From owner-svn-src-all@FreeBSD.ORG Thu Feb 28 21:55:58 2013 Return-Path: Delivered-To: svn-src-all@freebsd.org Received: from mx1.freebsd.org (mx1.FreeBSD.org [8.8.178.115]) by hub.freebsd.org (Postfix) with ESMTP id 8270B7C3; Thu, 28 Feb 2013 21:55:58 +0000 (UTC) (envelope-from jfv@FreeBSD.org) Received: from svn.freebsd.org (svn.freebsd.org [IPv6:2001:1900:2254:2068::e6a:0]) by mx1.freebsd.org (Postfix) with ESMTP id 6EB9F402; Thu, 28 Feb 2013 21:55:58 +0000 (UTC) Received: from svn.freebsd.org ([127.0.1.70]) by svn.freebsd.org (8.14.5/8.14.5) with ESMTP id r1SLtwrF064119; Thu, 28 Feb 2013 21:55:58 GMT (envelope-from jfv@svn.freebsd.org) Received: (from jfv@localhost) by svn.freebsd.org (8.14.5/8.14.5/Submit) id r1SLtwBE064117; Thu, 28 Feb 2013 21:55:58 GMT (envelope-from jfv@svn.freebsd.org) Message-Id: <201302282155.r1SLtwBE064117@svn.freebsd.org> From: Jack F Vogel Date: Thu, 28 Feb 2013 21:55:58 +0000 (UTC) To: src-committers@freebsd.org, svn-src-all@freebsd.org, svn-src-stable@freebsd.org, svn-src-stable-8@freebsd.org Subject: svn commit: r247501 - stable/8/sys/dev/ixgbe X-SVN-Group: stable-8 MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit X-BeenThere: svn-src-all@freebsd.org X-Mailman-Version: 2.1.14 Precedence: list List-Id: "SVN commit messages for the entire src tree \(except for " user" and " projects" \)" List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , X-List-Received-Date: Thu, 28 Feb 2013 21:55:58 -0000 Author: jfv Date: Thu Feb 28 21:55:57 2013 New Revision: 247501 URL: http://svnweb.freebsd.org/changeset/base/247501 Log: MFC of the ixgbe driver including revisions: 217556,219902,228276,229767,229939,230572,231796,232238, 234137,234229,235547,235964,236627,236729,238149,239940, 240155,240366,240968,241037,241616,241646,241856,241885, 242403,242421,243714,243716,243718,243721,243724,243725, 243728,243729,243733,243735,243736,243741,243833,243857, 244514,245952,246128,246482,247035,247056 Modified: stable/8/sys/dev/ixgbe/ixgbe.c (contents, props changed) stable/8/sys/dev/ixgbe/ixgbe.h stable/8/sys/dev/ixgbe/ixgbe_82598.c stable/8/sys/dev/ixgbe/ixgbe_82598.h (contents, props changed) stable/8/sys/dev/ixgbe/ixgbe_82599.c stable/8/sys/dev/ixgbe/ixgbe_api.c stable/8/sys/dev/ixgbe/ixgbe_api.h stable/8/sys/dev/ixgbe/ixgbe_common.c stable/8/sys/dev/ixgbe/ixgbe_common.h stable/8/sys/dev/ixgbe/ixgbe_osdep.h stable/8/sys/dev/ixgbe/ixgbe_phy.c stable/8/sys/dev/ixgbe/ixgbe_type.h stable/8/sys/dev/ixgbe/ixgbe_vf.c stable/8/sys/dev/ixgbe/ixgbe_x540.c (contents, props changed) stable/8/sys/dev/ixgbe/ixv.c Directory Properties: stable/8/sys/dev/ixgbe/ (props changed) stable/8/sys/dev/ixgbe/ixgbe_82599.h (props changed) stable/8/sys/dev/ixgbe/ixgbe_x540.h (props changed) Modified: stable/8/sys/dev/ixgbe/ixgbe.c ============================================================================== --- stable/8/sys/dev/ixgbe/ixgbe.c Thu Feb 28 21:32:47 2013 (r247500) +++ stable/8/sys/dev/ixgbe/ixgbe.c Thu Feb 28 21:55:57 2013 (r247501) @@ -1,6 +1,6 @@ /****************************************************************************** - Copyright (c) 2001-2012, Intel Corporation + Copyright (c) 2001-2013, Intel Corporation All rights reserved. Redistribution and use in source and binary forms, with or without @@ -47,7 +47,7 @@ int ixgbe_display_debug_stat /********************************************************************* * Driver version *********************************************************************/ -char ixgbe_driver_version[] = "2.4.5"; +char ixgbe_driver_version[] = "2.5.0 - 8"; /********************************************************************* * PCI Device ID Table @@ -80,8 +80,10 @@ static ixgbe_vendor_info_t ixgbe_vendor_ {IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599_T3_LOM, 0, 0, 0}, {IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599_COMBO_BACKPLANE, 0, 0, 0}, {IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599_BACKPLANE_FCOE, 0, 0, 0}, + {IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599_SFP_SF2, 0, 0, 0}, {IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599_SFP_FCOE, 0, 0, 0}, {IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599EN_SFP, 0, 0, 0}, + {IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X540T1, 0, 0, 0}, {IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X540T, 0, 0, 0}, /* required last entry */ {0, 0, 0, 0, 0} @@ -102,16 +104,16 @@ static int ixgbe_probe(device_t); static int ixgbe_attach(device_t); static int ixgbe_detach(device_t); static int ixgbe_shutdown(device_t); -#if __FreeBSD_version >= 800000 +#ifdef IXGBE_LEGACY_TX +static void ixgbe_start(struct ifnet *); +static void ixgbe_start_locked(struct tx_ring *, struct ifnet *); +#else /* ! IXGBE_LEGACY_TX */ static int ixgbe_mq_start(struct ifnet *, struct mbuf *); static int ixgbe_mq_start_locked(struct ifnet *, struct tx_ring *, struct mbuf *); static void ixgbe_qflush(struct ifnet *); static void ixgbe_deferred_mq_start(void *, int); -#else -static void ixgbe_start(struct ifnet *); -static void ixgbe_start_locked(struct tx_ring *, struct ifnet *); -#endif +#endif /* IXGBE_LEGACY_TX */ static int ixgbe_ioctl(struct ifnet *, u_long, caddr_t); static void ixgbe_init(void *); static void ixgbe_init_locked(struct adapter *); @@ -148,7 +150,7 @@ static void ixgbe_enable_intr(struct static void ixgbe_disable_intr(struct adapter *); static void ixgbe_update_stats_counters(struct adapter *); static bool ixgbe_txeof(struct tx_ring *); -static bool ixgbe_rxeof(struct ix_queue *, int); +static bool ixgbe_rxeof(struct ix_queue *); static void ixgbe_rx_checksum(u32, struct mbuf *, u32); static void ixgbe_set_promisc(struct adapter *); static void ixgbe_set_multi(struct adapter *); @@ -161,10 +163,10 @@ static int ixgbe_set_thermal_test(SYSCTL static int ixgbe_dma_malloc(struct adapter *, bus_size_t, struct ixgbe_dma_alloc *, int); static void ixgbe_dma_free(struct adapter *, struct ixgbe_dma_alloc *); -static void ixgbe_add_rx_process_limit(struct adapter *, const char *, - const char *, int *, int); -static bool ixgbe_tx_ctx_setup(struct tx_ring *, struct mbuf *); -static bool ixgbe_tso_setup(struct tx_ring *, struct mbuf *, u32 *); +static int ixgbe_tx_ctx_setup(struct tx_ring *, + struct mbuf *, u32 *, u32 *); +static int ixgbe_tso_setup(struct tx_ring *, + struct mbuf *, u32 *, u32 *); static void ixgbe_set_ivar(struct adapter *, u8, u8, s8); static void ixgbe_configure_ivars(struct adapter *); static u8 * ixgbe_mc_array_itr(struct ixgbe_hw *, u8 **, u32 *); @@ -179,6 +181,9 @@ static __inline void ixgbe_rx_discard(st static __inline void ixgbe_rx_input(struct rx_ring *, struct ifnet *, struct mbuf *, u32); +static void ixgbe_enable_rx_drop(struct adapter *); +static void ixgbe_disable_rx_drop(struct adapter *); + /* Support for pluggable optic modules */ static bool ixgbe_sfp_probe(struct adapter *); static void ixgbe_setup_optics(struct adapter *); @@ -211,7 +216,8 @@ static device_method_t ixgbe_methods[] = DEVMETHOD(device_attach, ixgbe_attach), DEVMETHOD(device_detach, ixgbe_detach), DEVMETHOD(device_shutdown, ixgbe_shutdown), - {0, 0} + + DEVMETHOD_END }; static driver_t ixgbe_driver = { @@ -237,16 +243,16 @@ MODULE_DEPEND(ixgbe, ether, 1, 1, 1); static int ixgbe_enable_aim = TRUE; TUNABLE_INT("hw.ixgbe.enable_aim", &ixgbe_enable_aim); -static int ixgbe_max_interrupt_rate = (8000000 / IXGBE_LOW_LATENCY); +static int ixgbe_max_interrupt_rate = (4000000 / IXGBE_LOW_LATENCY); TUNABLE_INT("hw.ixgbe.max_interrupt_rate", &ixgbe_max_interrupt_rate); /* How many packets rxeof tries to clean at a time */ -static int ixgbe_rx_process_limit = 128; +static int ixgbe_rx_process_limit = 256; TUNABLE_INT("hw.ixgbe.rx_process_limit", &ixgbe_rx_process_limit); -/* Flow control setting, default to full */ -static int ixgbe_flow_control = ixgbe_fc_full; -TUNABLE_INT("hw.ixgbe.flow_control", &ixgbe_flow_control); +/* How many packets txeof tries to clean at a time */ +static int ixgbe_tx_process_limit = 256; +TUNABLE_INT("hw.ixgbe.tx_process_limit", &ixgbe_tx_process_limit); /* ** Smart speed setting, default to on @@ -265,15 +271,6 @@ static int ixgbe_enable_msix = 1; TUNABLE_INT("hw.ixgbe.enable_msix", &ixgbe_enable_msix); /* - * Header split: this causes the hardware to DMA - * the header into a separate mbuf from the payload, - * it can be a performance win in some workloads, but - * in others it actually hurts, its off by default. - */ -static int ixgbe_header_split = FALSE; -TUNABLE_INT("hw.ixgbe.hdr_split", &ixgbe_header_split); - -/* * Number of Queues, can be set to 0, * it then autoconfigures based on the * number of cpus with a max of 8. This @@ -294,6 +291,20 @@ TUNABLE_INT("hw.ixgbe.txd", &ixgbe_txd); static int ixgbe_rxd = PERFORM_RXD; TUNABLE_INT("hw.ixgbe.rxd", &ixgbe_rxd); +/* +** HW RSC control: +** this feature only works with +** IPv4, and only on 82599 and later. +** Also this will cause IP forwarding to +** fail and that can't be controlled by +** the stack as LRO can. For all these +** reasons I've deemed it best to leave +** this off and not bother with a tuneable +** interface, this would need to be compiled +** to enable. +*/ +static bool ixgbe_rsc_enable = FALSE; + /* Keep running tab on them for sanity check */ static int ixgbe_total_ports; @@ -318,6 +329,18 @@ static int atr_sample_rate = 20; static int fdir_pballoc = 1; #endif +#ifdef DEV_NETMAP +/* + * The #ifdef DEV_NETMAP / #endif blocks in this file are meant to + * be a reference on how to implement netmap support in a driver. + * Additional comments are in ixgbe_netmap.h . + * + * contains functions for netmap support + * that extend the standard driver. + */ +#include +#endif /* DEV_NETMAP */ + /********************************************************************* * Device identification routine * @@ -406,11 +429,6 @@ ixgbe_attach(device_t dev) OID_AUTO, "fc", CTLTYPE_INT | CTLFLAG_RW, adapter, 0, ixgbe_set_flowcntl, "I", "Flow Control"); - SYSCTL_ADD_PROC(device_get_sysctl_ctx(dev), - SYSCTL_CHILDREN(device_get_sysctl_tree(dev)), - OID_AUTO, "advertise_gig", CTLTYPE_INT | CTLFLAG_RW, - adapter, 0, ixgbe_set_advertise, "I", "1G Link"); - SYSCTL_ADD_INT(device_get_sysctl_ctx(dev), SYSCTL_CHILDREN(device_get_sysctl_tree(dev)), OID_AUTO, "enable_aim", CTLTYPE_INT|CTLFLAG_RW, @@ -516,28 +534,24 @@ ixgbe_attach(device_t dev) goto err_late; } - /* Get Hardware Flow Control setting */ - hw->fc.requested_mode = ixgbe_fc_full; - adapter->fc = hw->fc.requested_mode; - hw->fc.pause_time = IXGBE_FC_PAUSE; - hw->fc.low_water = IXGBE_FC_LO; - hw->fc.high_water[0] = IXGBE_FC_HI; - hw->fc.send_xon = TRUE; - error = ixgbe_init_hw(hw); - if (error == IXGBE_ERR_EEPROM_VERSION) { + switch (error) { + case IXGBE_ERR_EEPROM_VERSION: device_printf(dev, "This device is a pre-production adapter/" "LOM. Please be aware there may be issues associated " "with your hardware.\n If you are experiencing problems " "please contact your Intel or hardware representative " "who provided you with this hardware.\n"); - } else if (error == IXGBE_ERR_SFP_NOT_SUPPORTED) + break; + case IXGBE_ERR_SFP_NOT_SUPPORTED: device_printf(dev,"Unsupported SFP+ Module\n"); - - if (error) { error = EIO; - device_printf(dev,"Hardware Initialization Failure\n"); goto err_late; + case IXGBE_ERR_SFP_NOT_PRESENT: + device_printf(dev,"No SFP+ Module found\n"); + /* falls thru */ + default: + break; } /* Detect and set physical type */ @@ -554,11 +568,6 @@ ixgbe_attach(device_t dev) if (ixgbe_setup_interface(dev, adapter) != 0) goto err_late; - /* Sysctl for limiting the amount of work done in the taskqueue */ - ixgbe_add_rx_process_limit(adapter, "rx_processing_limit", - "max number of rx packets to process", &adapter->rx_process_limit, - ixgbe_rx_process_limit); - /* Initialize statistics */ ixgbe_update_stats_counters(adapter); @@ -594,6 +603,9 @@ ixgbe_attach(device_t dev) ixgbe_add_hw_stats(adapter); +#ifdef DEV_NETMAP + ixgbe_netmap_attach(adapter); +#endif /* DEV_NETMAP */ INIT_DEBUGOUT("ixgbe_attach: end"); return (0); err_late: @@ -640,7 +652,7 @@ ixgbe_detach(device_t dev) for (int i = 0; i < adapter->num_queues; i++, que++, txr++) { if (que->tq) { -#if __FreeBSD_version >= 800000 +#ifdef IXGBE_LEGACY_TX taskqueue_drain(que->tq, &txr->txq_task); #endif taskqueue_drain(que->tq, &que->que_task); @@ -672,6 +684,9 @@ ixgbe_detach(device_t dev) ether_ifdetach(adapter->ifp); callout_drain(&adapter->timer); +#ifdef DEV_NETMAP + netmap_detach(adapter->ifp); +#endif /* DEV_NETMAP */ ixgbe_free_pci_resources(adapter); bus_generic_detach(dev); if_free(adapter->ifp); @@ -701,7 +716,7 @@ ixgbe_shutdown(device_t dev) } -#if __FreeBSD_version < 800000 +#ifdef IXGBE_LEGACY_TX /********************************************************************* * Transmit entry point * @@ -720,17 +735,14 @@ ixgbe_start_locked(struct tx_ring *txr, IXGBE_TX_LOCK_ASSERT(txr); - if ((ifp->if_drv_flags & (IFF_DRV_RUNNING|IFF_DRV_OACTIVE)) != - IFF_DRV_RUNNING) + if ((ifp->if_drv_flags & IFF_DRV_RUNNING) == 0) return; if (!adapter->link_active) return; while (!IFQ_DRV_IS_EMPTY(&ifp->if_snd)) { - if (txr->tx_avail <= IXGBE_QUEUE_MIN_FREE) { - txr->queue_status |= IXGBE_QUEUE_DEPLETED; + if (txr->tx_avail <= IXGBE_QUEUE_MIN_FREE) break; - } IFQ_DRV_DEQUEUE(&ifp->if_snd, m_head); if (m_head == NULL) @@ -739,8 +751,6 @@ ixgbe_start_locked(struct tx_ring *txr, if (ixgbe_xmit(txr, &m_head)) { if (m_head != NULL) IFQ_DRV_PREPEND(&ifp->if_snd, m_head); - if (txr->tx_avail <= IXGBE_QUEUE_MIN_FREE) - txr->queue_status |= IXGBE_QUEUE_DEPLETED; break; } /* Send a copy of the frame to the BPF listener */ @@ -773,7 +783,8 @@ ixgbe_start(struct ifnet *ifp) return; } -#else +#else /* ! IXGBE_LEGACY_TX */ + /* ** Multiqueue Transmit driver ** @@ -795,8 +806,7 @@ ixgbe_mq_start(struct ifnet *ifp, struct txr = &adapter->tx_rings[i]; que = &adapter->queues[i]; - if (((txr->queue_status & IXGBE_QUEUE_DEPLETED) == 0) && - IXGBE_TX_TRYLOCK(txr)) { + if (IXGBE_TX_TRYLOCK(txr)) { err = ixgbe_mq_start_locked(ifp, txr, m); IXGBE_TX_UNLOCK(txr); } else { @@ -815,17 +825,12 @@ ixgbe_mq_start_locked(struct ifnet *ifp, int enqueued, err = 0; if (((ifp->if_drv_flags & IFF_DRV_RUNNING) == 0) || - (txr->queue_status == IXGBE_QUEUE_DEPLETED) || adapter->link_active == 0) { if (m != NULL) err = drbr_enqueue(ifp, txr->br, m); return (err); } - /* Call cleanup if number of TX descriptors low */ - if (txr->tx_avail <= IXGBE_TX_CLEANUP_THRESHOLD) - ixgbe_txeof(txr); - enqueued = 0; if (m == NULL) { next = drbr_dequeue(ifp, txr->br); @@ -844,23 +849,21 @@ ixgbe_mq_start_locked(struct ifnet *ifp, break; } enqueued++; - drbr_stats_update(ifp, next->m_pkthdr.len, next->m_flags); + ifp->if_obytes += next->m_pkthdr.len; + if (next->m_flags & M_MCAST) + ifp->if_omcasts++; /* Send a copy of the frame to the BPF listener */ ETHER_BPF_MTAP(ifp, next); if ((ifp->if_drv_flags & IFF_DRV_RUNNING) == 0) break; if (txr->tx_avail < IXGBE_TX_OP_THRESHOLD) ixgbe_txeof(txr); - if (txr->tx_avail < IXGBE_TX_OP_THRESHOLD) { - txr->queue_status |= IXGBE_QUEUE_DEPLETED; - break; - } next = drbr_dequeue(ifp, txr->br); } if (enqueued > 0) { /* Set watchdog on */ - txr->queue_status |= IXGBE_QUEUE_WORKING; + txr->queue_status = IXGBE_QUEUE_WORKING; txr->watchdog_time = ticks; } @@ -904,7 +907,7 @@ ixgbe_qflush(struct ifnet *ifp) } if_qflush(ifp); } -#endif /* __FreeBSD_version >= 800000 */ +#endif /* IXGBE_LEGACY_TX */ /********************************************************************* * Ioctl entry point @@ -919,6 +922,7 @@ static int ixgbe_ioctl(struct ifnet * ifp, u_long command, caddr_t data) { struct adapter *adapter = ifp->if_softc; + struct ixgbe_hw *hw = &adapter->hw; struct ifreq *ifr = (struct ifreq *) data; #if defined(INET) || defined(INET6) struct ifaddr *ifa = (struct ifaddr *)data; @@ -1006,6 +1010,8 @@ ixgbe_ioctl(struct ifnet * ifp, u_long c ifp->if_capenable ^= IFCAP_HWCSUM; if (mask & IFCAP_TSO4) ifp->if_capenable ^= IFCAP_TSO4; + if (mask & IFCAP_TSO6) + ifp->if_capenable ^= IFCAP_TSO6; if (mask & IFCAP_LRO) ifp->if_capenable ^= IFCAP_LRO; if (mask & IFCAP_VLAN_HWTAGGING) @@ -1022,7 +1028,22 @@ ixgbe_ioctl(struct ifnet * ifp, u_long c VLAN_CAPABILITIES(ifp); break; } - + case SIOCGI2C: + { + struct ixgbe_i2c_req i2c; + IOCTL_DEBUGOUT("ioctl: SIOCGI2C (Get I2C Data)"); + error = copyin(ifr->ifr_data, &i2c, sizeof(i2c)); + if (error) + break; + if ((i2c.dev_addr != 0xA0) || (i2c.dev_addr != 0xA2)){ + error = EINVAL; + break; + } + hw->phy.ops.read_i2c_byte(hw, i2c.offset, + i2c.dev_addr, i2c.data); + error = copyout(&i2c, ifr->ifr_data, sizeof(i2c)); + break; + } default: IOCTL_DEBUGOUT1("ioctl: UNKNOWN (0x%X)\n", (int)command); error = ether_ioctl(ifp, command, data); @@ -1070,7 +1091,7 @@ ixgbe_init_locked(struct adapter *adapte /* Set the various hardware offload abilities */ ifp->if_hwassist = 0; - if (ifp->if_capenable & IFCAP_TSO4) + if (ifp->if_capenable & IFCAP_TSO) ifp->if_hwassist |= CSUM_TSO; if (ifp->if_capenable & IFCAP_TXCSUM) { ifp->if_hwassist |= (CSUM_TCP | CSUM_UDP); @@ -1095,7 +1116,7 @@ ixgbe_init_locked(struct adapter *adapte /* ** Determine the correct mbuf pool - ** for doing jumbo/headersplit + ** for doing jumbo frames */ if (adapter->max_frame_size <= 2048) adapter->rx_mbuf_sz = MCLBYTES; @@ -1152,6 +1173,14 @@ ixgbe_init_locked(struct adapter *adapte txdctl |= IXGBE_TXDCTL_ENABLE; /* Set WTHRESH to 8, burst writeback */ txdctl |= (8 << 16); + /* + * When the internal queue falls below PTHRESH (32), + * start prefetching as long as there are at least + * HTHRESH (1) buffers ready. The values are taken + * from the Intel linux driver 3.8.21. + * Prefetching enables tx line rate even with 1 queue. + */ + txdctl |= (32 << 0) | (1 << 8); IXGBE_WRITE_REG(hw, IXGBE_TXDCTL(i), txdctl); } @@ -1176,6 +1205,31 @@ ixgbe_init_locked(struct adapter *adapte msec_delay(1); } wmb(); +#ifdef DEV_NETMAP + /* + * In netmap mode, we must preserve the buffers made + * available to userspace before the if_init() + * (this is true by default on the TX side, because + * init makes all buffers available to userspace). + * + * netmap_reset() and the device specific routines + * (e.g. ixgbe_setup_receive_rings()) map these + * buffers at the end of the NIC ring, so here we + * must set the RDT (tail) register to make sure + * they are not overwritten. + * + * In this driver the NIC ring starts at RDH = 0, + * RDT points to the last slot available for reception (?), + * so RDT = num_rx_desc - 1 means the whole ring is available. + */ + if (ifp->if_capenable & IFCAP_NETMAP) { + struct netmap_adapter *na = NA(adapter->ifp); + struct netmap_kring *kring = &na->rx_rings[i]; + int t = na->num_rx_desc - 1 - kring->nr_hwavail; + + IXGBE_WRITE_REG(hw, IXGBE_RDT(i), t); + } else +#endif /* DEV_NETMAP */ IXGBE_WRITE_REG(hw, IXGBE_RDT(i), adapter->num_rx_desc - 1); } @@ -1210,7 +1264,7 @@ ixgbe_init_locked(struct adapter *adapte #ifdef IXGBE_FDIR /* Init Flow director */ if (hw->mac.type != ixgbe_mac_82598EB) { - u32 hdrm = 64 << fdir_pballoc; + u32 hdrm = 32 << fdir_pballoc; hw->mac.ops.setup_rxpba(hw, 0, hdrm, PBA_STRATEGY_EQUAL); ixgbe_init_fdir_signature_82599(&adapter->hw, fdir_pballoc); @@ -1236,12 +1290,40 @@ ixgbe_init_locked(struct adapter *adapte /* Config/Enable Link */ ixgbe_config_link(adapter); + /* Hardware Packet Buffer & Flow Control setup */ + { + u32 rxpb, frame, size, tmp; + + frame = adapter->max_frame_size; + + /* Calculate High Water */ + if (hw->mac.type == ixgbe_mac_X540) + tmp = IXGBE_DV_X540(frame, frame); + else + tmp = IXGBE_DV(frame, frame); + size = IXGBE_BT2KB(tmp); + rxpb = IXGBE_READ_REG(hw, IXGBE_RXPBSIZE(0)) >> 10; + hw->fc.high_water[0] = rxpb - size; + + /* Now calculate Low Water */ + if (hw->mac.type == ixgbe_mac_X540) + tmp = IXGBE_LOW_DV_X540(frame); + else + tmp = IXGBE_LOW_DV(frame); + hw->fc.low_water[0] = IXGBE_BT2KB(tmp); + + adapter->fc = hw->fc.requested_mode = ixgbe_fc_full; + hw->fc.pause_time = IXGBE_FC_PAUSE; + hw->fc.send_xon = TRUE; + } + /* Initialize the FC settings */ + ixgbe_start_hw(hw); + /* And now turn on interrupts */ ixgbe_enable_intr(adapter); /* Now inform the stack we're ready */ ifp->if_drv_flags |= IFF_DRV_RUNNING; - ifp->if_drv_flags &= ~IFF_DRV_OACTIVE; return; } @@ -1331,10 +1413,10 @@ ixgbe_handle_que(void *context, int pend bool more; if (ifp->if_drv_flags & IFF_DRV_RUNNING) { - more = ixgbe_rxeof(que, adapter->rx_process_limit); + more = ixgbe_rxeof(que); IXGBE_TX_LOCK(txr); ixgbe_txeof(txr); -#if __FreeBSD_version >= 800000 +#ifndef IXGBE_LEGACY_TX if (!drbr_empty(ifp, txr->br)) ixgbe_mq_start_locked(ifp, txr, NULL); #else @@ -1342,7 +1424,7 @@ ixgbe_handle_que(void *context, int pend ixgbe_start_locked(txr, ifp); #endif IXGBE_TX_UNLOCK(txr); - if (more || (ifp->if_drv_flags & IFF_DRV_OACTIVE)) { + if (more) { taskqueue_enqueue(que->tq, &que->que_task); return; } @@ -1379,7 +1461,7 @@ ixgbe_legacy_irq(void *arg) return; } - more_rx = ixgbe_rxeof(que, adapter->rx_process_limit); + more_rx = ixgbe_rxeof(que); IXGBE_TX_LOCK(txr); do { @@ -1425,7 +1507,7 @@ ixgbe_msix_que(void *arg) ixgbe_disable_queue(adapter, que->msix); ++que->irqs; - more_rx = ixgbe_rxeof(que, adapter->rx_process_limit); + more_rx = ixgbe_rxeof(que); IXGBE_TX_LOCK(txr); more_tx = ixgbe_txeof(txr); @@ -1434,7 +1516,7 @@ ixgbe_msix_que(void *arg) ** has anything queued the task gets ** scheduled to handle it. */ -#if __FreeBSD_version < 800000 +#ifdef IXGBE_LEGACY_TX if (!IFQ_DRV_IS_EMPTY(&adapter->ifp->if_snd)) #else if (!drbr_empty(adapter->ifp, txr->br)) @@ -1525,10 +1607,8 @@ ixgbe_msix_link(void *arg) /* This is probably overkill :) */ if (!atomic_cmpset_int(&adapter->fdir_reinit, 0, 1)) return; - /* Clear the interrupt */ - IXGBE_WRITE_REG(hw, IXGBE_EICR, IXGBE_EICR_FLOW_DIR); - /* Turn off the interface */ - adapter->ifp->if_drv_flags &= ~IFF_DRV_RUNNING; + /* Disable the interrupt */ + IXGBE_WRITE_REG(hw, IXGBE_EIMC, IXGBE_EICR_FLOW_DIR); taskqueue_enqueue(adapter->tq, &adapter->fdir_task); } else #endif @@ -1602,7 +1682,7 @@ ixgbe_media_status(struct ifnet * ifp, s ifmr->ifm_active |= IFM_100_TX | IFM_FDX; break; case IXGBE_LINK_SPEED_1GB_FULL: - ifmr->ifm_active |= IFM_1000_T | IFM_FDX; + ifmr->ifm_active |= adapter->optics | IFM_FDX; break; case IXGBE_LINK_SPEED_10GB_FULL: ifmr->ifm_active |= adapter->optics | IFM_FDX; @@ -1661,9 +1741,9 @@ ixgbe_xmit(struct tx_ring *txr, struct m { struct adapter *adapter = txr->adapter; u32 olinfo_status = 0, cmd_type_len; - u32 paylen = 0; int i, j, error, nsegs; - int first, last = 0; + int first; + bool remap = TRUE; struct mbuf *m_head; bus_dma_segment_t segs[adapter->num_segs]; bus_dmamap_t map; @@ -1691,74 +1771,58 @@ ixgbe_xmit(struct tx_ring *txr, struct m /* * Map the packet for DMA. */ +retry: error = bus_dmamap_load_mbuf_sg(txr->txtag, map, *m_headp, segs, &nsegs, BUS_DMA_NOWAIT); - if (error == EFBIG) { + if (__predict_false(error)) { struct mbuf *m; - m = m_defrag(*m_headp, M_DONTWAIT); - if (m == NULL) { - adapter->mbuf_defrag_failed++; - m_freem(*m_headp); - *m_headp = NULL; - return (ENOBUFS); - } - *m_headp = m; - - /* Try it again */ - error = bus_dmamap_load_mbuf_sg(txr->txtag, map, - *m_headp, segs, &nsegs, BUS_DMA_NOWAIT); - - if (error == ENOMEM) { - adapter->no_tx_dma_setup++; + switch (error) { + case EFBIG: + /* Try it again? - one try */ + if (remap == TRUE) { + remap = FALSE; + m = m_defrag(*m_headp, M_NOWAIT); + if (m == NULL) { + adapter->mbuf_defrag_failed++; + m_freem(*m_headp); + *m_headp = NULL; + return (ENOBUFS); + } + *m_headp = m; + goto retry; + } else + return (error); + case ENOMEM: + txr->no_tx_dma_setup++; return (error); - } else if (error != 0) { - adapter->no_tx_dma_setup++; + default: + txr->no_tx_dma_setup++; m_freem(*m_headp); *m_headp = NULL; return (error); } - } else if (error == ENOMEM) { - adapter->no_tx_dma_setup++; - return (error); - } else if (error != 0) { - adapter->no_tx_dma_setup++; - m_freem(*m_headp); - *m_headp = NULL; - return (error); } /* Make certain there are enough descriptors */ if (nsegs > txr->tx_avail - 2) { txr->no_desc_avail++; - error = ENOBUFS; - goto xmit_fail; + bus_dmamap_unload(txr->txtag, map); + return (ENOBUFS); } m_head = *m_headp; /* ** Set up the appropriate offload context - ** this becomes the first descriptor of - ** a packet. + ** this will consume the first descriptor */ - if (m_head->m_pkthdr.csum_flags & CSUM_TSO) { - if (ixgbe_tso_setup(txr, m_head, &paylen)) { - cmd_type_len |= IXGBE_ADVTXD_DCMD_TSE; - olinfo_status |= IXGBE_TXD_POPTS_IXSM << 8; - olinfo_status |= IXGBE_TXD_POPTS_TXSM << 8; - olinfo_status |= paylen << IXGBE_ADVTXD_PAYLEN_SHIFT; - ++adapter->tso_tx; - } else - return (ENXIO); - } else if (ixgbe_tx_ctx_setup(txr, m_head)) - olinfo_status |= IXGBE_TXD_POPTS_TXSM << 8; - -#ifdef IXGBE_IEEE1588 - /* This is changing soon to an mtag detection */ - if (we detect this mbuf has a TSTAMP mtag) - cmd_type_len |= IXGBE_ADVTXD_MAC_TSTAMP; -#endif + error = ixgbe_tx_ctx_setup(txr, m_head, &cmd_type_len, &olinfo_status); + if (__predict_false(error)) { + if (error == ENOBUFS) + *m_headp = NULL; + return (error); + } #ifdef IXGBE_FDIR /* Do the flow director magic */ @@ -1770,10 +1834,6 @@ ixgbe_xmit(struct tx_ring *txr, struct m } } #endif - /* Record payload length */ - if (paylen == 0) - olinfo_status |= m_head->m_pkthdr.len << - IXGBE_ADVTXD_PAYLEN_SHIFT; i = txr->next_avail_desc; for (j = 0; j < nsegs; j++) { @@ -1789,13 +1849,9 @@ ixgbe_xmit(struct tx_ring *txr, struct m txd->read.cmd_type_len = htole32(txr->txd_cmd | cmd_type_len |seglen); txd->read.olinfo_status = htole32(olinfo_status); - last = i; /* descriptor that will get completion IRQ */ - if (++i == adapter->num_tx_desc) + if (++i == txr->num_desc) i = 0; - - txbuf->m_head = NULL; - txbuf->eop_index = -1; } txd->read.cmd_type_len |= @@ -1804,14 +1860,19 @@ ixgbe_xmit(struct tx_ring *txr, struct m txr->next_avail_desc = i; txbuf->m_head = m_head; - /* Swap the dma map between the first and last descriptor */ + /* + ** Here we swap the map so the last descriptor, + ** which gets the completion interrupt has the + ** real map, and the first descriptor gets the + ** unused map from this descriptor. + */ txr->tx_buffers[first].map = txbuf->map; txbuf->map = map; bus_dmamap_sync(txr->txtag, map, BUS_DMASYNC_PREWRITE); - /* Set the index of the descriptor that will be marked done */ + /* Set the EOP descriptor that will be marked done */ txbuf = &txr->tx_buffers[first]; - txbuf->eop_index = last; + txbuf->eop = txd; bus_dmamap_sync(txr->txdma.dma_tag, txr->txdma.dma_map, BUS_DMASYNC_PREREAD | BUS_DMASYNC_PREWRITE); @@ -1824,10 +1885,6 @@ ixgbe_xmit(struct tx_ring *txr, struct m return (0); -xmit_fail: - bus_dmamap_unload(txr->txtag, txbuf->map); - return (error); - } static void @@ -1946,13 +2003,11 @@ ixgbe_local_timer(void *arg) { struct adapter *adapter = arg; device_t dev = adapter->dev; - struct ifnet *ifp = adapter->ifp; struct ix_queue *que = adapter->queues; struct tx_ring *txr = adapter->tx_rings; - int hung, busy, paused; + int hung = 0, paused = 0; mtx_assert(&adapter->core_mtx, MA_OWNED); - hung = busy = paused = 0; /* Check for pluggable optics */ if (adapter->sfp_probe) @@ -1971,27 +2026,18 @@ ixgbe_local_timer(void *arg) /* ** Check the TX queues status - ** - central locked handling of OACTIVE ** - watchdog only if all queues show hung */ for (int i = 0; i < adapter->num_queues; i++, que++, txr++) { - if ((txr->queue_status & IXGBE_QUEUE_HUNG) && + if ((txr->queue_status == IXGBE_QUEUE_HUNG) && (paused == 0)) ++hung; - if (txr->queue_status & IXGBE_QUEUE_DEPLETED) - ++busy; - if ((txr->queue_status & IXGBE_QUEUE_IDLE) == 0) + else if (txr->queue_status == IXGBE_QUEUE_WORKING) taskqueue_enqueue(que->tq, &que->que_task); } /* Only truely watchdog if all queues show hung */ if (hung == adapter->num_queues) goto watchdog; - /* Only turn off the stack flow when ALL are depleted */ - if (busy == adapter->num_queues) - ifp->if_drv_flags |= IFF_DRV_OACTIVE; - else if ((ifp->if_drv_flags & IFF_DRV_OACTIVE) && - (busy < adapter->num_queues)) - ifp->if_drv_flags &= ~IFF_DRV_OACTIVE; out: ixgbe_rearm_queues(adapter, adapter->que_mask); @@ -2020,7 +2066,6 @@ static void ixgbe_update_link_status(struct adapter *adapter) { struct ifnet *ifp = adapter->ifp; - struct tx_ring *txr = adapter->tx_rings; device_t dev = adapter->dev; @@ -2031,6 +2076,8 @@ ixgbe_update_link_status(struct adapter ((adapter->link_speed == 128)? 10:1), "Full Duplex"); adapter->link_active = TRUE; + /* Update any Flow Control changes */ + ixgbe_fc_enable(&adapter->hw); if_link_state_change(ifp, LINK_STATE_UP); } } else { /* Link down */ @@ -2039,9 +2086,6 @@ ixgbe_update_link_status(struct adapter device_printf(dev,"Link is Down\n"); if_link_state_change(ifp, LINK_STATE_DOWN); adapter->link_active = FALSE; - for (int i = 0; i < adapter->num_queues; - i++, txr++) - txr->queue_status = IXGBE_QUEUE_IDLE; } } @@ -2072,7 +2116,6 @@ ixgbe_stop(void *arg) /* Let the stack know...*/ ifp->if_drv_flags &= ~IFF_DRV_RUNNING; - ifp->if_drv_flags |= IFF_DRV_OACTIVE; ixgbe_reset_hw(hw); hw->adapter_stopped = FALSE; @@ -2144,6 +2187,11 @@ ixgbe_setup_optics(struct adapter *adapt return; } + if (layer & IXGBE_PHYSICAL_LAYER_1000BASE_SX) { + adapter->optics = IFM_1000_SX; + return; + } + if (layer & (IXGBE_PHYSICAL_LAYER_10GBASE_LR | IXGBE_PHYSICAL_LAYER_10GBASE_LRM)) { adapter->optics = IFM_10G_LR; @@ -2179,12 +2227,12 @@ ixgbe_setup_optics(struct adapter *adapt static int ixgbe_allocate_legacy(struct adapter *adapter) { - device_t dev = adapter->dev; + device_t dev = adapter->dev; struct ix_queue *que = adapter->queues; -#if __FreeBSD_version >= 800000 +#ifndef IXGBE_LEGACY_TX struct tx_ring *txr = adapter->tx_rings; #endif - int error, rid = 0; + int error, rid = 0; /* MSI RID at 1 */ if (adapter->msix == 1) @@ -2203,7 +2251,7 @@ ixgbe_allocate_legacy(struct adapter *ad * Try allocating a fast interrupt and the associated deferred * processing contexts. */ -#if __FreeBSD_version >= 800000 +#ifndef IXGBE_LEGACY_TX TASK_INIT(&txr->txq_task, 0, ixgbe_deferred_mq_start, txr); #endif TASK_INIT(&que->que_task, 0, ixgbe_handle_que, que); @@ -2285,7 +2333,7 @@ ixgbe_allocate_msix(struct adapter *adap if (adapter->num_queues > 1) bus_bind_intr(dev, que->res, i); -#if __FreeBSD_version >= 800000 +#ifndef IXGBE_LEGACY_TX TASK_INIT(&txr->txq_task, 0, ixgbe_deferred_mq_start, txr); #endif TASK_INIT(&que->que_task, 0, ixgbe_handle_que, que); @@ -2526,13 +2574,12 @@ ixgbe_setup_interface(device_t dev, stru return (-1); } if_initname(ifp, device_get_name(dev), device_get_unit(dev)); - ifp->if_mtu = ETHERMTU; ifp->if_baudrate = 1000000000; ifp->if_init = ixgbe_init; ifp->if_softc = adapter; ifp->if_flags = IFF_BROADCAST | IFF_SIMPLEX | IFF_MULTICAST; ifp->if_ioctl = ixgbe_ioctl; -#if __FreeBSD_version >= 800000 +#ifndef IXGBE_LEGACY_TX ifp->if_transmit = ixgbe_mq_start; ifp->if_qflush = ixgbe_qflush; #else @@ -2550,16 +2597,14 @@ ixgbe_setup_interface(device_t dev, stru */ ifp->if_data.ifi_hdrlen = sizeof(struct ether_vlan_header); - ifp->if_capabilities |= IFCAP_HWCSUM | IFCAP_TSO4 | IFCAP_VLAN_HWCSUM; + ifp->if_capabilities |= IFCAP_HWCSUM | IFCAP_TSO | IFCAP_VLAN_HWCSUM; ifp->if_capabilities |= IFCAP_JUMBO_MTU; + ifp->if_capabilities |= IFCAP_LRO; ifp->if_capabilities |= IFCAP_VLAN_HWTAGGING | IFCAP_VLAN_HWTSO | IFCAP_VLAN_MTU; ifp->if_capenable = ifp->if_capabilities; - /* Don't enable LRO by default */ - ifp->if_capabilities |= IFCAP_LRO; - /* ** Don't turn this on by default, if vlans are ** created on another pseudo device (eg. lagg) @@ -2759,6 +2804,7 @@ ixgbe_allocate_queues(struct adapter *ad txr = &adapter->tx_rings[i]; txr->adapter = adapter; txr->me = i; + txr->num_desc = adapter->num_tx_desc; /* Initialize the TX side lock */ snprintf(txr->mtx_name, sizeof(txr->mtx_name), "%s:tx(%d)", @@ -2782,7 +2828,7 @@ ixgbe_allocate_queues(struct adapter *ad error = ENOMEM; goto err_tx_desc; } -#if __FreeBSD_version >= 800000 +#ifndef IXGBE_LEGACY_TX /* Allocate a buf ring */ txr->br = buf_ring_alloc(IXGBE_BR_SIZE, M_DEVBUF, M_WAITOK, &txr->tx_mtx); @@ -2805,6 +2851,7 @@ ixgbe_allocate_queues(struct adapter *ad /* Set up some basics */ rxr->adapter = adapter; rxr->me = i; + rxr->num_desc = adapter->num_rx_desc; /* Initialize the RX side lock */ snprintf(rxr->mtx_name, sizeof(rxr->mtx_name), "%s:rx(%d)", @@ -2928,9 +2975,20 @@ ixgbe_setup_transmit_ring(struct tx_ring struct adapter *adapter = txr->adapter; struct ixgbe_tx_buf *txbuf; int i; +#ifdef DEV_NETMAP + struct netmap_adapter *na = NA(adapter->ifp); + struct netmap_slot *slot; +#endif /* DEV_NETMAP */ /* Clear the old ring contents */ IXGBE_TX_LOCK(txr); +#ifdef DEV_NETMAP + /* + * (under lock): if in netmap mode, do some consistency + * checks and set slot to entry 0 of the netmap ring. + */ + slot = netmap_reset(na, NR_TX, txr->me, 0); +#endif /* DEV_NETMAP */ bzero((void *)txr->tx_base, (sizeof(union ixgbe_adv_tx_desc)) * adapter->num_tx_desc); /* Reset indices */ @@ -2939,7 +2997,7 @@ ixgbe_setup_transmit_ring(struct tx_ring /* Free any existing tx buffers. */ txbuf = txr->tx_buffers; - for (i = 0; i < adapter->num_tx_desc; i++, txbuf++) { + for (i = 0; i < txr->num_desc; i++, txbuf++) { if (txbuf->m_head != NULL) { bus_dmamap_sync(txr->txtag, txbuf->map, BUS_DMASYNC_POSTWRITE); @@ -2947,8 +3005,24 @@ ixgbe_setup_transmit_ring(struct tx_ring *** DIFF OUTPUT TRUNCATED AT 1000 LINES ***