our first ingress queue + * ... + */ + if (!adapter->params.rss.u.basicvirtual.tnlalllookup) { + union rss_vi_config config; + err = t4vf_read_rss_vi_config(adapter, + pi->viid, + &config); + if (err) + return err; + config.basicvirtual.defaultq = + rxq[0].rspq.abs_id; + err = t4vf_write_rss_vi_config(adapter, + pi->viid, + &config); + if (err) + return err; + } + break; + } + } + + return 0; +} + +/* + * Bring the adapter up. Called whenever we go from no "ports" open to having + * one open. This function performs the actions necessary to make an adapter + * operational, such as completing the initialization of HW modules, and + * enabling interrupts. Must be called with the rtnl lock held. (Note that + * this is called "cxgb_up" in the PF Driver.) + */ +static int adapter_up(struct adapter *adapter) +{ + int err; + + /* + * If this is the first time we've been called, perform basic + * adapter setup. Once we've done this, many of our adapter + * parameters can no longer be changed ... + */ + if ((adapter->flags & FULL_INIT_DONE) == 0) { + err = setup_sge_queues(adapter); + if (err) + return err; + err = setup_rss(adapter); + if (err) { + t4vf_free_sge_resources(adapter); + return err; + } + + if (adapter->flags & USING_MSIX) + name_msix_vecs(adapter); + adapter->flags |= FULL_INIT_DONE; + } + + /* + * Acquire our interrupt resources. We only support MSI-X and MSI. + */ + BUG_ON((adapter->flags & (USING_MSIX|USING_MSI)) == 0); + if (adapter->flags & USING_MSIX) + err = request_msix_queue_irqs(adapter); + else + err = request_irq(adapter->pdev->irq, + t4vf_intr_handler(adapter), 0, + adapter->name, adapter); + if (err) { + dev_err(adapter->pdev_dev, "request_irq failed, err %d\n", + err); + return err; + } + + /* + * Enable NAPI ingress processing and return success. + */ + enable_rx(adapter); + t4vf_sge_start(adapter); + return 0; +} + +/* + * Bring the adapter down. Called whenever the last "port" (Virtual + * Interface) closed. (Note that this routine is called "cxgb_down" in the PF + * Driver.) + */ +static void adapter_down(struct adapter *adapter) +{ + /* + * Free interrupt resources. + */ + if (adapter->flags & USING_MSIX) + free_msix_queue_irqs(adapter); + else + free_irq(adapter->pdev->irq, adapter); + + /* + * Wait for NAPI handlers to finish. + */ + quiesce_rx(adapter); +} + +/* + * Start up a net device. + */ +static int cxgb4vf_open(struct net_device *dev) +{ + int err; + struct port_info *pi = netdev_priv(dev); + struct adapter *adapter = pi->adapter; + + /* + * If this is the first interface that we're opening on the "adapter", + * bring the "adapter" up now. + */ + if (adapter->open_device_map == 0) { + err = adapter_up(adapter); + if (err) + return err; + } + + /* + * Note that this interface is up and start everything up ... + */ + dev->real_num_tx_queues = pi->nqsets; + set_bit(pi->port_id, &adapter->open_device_map); + link_start(dev); + netif_tx_start_all_queues(dev); + return 0; +} + +/* + * Shut down a net device. This routine is called "cxgb_close" in the PF + * Driver ... + */ +static int cxgb4vf_stop(struct net_device *dev) +{ + int ret; + struct port_info *pi = netdev_priv(dev); + struct adapter *adapter = pi->adapter; + + netif_tx_stop_all_queues(dev); + netif_carrier_off(dev); + ret = t4vf_enable_vi(adapter, pi->viid, false, false); + pi->link_cfg.link_ok = 0; + + clear_bit(pi->port_id, &adapter->open_device_map); + if (adapter->open_device_map == 0) + adapter_down(adapter); + return 0; +} + +/* + * Translate our basic statistics into the standard "ifconfig" statistics. + */ +static struct net_device_stats *cxgb4vf_get_stats(struct net_device *dev) +{ + struct t4vf_port_stats stats; + struct port_info *pi = netdev2pinfo(dev); + struct adapter *adapter = pi->adapter; + struct net_device_stats *ns = &dev->stats; + int err; + + spin_lock(&adapter->stats_lock); + err = t4vf_get_port_stats(adapter, pi->pidx, &stats); + spin_unlock(&adapter->stats_lock); + + memset(ns, 0, sizeof(*ns)); + if (err) + return ns; + + ns->tx_bytes = (stats.tx_bcast_bytes + stats.tx_mcast_bytes + + stats.tx_ucast_bytes + stats.tx_offload_bytes); + ns->tx_packets = (stats.tx_bcast_frames + stats.tx_mcast_frames + + stats.tx_ucast_frames + stats.tx_offload_frames); + ns->rx_bytes = (stats.rx_bcast_bytes + stats.rx_mcast_bytes + + stats.rx_ucast_bytes); + ns->rx_packets = (stats.rx_bcast_frames + stats.rx_mcast_frames + + stats.rx_ucast_frames); + ns->multicast = stats.rx_mcast_frames; + ns->tx_errors = stats.tx_drop_frames; + ns->rx_errors = stats.rx_err_frames; + + return ns; +} + +/* + * Collect up to maxaddrs worth of a netdevice's unicast addresses into an + * array of addrss pointers and return the number collected. + */ +static inline int collect_netdev_uc_list_addrs(const struct net_device *dev, + const u8 **addr, + unsigned int maxaddrs) +{ + unsigned int naddr = 0; + const struct netdev_hw_addr *ha; + + for_each_dev_addr(dev, ha) { + addr[naddr++] = ha->addr; + if (naddr >= maxaddrs) + break; + } + return naddr; +} + +/* + * Collect up to maxaddrs worth of a netdevice's multicast addresses into an + * array of addrss pointers and return the number collected. + */ +static inline int collect_netdev_mc_list_addrs(const struct net_device *dev, + const u8 **addr, + unsigned int maxaddrs) +{ + unsigned int naddr = 0; + const struct netdev_hw_addr *ha; + + netdev_for_each_mc_addr(ha, dev) { + addr[naddr++] = ha->addr; + if (naddr >= maxaddrs) + break; + } + return naddr; +} + +/* + * Configure the exact and hash address filters to handle a port's multicast + * and secondary unicast MAC addresses. + */ +static int set_addr_filters(const struct net_device *dev, bool sleep) +{ + u64 mhash = 0; + u64 uhash = 0; + bool free = true; + u16 filt_idx[7]; + const u8 *addr[7]; + int ret, naddr = 0; + const struct port_info *pi = netdev_priv(dev); + + /* first do the secondary unicast addresses */ + naddr = collect_netdev_uc_list_addrs(dev, addr, ARRAY_SIZE(addr)); + if (naddr > 0) { + ret = t4vf_alloc_mac_filt(pi->adapter, pi->viid, free, + naddr, addr, filt_idx, &uhash, sleep); + if (ret < 0) + return ret; + + free = false; + } + + /* next set up the multicast addresses */ + naddr = collect_netdev_mc_list_addrs(dev, addr, ARRAY_SIZE(addr)); + if (naddr > 0) { + ret = t4vf_alloc_mac_filt(pi->adapter, pi->viid, free, + naddr, addr, filt_idx, &mhash, sleep); + if (ret < 0) + return ret; + } + + return t4vf_set_addr_hash(pi->adapter, pi->viid, uhash != 0, + uhash | mhash, sleep); +} + +/* + * Set RX properties of a port, such as promiscruity, address filters, and MTU. + * If @mtu is -1 it is left unchanged. + */ +static int set_rxmode(struct net_device *dev, int mtu, bool sleep_ok) +{ + int ret; + struct port_info *pi = netdev_priv(dev); + + ret = set_addr_filters(dev, sleep_ok); + if (ret == 0) + ret = t4vf_set_rxmode(pi->adapter, pi->viid, -1, + (dev->flags & IFF_PROMISC) != 0, + (dev->flags & IFF_ALLMULTI) != 0, + 1, -1, sleep_ok); + return ret; +} + +/* + * Set the current receive modes on the device. + */ +static void cxgb4vf_set_rxmode(struct net_device *dev) +{ + /* unfortunately we can't return errors to the stack */ + set_rxmode(dev, -1, false); +} + +/* + * Find the entry in the interrupt holdoff timer value array which comes + * closest to the specified interrupt holdoff value. + */ +static int closest_timer(const struct sge *s, int us) +{ + int i, timer_idx = 0, min_delta = INT_MAX; + + for (i = 0; i < ARRAY_SIZE(s->timer_val); i++) { + int delta = us - s->timer_val[i]; + if (delta < 0) + delta = -delta; + if (delta < min_delta) { + min_delta = delta; + timer_idx = i; + } + } + return timer_idx; +} + +static int closest_thres(const struct sge *s, int thres) +{ + int i, delta, pktcnt_idx = 0, min_delta = INT_MAX; + + for (i = 0; i < ARRAY_SIZE(s->counter_val); i++) { + delta = thres - s->counter_val[i]; + if (delta < 0) + delta = -delta; + if (delta < min_delta) { + min_delta = delta; + pktcnt_idx = i; + } + } + return pktcnt_idx; +} + +/* + * Return a queue's interrupt hold-off time in us. 0 means no timer. + */ +static unsigned int qtimer_val(const struct adapter *adapter, + const struct sge_rspq *rspq) +{ + unsigned int timer_idx = QINTR_TIMER_IDX_GET(rspq->intr_params); + + return timer_idx < SGE_NTIMERS + ? adapter->sge.timer_val[timer_idx] + : 0; +} + +/** + * set_rxq_intr_params - set a queue's interrupt holdoff parameters + * @adapter: the adapter + * @rspq: the RX response queue + * @us: the hold-off time in us, or 0 to disable timer + * @cnt: the hold-off packet count, or 0 to disable counter + * + * Sets an RX response queue's interrupt hold-off time and packet count. + * At least one of the two needs to be enabled for the queue to generate + * interrupts. + */ +static int set_rxq_intr_params(struct adapter *adapter, struct sge_rspq *rspq, + unsigned int us, unsigned int cnt) +{ + unsigned int timer_idx; + + /* + * If both the interrupt holdoff timer and count are specified as + * zero, default to a holdoff count of 1 ... + */ + if ((us | cnt) == 0) + cnt = 1; + + /* + * If an interrupt holdoff count has been specified, then find the + * closest configured holdoff count and use that. If the response + * queue has already been created, then update its queue context + * parameters ... + */ + if (cnt) { + int err; + u32 v, pktcnt_idx; + + pktcnt_idx = closest_thres(&adapter->sge, cnt); + if (rspq->desc && rspq->pktcnt_idx != pktcnt_idx) { + v = FW_PARAMS_MNEM(FW_PARAMS_MNEM_DMAQ) | + FW_PARAMS_PARAM_X( + FW_PARAMS_PARAM_DMAQ_IQ_INTCNTTHRESH) | + FW_PARAMS_PARAM_YZ(rspq->cntxt_id); + err = t4vf_set_params(adapter, 1, &v, &pktcnt_idx); + if (err) + return err; + } + rspq->pktcnt_idx = pktcnt_idx; + } + + /* + * Compute the closest holdoff timer index from the supplied holdoff + * timer value. + */ + timer_idx = (us == 0 + ? SGE_TIMER_RSTRT_CNTR + : closest_timer(&adapter->sge, us)); + + /* + * Update the response queue's interrupt coalescing parameters and + * return success. + */ + rspq->intr_params = (QINTR_TIMER_IDX(timer_idx) | + (cnt > 0 ? QINTR_CNT_EN : 0)); + return 0; +} + +/* + * Return a version number to identify the type of adapter. The scheme is: + * - bits 0..9: chip version + * - bits 10..15: chip revision + */ +static inline unsigned int mk_adap_vers(const struct adapter *adapter) +{ + /* + * Chip version 4, revision 0x3f (cxgb4vf). + */ + return 4 | (0x3f << 10); +} + +/* + * Execute the specified ioctl command. + */ +static int cxgb4vf_do_ioctl(struct net_device *dev, struct ifreq *ifr, int cmd) +{ + int ret = 0; + + switch (cmd) { + /* + * The VF Driver doesn't have access to any of the other + * common Ethernet device ioctl()'s (like reading/writing + * PHY registers, etc. + */ + + default: + ret = -EOPNOTSUPP; + break; + } + return ret; +} + +/* + * Change the device's MTU. + */ +static int cxgb4vf_change_mtu(struct net_device *dev, int new_mtu) +{ + int ret; + struct port_info *pi = netdev_priv(dev); + + /* accommodate SACK */ + if (new_mtu < 81) + return -EINVAL; + + ret = t4vf_set_rxmode(pi->adapter, pi->viid, new_mtu, + -1, -1, -1, -1, true); + if (!ret) + dev->mtu = new_mtu; + return ret; +} + +/* + * Change the devices MAC address. + */ +static int cxgb4vf_set_mac_addr(struct net_device *dev, void *_addr) +{ + int ret; + struct sockaddr *addr = _addr; + struct port_info *pi = netdev_priv(dev); + + if (!is_valid_ether_addr(addr->sa_data)) + return -EINVAL; + + ret = t4vf_change_mac(pi->adapter, pi->viid, pi->xact_addr_filt, + addr->sa_data, true); + if (ret < 0) + return ret; + + memcpy(dev->dev_addr, addr->sa_data, dev->addr_len); + pi->xact_addr_filt = ret; + return 0; +} + +/* + * Return a TX Queue on which to send the specified skb. + */ +static u16 cxgb4vf_select_queue(struct net_device *dev, struct sk_buff *skb) +{ + /* + * XXX For now just use the default hash but we probably want to + * XXX look at other possibilities ... + */ + return skb_tx_hash(dev, skb); +} + +#ifdef CONFIG_NET_POLL_CONTROLLER +/* + * Poll all of our receive queues. This is called outside of normal interrupt + * context. + */ +static void cxgb4vf_poll_controller(struct net_device *dev) +{ + struct port_info *pi = netdev_priv(dev); + struct adapter *adapter = pi->adapter; + + if (adapter->flags & USING_MSIX) { + struct sge_eth_rxq *rxq; + int nqsets; + + rxq = &adapter->sge.ethrxq[pi->first_qset]; + for (nqsets = pi->nqsets; nqsets; nqsets--) { + t4vf_sge_intr_msix(0, &rxq->rspq); + rxq++; + } + } else + t4vf_intr_handler(adapter)(0, adapter); +} +#endif + +/* + * Ethtool operations. + * =================== + * + * Note that we don't support any ethtool operations which change the physical + * state of the port to which we're linked. + */ + +/* + * Return current port link settings. + */ +static int cxgb4vf_get_settings(struct net_device *dev, + struct ethtool_cmd *cmd) +{ + const struct port_info *pi = netdev_priv(dev); + + cmd->supported = pi->link_cfg.supported; + cmd->advertising = pi->link_cfg.advertising; + cmd->speed = netif_carrier_ok(dev) ? pi->link_cfg.speed : -1; + cmd->duplex = DUPLEX_FULL; + + cmd->port = (cmd->supported & SUPPORTED_TP) ? PORT_TP : PORT_FIBRE; + cmd->phy_address = pi->port_id; + cmd->transceiver = XCVR_EXTERNAL; + cmd->autoneg = pi->link_cfg.autoneg; + cmd->maxtxpkt = 0; + cmd->maxrxpkt = 0; + return 0; +} + +/* + * Return our driver information. + */ +static void cxgb4vf_get_drvinfo(struct net_device *dev, + struct ethtool_drvinfo *drvinfo) +{ + struct adapter *adapter = netdev2adap(dev); + + strcpy(drvinfo->driver, KBUILD_MODNAME); + strcpy(drvinfo->version, DRV_VERSION); + strcpy(drvinfo->bus_info, pci_name(to_pci_dev(dev->dev.parent))); + snprintf(drvinfo->fw_version, sizeof(drvinfo->fw_version), + "%u.%u.%u.%u, TP %u.%u.%u.%u", + FW_HDR_FW_VER_MAJOR_GET(adapter->params.dev.fwrev), + FW_HDR_FW_VER_MINOR_GET(adapter->params.dev.fwrev), + FW_HDR_FW_VER_MICRO_GET(adapter->params.dev.fwrev), + FW_HDR_FW_VER_BUILD_GET(adapter->params.dev.fwrev), + FW_HDR_FW_VER_MAJOR_GET(adapter->params.dev.tprev), + FW_HDR_FW_VER_MINOR_GET(adapter->params.dev.tprev), + FW_HDR_FW_VER_MICRO_GET(adapter->params.dev.tprev), + FW_HDR_FW_VER_BUILD_GET(adapter->params.dev.tprev)); +} + +/* + * Return current adapter message level. + */ +static u32 cxgb4vf_get_msglevel(struct net_device *dev) +{ + return netdev2adap(dev)->msg_enable; +} |