Date: Mon, 31 Oct 2016 22:45:12 +0000 (UTC) From: John Baldwin <jhb@FreeBSD.org> To: src-committers@freebsd.org, svn-src-all@freebsd.org, svn-src-stable@freebsd.org, svn-src-stable-10@freebsd.org Subject: svn commit: r308154 - in stable/10: share/man/man4 sys/dev/cxgbe sys/dev/cxgbe/common sys/dev/cxgbe/iw_cxgbe sys/dev/cxgbe/tom Message-ID: <201610312245.u9VMjCTB019847@repo.freebsd.org>
next in thread | raw e-mail | index | archive | help
Author: jhb Date: Mon Oct 31 22:45:11 2016 New Revision: 308154 URL: https://svnweb.freebsd.org/changeset/base/308154 Log: MFC 291665,291685,291856,297467,302110,302263: Add support for VIs. 291665: Add support for configuring additional virtual interfaces (VIs) on a port. Each virtual interface has its own MAC address, queues, and statistics. The dedicated netmap interfaces (ncxgbeX / ncxlX) were already implemented as additional VIs on each port. This change allows additional non-netmap interfaces to be configured on each port. Additional virtual interfaces use the naming scheme vcxgbeX or vcxlX. Additional VIs are enabled by setting the hw.cxgbe.num_vis tunable to a value greater than 1 before loading the cxgbe(4) or cxl(4) driver. NB: The first VI on each port is the "main" interface (cxgbeX or cxlX). T4/T5 NICs provide a limited number of MAC addresses for each physical port. As a result, a maximum of six VIs can be configured on each port (including the "main" interface and the netmap interface when netmap is enabled). One user-visible result is that when netmap is enabled, packets received or transmitted via the netmap interface are no longer counted in the stats for the "main" interface, but are not accounted to the netmap interface. The netmap interfaces now also have a new-bus device and export various information sysctl nodes via dev.n(cxgbe|cxl).X. The cxgbetool 'clearstats' command clears the stats for all VIs on the specified port along with the port's stats. There is currently no way to clear the stats of an individual VI. 291685: Fix build for !TCP_OFFLOAD case. 291856: Fix RSS build. 297467: Remove #ifdef's from various structures used in the cxgbe/cxl driver. This provides a constant ABI and layout for these structures (especially struct adapter) avoiding some foot shooting. 302110: cxgbe(4): Merge netmap support from the ncxgbe/ncxl interfaces to the vcxgbe/vcxl interfaces and retire the 'n' interfaces. The main cxgbe/cxl interfaces and tunables related to them are not affected by any of this and will continue to operate as usual. The driver used to create an additional 'n' interface for every cxgbe/cxl interface if "device netmap" was in the kernel. The 'n' interface shared the wire with the main interface but was otherwise autonomous (with its own MAC address, etc.). It did not have normal tx/rx but had a specialized netmap-only data path. r291665 added another set of virtual interfaces (the 'v' interfaces) to the driver. These had normal tx/rx but no netmap support. This revision consolidates the features of both the interfaces into the 'v' interface which now has a normal data path, TOE support, and native netmap support. The 'v' interfaces need to be created explicitly with the hw.cxgbe.num_vis tunable. This means "device netmap" will not result in the automatic creation of any virtual interfaces. The following tunables can be used to override the default number of queues allocated for each 'v' interface. nofld* = 0 will disable TOE on the virtual interface and nnm* = 0 to will disable native netmap support. # number of normal NIC queues hw.cxgbe.ntxq_vi hw.cxgbe.nrxq_vi # number of TOE queues hw.cxgbe.nofldtxq_vi hw.cxgbe.nofldrxq_vi # number of netmap queues hw.cxgbe.nnmtxq_vi hw.cxgbe.nnmrxq_vi hw.cxgbe.nnm{t,r}xq{10,1}g tunables have been removed. --- tl;dr version --- The workflow for netmap on cxgbe starting with FreeBSD 11 is: 1) "device netmap" in the kernel config. 2) "hw.cxgbe.num_vis=2" in loader.conf. num_vis > 2 is ok too, you'll end up with multiple autonomous netmap-capable interfaces for every port. 3) "dmesg | grep vcxl | grep netmap" to verify that the interface has netmap queues. 4) Use any of the 'v' interfaces for netmap. pkt-gen -i vcxl<n>... . One major improvement is that the netmap interface has a normal data path as expected. 5) Just ignore the cxl interfaces if you want to use netmap only. No need to bring them up. The vcxl interfaces are completely independent and everything should just work. --------------------- 302263: cxgbe(4): Do not bring up an interface when IFCAP_TOE is enabled on it. The interface's queues are functional after VI_INIT_DONE (which is short of interface-up) and that's all that's needed for t4_tom to communicate with the chip. Relnotes: yes Sponsored by: Chelsio Communications Modified: stable/10/share/man/man4/cxgbe.4 stable/10/sys/dev/cxgbe/adapter.h stable/10/sys/dev/cxgbe/common/t4_hw.c stable/10/sys/dev/cxgbe/iw_cxgbe/provider.c stable/10/sys/dev/cxgbe/offload.h stable/10/sys/dev/cxgbe/t4_main.c stable/10/sys/dev/cxgbe/t4_netmap.c stable/10/sys/dev/cxgbe/t4_sge.c stable/10/sys/dev/cxgbe/tom/t4_connect.c stable/10/sys/dev/cxgbe/tom/t4_cpl_io.c stable/10/sys/dev/cxgbe/tom/t4_listen.c stable/10/sys/dev/cxgbe/tom/t4_tom.c stable/10/sys/dev/cxgbe/tom/t4_tom.h Directory Properties: stable/10/ (props changed) Modified: stable/10/share/man/man4/cxgbe.4 ============================================================================== --- stable/10/share/man/man4/cxgbe.4 Mon Oct 31 22:03:44 2016 (r308153) +++ stable/10/share/man/man4/cxgbe.4 Mon Oct 31 22:45:11 2016 (r308154) @@ -31,7 +31,7 @@ .\" .\" $FreeBSD$ .\" -.Dd March 20, 2014 +.Dd December 2, 2015 .Dt CXGBE 4 .Os .Sh NAME @@ -171,6 +171,16 @@ number of CPU cores in the system, which .It Va hw.cxgbe.nofldrxq1g The number of TOE rx queues to use for a 1Gb port. The default is 1. +.It Va hw.cxgbe.num_vis +The number of virtual interfaces (VIs) created for each port. +Each virtual interface creates a separate network interface. +The first virtual interface on each port is required and represents +the primary network interface on the port. +Additional virtual interfaces on a port are named vcxgbe (T4) or +vcxl (T5) and only use a single rx and tx queue. +Additional virtual interfaces use a single pair of queues +for rx and tx as well an additional pair of queues for TOE rx and tx. +The default is 1. .It Va hw.cxgbe.holdoff_timer_idx_10G .It Va hw.cxgbe.holdoff_timer_idx_1G The timer index value to use to delay interrupts. Modified: stable/10/sys/dev/cxgbe/adapter.h ============================================================================== --- stable/10/sys/dev/cxgbe/adapter.h Mon Oct 31 22:03:44 2016 (r308153) +++ stable/10/sys/dev/cxgbe/adapter.h Mon Oct 31 22:45:11 2016 (r308154) @@ -192,49 +192,46 @@ enum { CXGBE_BUSY = (1 << 9), /* port flags */ - DOOMED = (1 << 0), - PORT_INIT_DONE = (1 << 1), - PORT_SYSCTL_CTX = (1 << 2), HAS_TRACEQ = (1 << 3), + + /* VI flags */ + DOOMED = (1 << 0), + VI_INIT_DONE = (1 << 1), + VI_SYSCTL_CTX = (1 << 2), INTR_RXQ = (1 << 4), /* All NIC rxq's take interrupts */ INTR_OFLD_RXQ = (1 << 5), /* All TOE rxq's take interrupts */ - INTR_NM_RXQ = (1 << 6), /* All netmap rxq's take interrupts */ - INTR_ALL = (INTR_RXQ | INTR_OFLD_RXQ | INTR_NM_RXQ), + INTR_ALL = (INTR_RXQ | INTR_OFLD_RXQ), /* adapter debug_flags */ DF_DUMP_MBOX = (1 << 0), }; -#define IS_DOOMED(pi) ((pi)->flags & DOOMED) -#define SET_DOOMED(pi) do {(pi)->flags |= DOOMED;} while (0) +#define IS_DOOMED(vi) ((vi)->flags & DOOMED) +#define SET_DOOMED(vi) do {(vi)->flags |= DOOMED;} while (0) #define IS_BUSY(sc) ((sc)->flags & CXGBE_BUSY) #define SET_BUSY(sc) do {(sc)->flags |= CXGBE_BUSY;} while (0) #define CLR_BUSY(sc) do {(sc)->flags &= ~CXGBE_BUSY;} while (0) -struct port_info { +struct vi_info { device_t dev; - struct adapter *adapter; + struct port_info *pi; struct ifnet *ifp; struct ifmedia media; - struct mtx pi_lock; - char lockname[16]; unsigned long flags; int if_flags; - uint16_t *rss; + uint16_t *rss, *nm_rss; uint16_t viid; int16_t xact_addr_filt;/* index of exact MAC address filter */ uint16_t rss_size; /* size of VI's RSS table slice */ uint16_t rss_base; /* start of VI's RSS table slice */ - uint8_t lport; /* associated offload logical port */ - int8_t mdio_addr; - uint8_t port_type; - uint8_t mod_type; - uint8_t port_id; - uint8_t tx_chan; - uint8_t rx_chan_map; /* rx MPS channel bitmap */ + + eventhandler_tag vlan_c; + + int nintr; + int first_intr; /* These need to be int as they are used in sysctl */ int ntxq; /* # of tx queues */ @@ -242,30 +239,49 @@ struct port_info { int rsrv_noflowq; /* Reserve queue 0 for non-flowid packets */ int nrxq; /* # of rx queues */ int first_rxq; /* index of first rx queue */ -#ifdef TCP_OFFLOAD int nofldtxq; /* # of offload tx queues */ int first_ofld_txq; /* index of first offload tx queue */ int nofldrxq; /* # of offload rx queues */ int first_ofld_rxq; /* index of first offload rx queue */ -#endif -#ifdef DEV_NETMAP - int nnmtxq; /* # of netmap tx queues */ - int first_nm_txq; /* index of first netmap tx queue */ - int nnmrxq; /* # of netmap rx queues */ - int first_nm_rxq; /* index of first netmap rx queue */ - - struct ifnet *nm_ifp; - struct ifmedia nm_media; - int nmif_flags; - uint16_t nm_viid; - int16_t nm_xact_addr_filt; - uint16_t nm_rss_size; /* size of netmap VI's RSS table slice */ -#endif + int nnmtxq; + int first_nm_txq; + int nnmrxq; + int first_nm_rxq; int tmr_idx; int pktc_idx; int qsize_rxq; int qsize_txq; + struct timeval last_refreshed; + struct fw_vi_stats_vf stats; + + struct callout tick; + struct sysctl_ctx_list ctx; /* from ifconfig up to driver detach */ + + uint8_t hw_addr[ETHER_ADDR_LEN]; /* factory MAC address, won't change */ +}; + +struct port_info { + device_t dev; + struct adapter *adapter; + + struct vi_info *vi; + int nvi; + int up_vis; + int uld_vis; + + struct mtx pi_lock; + char lockname[16]; + unsigned long flags; + + uint8_t lport; /* associated offload logical port */ + int8_t mdio_addr; + uint8_t port_type; + uint8_t mod_type; + uint8_t port_id; + uint8_t tx_chan; + uint8_t rx_chan_map; /* rx MPS channel bitmap */ + int linkdnrc; struct link_config link_cfg; @@ -273,14 +289,11 @@ struct port_info { struct port_stats stats; u_int tx_parse_error; - eventhandler_tag vlan_c; - struct callout tick; - struct sysctl_ctx_list ctx; /* from ifconfig up to driver detach */ - - uint8_t hw_addr[ETHER_ADDR_LEN]; /* factory MAC address, won't change */ }; +#define IS_MAIN_VI(vi) ((vi) == &((vi)->pi->vi[0])) + /* Where the cluster came from, how it has been carved up. */ struct cluster_layout { int8_t zidx; @@ -292,9 +305,7 @@ struct cluster_layout { struct cluster_metadata { u_int refcount; -#ifdef INVARIANTS struct fl_sdesc *sd; /* For debug only. Could easily be stale */ -#endif }; struct fl_sdesc { @@ -333,6 +344,11 @@ enum { IQS_DISABLED = 0, IQS_BUSY = 1, IQS_IDLE = 2, + + /* netmap related flags */ + NM_OFF = 0, + NM_ON = 1, + NM_BUSY = 2, }; /* @@ -529,7 +545,6 @@ iq_to_rxq(struct sge_iq *iq) } -#ifdef TCP_OFFLOAD /* ofld_rxq: SGE ingress queue + SGE free list + miscellaneous items */ struct sge_ofld_rxq { struct sge_iq iq; /* MUST be first */ @@ -542,7 +557,6 @@ iq_to_ofld_rxq(struct sge_iq *iq) return (__containerof(iq, struct sge_ofld_rxq, iq)); } -#endif struct wrqe { STAILQ_ENTRY(wrqe) link; @@ -594,9 +608,8 @@ struct sge_wrq { } __aligned(CACHE_LINE_SIZE); -#ifdef DEV_NETMAP struct sge_nm_rxq { - struct port_info *pi; + struct vi_info *vi; struct iq_desc *iq_desc; uint16_t iq_abs_id; @@ -649,7 +662,6 @@ struct sge_nm_txq { bus_addr_t ba; int iqidx; } __aligned(CACHE_LINE_SIZE); -#endif struct sge { int timer_val[SGE_NTIMERS]; @@ -661,14 +673,10 @@ struct sge { int nrxq; /* total # of Ethernet rx queues */ int ntxq; /* total # of Ethernet tx tx queues */ -#ifdef TCP_OFFLOAD int nofldrxq; /* total # of TOE rx queues */ int nofldtxq; /* total # of TOE tx queues */ -#endif -#ifdef DEV_NETMAP int nnmrxq; /* total # of netmap rx queues */ int nnmtxq; /* total # of netmap tx queues */ -#endif int niq; /* total # of ingress queues */ int neq; /* total # of egress queues */ @@ -677,14 +685,10 @@ struct sge { struct sge_wrq *ctrlq; /* Control queues */ struct sge_txq *txq; /* NIC tx queues */ struct sge_rxq *rxq; /* NIC rx queues */ -#ifdef TCP_OFFLOAD struct sge_wrq *ofld_txq; /* TOE tx queues */ struct sge_ofld_rxq *ofld_rxq; /* TOE rx queues */ -#endif -#ifdef DEV_NETMAP struct sge_nm_txq *nm_txq; /* netmap tx queues */ struct sge_nm_rxq *nm_rxq; /* netmap rx queues */ -#endif uint16_t iq_start; int eq_start; @@ -731,8 +735,11 @@ struct adapter { struct irq { struct resource *res; int rid; + volatile int nm_state; /* NM_OFF, NM_ON, or NM_BUSY */ void *tag; - } *irq; + struct sge_rxq *rxq; + struct sge_nm_rxq *nm_rxq; + } __aligned(CACHE_LINE_SIZE) *irq; bus_dma_tag_t dmat; /* Parent DMA tag */ @@ -743,21 +750,16 @@ struct adapter { struct port_info *port[MAX_NPORTS]; uint8_t chan_map[NCHAN]; -#ifdef TCP_OFFLOAD void *tom_softc; /* (struct tom_data *) */ struct tom_tunables tt; void *iwarp_softc; /* (struct c4iw_dev *) */ void *iscsi_softc; -#endif struct l2t_data *l2t; /* L2 table */ struct tid_info tids; uint16_t doorbells; - int open_device_map; -#ifdef TCP_OFFLOAD int offload_map; /* ports with IFCAP_TOE enabled */ int active_ulds; /* ULDs activated on this adapter */ -#endif int flags; int debug_flags; @@ -798,11 +800,9 @@ struct adapter { fw_msg_handler_t fw_msg_handler[5]; /* NUM_FW6_TYPES */ cpl_handler_t cpl_handler[0xef]; /* NUM_CPL_CMDS */ -#ifdef INVARIANTS const char *last_op; const void *last_op_thr; int last_op_flags; -#endif int sc_do_rxcopy; }; @@ -863,24 +863,27 @@ struct adapter { } \ } while (0) -#define for_each_txq(pi, iter, q) \ - for (q = &pi->adapter->sge.txq[pi->first_txq], iter = 0; \ - iter < pi->ntxq; ++iter, ++q) -#define for_each_rxq(pi, iter, q) \ - for (q = &pi->adapter->sge.rxq[pi->first_rxq], iter = 0; \ - iter < pi->nrxq; ++iter, ++q) -#define for_each_ofld_txq(pi, iter, q) \ - for (q = &pi->adapter->sge.ofld_txq[pi->first_ofld_txq], iter = 0; \ - iter < pi->nofldtxq; ++iter, ++q) -#define for_each_ofld_rxq(pi, iter, q) \ - for (q = &pi->adapter->sge.ofld_rxq[pi->first_ofld_rxq], iter = 0; \ - iter < pi->nofldrxq; ++iter, ++q) -#define for_each_nm_txq(pi, iter, q) \ - for (q = &pi->adapter->sge.nm_txq[pi->first_nm_txq], iter = 0; \ - iter < pi->nnmtxq; ++iter, ++q) -#define for_each_nm_rxq(pi, iter, q) \ - for (q = &pi->adapter->sge.nm_rxq[pi->first_nm_rxq], iter = 0; \ - iter < pi->nnmrxq; ++iter, ++q) +#define for_each_txq(vi, iter, q) \ + for (q = &vi->pi->adapter->sge.txq[vi->first_txq], iter = 0; \ + iter < vi->ntxq; ++iter, ++q) +#define for_each_rxq(vi, iter, q) \ + for (q = &vi->pi->adapter->sge.rxq[vi->first_rxq], iter = 0; \ + iter < vi->nrxq; ++iter, ++q) +#define for_each_ofld_txq(vi, iter, q) \ + for (q = &vi->pi->adapter->sge.ofld_txq[vi->first_ofld_txq], iter = 0; \ + iter < vi->nofldtxq; ++iter, ++q) +#define for_each_ofld_rxq(vi, iter, q) \ + for (q = &vi->pi->adapter->sge.ofld_rxq[vi->first_ofld_rxq], iter = 0; \ + iter < vi->nofldrxq; ++iter, ++q) +#define for_each_nm_txq(vi, iter, q) \ + for (q = &vi->pi->adapter->sge.nm_txq[vi->first_nm_txq], iter = 0; \ + iter < vi->nnmtxq; ++iter, ++q) +#define for_each_nm_rxq(vi, iter, q) \ + for (q = &vi->pi->adapter->sge.nm_rxq[vi->first_nm_rxq], iter = 0; \ + iter < vi->nnmrxq; ++iter, ++q) +#define for_each_vi(_pi, _iter, _vi) \ + for ((_vi) = (_pi)->vi, (_iter) = 0; (_iter) < (_pi)->nvi; \ + ++(_iter), ++(_vi)) #define IDXINCR(idx, incr, wrap) do { \ idx = wrap - idx > incr ? idx + incr : incr - (wrap - idx); \ @@ -972,7 +975,7 @@ static inline void t4_os_set_hw_addr(struct adapter *sc, int idx, uint8_t hw_addr[]) { - bcopy(hw_addr, sc->port[idx]->hw_addr, ETHER_ADDR_LEN); + bcopy(hw_addr, sc->port[idx]->vi[0].hw_addr, ETHER_ADDR_LEN); } static inline bool @@ -1008,18 +1011,22 @@ int t4_register_cpl_handler(struct adapt int t4_register_an_handler(struct adapter *, an_handler_t); int t4_register_fw_msg_handler(struct adapter *, int, fw_msg_handler_t); int t4_filter_rpl(struct sge_iq *, const struct rss_header *, struct mbuf *); -int begin_synchronized_op(struct adapter *, struct port_info *, int, char *); +int begin_synchronized_op(struct adapter *, struct vi_info *, int, char *); +void doom_vi(struct adapter *, struct vi_info *); void end_synchronized_op(struct adapter *, int); int update_mac_settings(struct ifnet *, int); int adapter_full_init(struct adapter *); int adapter_full_uninit(struct adapter *); -int port_full_init(struct port_info *); -int port_full_uninit(struct port_info *); +uint64_t cxgbe_get_counter(struct ifnet *, ift_counter); +int vi_full_init(struct vi_info *); +int vi_full_uninit(struct vi_info *); +void vi_sysctls(struct vi_info *); +void vi_tick(void *); #ifdef DEV_NETMAP /* t4_netmap.c */ -int create_netmap_ifnet(struct port_info *); -int destroy_netmap_ifnet(struct port_info *); +void cxgbe_nm_attach(struct vi_info *); +void cxgbe_nm_detach(struct vi_info *); void t4_nm_intr(void *); #endif @@ -1036,10 +1043,11 @@ void t4_sge_sysctls(struct adapter *, st int t4_destroy_dma_tag(struct adapter *); int t4_setup_adapter_queues(struct adapter *); int t4_teardown_adapter_queues(struct adapter *); -int t4_setup_port_queues(struct port_info *); -int t4_teardown_port_queues(struct port_info *); +int t4_setup_vi_queues(struct vi_info *); +int t4_teardown_vi_queues(struct vi_info *); void t4_intr_all(void *); void t4_intr(void *); +void t4_vi_intr(void *); void t4_intr_err(void *); void t4_intr_evt(void *); void t4_wrq_tx_locked(struct adapter *, struct sge_wrq *, struct wrqe *); Modified: stable/10/sys/dev/cxgbe/common/t4_hw.c ============================================================================== --- stable/10/sys/dev/cxgbe/common/t4_hw.c Mon Oct 31 22:03:44 2016 (r308153) +++ stable/10/sys/dev/cxgbe/common/t4_hw.c Mon Oct 31 22:45:11 2016 (r308154) @@ -5720,11 +5720,11 @@ int __devinit t4_port_init(struct port_i if (ret < 0) return ret; - p->viid = ret; + p->vi[0].viid = ret; p->tx_chan = j; p->rx_chan_map = get_mps_bg_map(adap, j); p->lport = j; - p->rss_size = rss_size; + p->vi[0].rss_size = rss_size; t4_os_set_hw_addr(adap, p->port_id, addr); ret = ntohl(c.u.info.lstatus_to_modtype); @@ -5737,13 +5737,13 @@ int __devinit t4_port_init(struct port_i param = V_FW_PARAMS_MNEM(FW_PARAMS_MNEM_DEV) | V_FW_PARAMS_PARAM_X(FW_PARAMS_PARAM_DEV_RSSINFO) | - V_FW_PARAMS_PARAM_YZ(p->viid); + V_FW_PARAMS_PARAM_YZ(p->vi[0].viid); ret = t4_query_params(adap, mbox, pf, vf, 1, ¶m, &val); if (ret) - p->rss_base = 0xffff; + p->vi[0].rss_base = 0xffff; else { /* MPASS((val >> 16) == rss_size); */ - p->rss_base = val & 0xffff; + p->vi[0].rss_base = val & 0xffff; } return 0; Modified: stable/10/sys/dev/cxgbe/iw_cxgbe/provider.c ============================================================================== --- stable/10/sys/dev/cxgbe/iw_cxgbe/provider.c Mon Oct 31 22:03:44 2016 (r308153) +++ stable/10/sys/dev/cxgbe/iw_cxgbe/provider.c Mon Oct 31 22:45:11 2016 (r308154) @@ -296,7 +296,7 @@ c4iw_query_gid(struct ib_device *ibdev, if (port == 0 || port > sc->params.nports) return (-EINVAL); pi = sc->port[port - 1]; - memcpy(&gid->raw[0], pi->hw_addr, sizeof(pi->hw_addr)); + memcpy(&gid->raw[0], pi->vi[0].hw_addr, ETHER_ADDR_LEN); return (0); } @@ -309,7 +309,8 @@ c4iw_query_device(struct ib_device *ibde CTR3(KTR_IW_CXGBE, "%s ibdev %p, props %p", __func__, ibdev, props); memset(props, 0, sizeof *props); - memcpy(&props->sys_image_guid, sc->port[0]->hw_addr, 6); + memcpy(&props->sys_image_guid, sc->port[0]->vi[0].hw_addr, + ETHER_ADDR_LEN); props->hw_ver = sc->params.chipid; props->fw_ver = sc->params.fw_vers; props->device_cap_flags = dev->device_cap_flags; @@ -352,7 +353,7 @@ c4iw_query_port(struct ib_device *ibdev, if (port > sc->params.nports) return (-EINVAL); pi = sc->port[port - 1]; - ifp = pi->ifp; + ifp = pi->vi[0].ifp; memset(props, 0, sizeof(struct ib_port_attr)); props->max_mtu = IB_MTU_4096; @@ -397,7 +398,7 @@ c4iw_register_device(struct c4iw_dev *de BUG_ON(!sc->port[0]); strlcpy(ibdev->name, device_get_nameunit(sc->dev), sizeof(ibdev->name)); memset(&ibdev->node_guid, 0, sizeof(ibdev->node_guid)); - memcpy(&ibdev->node_guid, sc->port[0]->hw_addr, 6); + memcpy(&ibdev->node_guid, sc->port[0]->vi[0].hw_addr, ETHER_ADDR_LEN); ibdev->owner = THIS_MODULE; dev->device_cap_flags = IB_DEVICE_LOCAL_DMA_LKEY | IB_DEVICE_MEM_WINDOW; if (fastreg_support) Modified: stable/10/sys/dev/cxgbe/offload.h ============================================================================== --- stable/10/sys/dev/cxgbe/offload.h Mon Oct 31 22:03:44 2016 (r308153) +++ stable/10/sys/dev/cxgbe/offload.h Mon Oct 31 22:45:11 2016 (r308154) @@ -125,7 +125,6 @@ struct t4_virt_res { struct t4_range l2t; }; -#ifdef TCP_OFFLOAD enum { ULD_TOM = 0, ULD_IWARP, @@ -152,6 +151,7 @@ struct tom_tunables { int tx_align; }; +#ifdef TCP_OFFLOAD int t4_register_uld(struct uld_info *); int t4_unregister_uld(struct uld_info *); int t4_activate_uld(struct adapter *, int); Modified: stable/10/sys/dev/cxgbe/t4_main.c ============================================================================== --- stable/10/sys/dev/cxgbe/t4_main.c Mon Oct 31 22:03:44 2016 (r308153) +++ stable/10/sys/dev/cxgbe/t4_main.c Mon Oct 31 22:45:11 2016 (r308154) @@ -107,6 +107,22 @@ static driver_t cxgbe_driver = { sizeof(struct port_info) }; +/* T4 VI (vcxgbe) interface */ +static int vcxgbe_probe(device_t); +static int vcxgbe_attach(device_t); +static int vcxgbe_detach(device_t); +static device_method_t vcxgbe_methods[] = { + DEVMETHOD(device_probe, vcxgbe_probe), + DEVMETHOD(device_attach, vcxgbe_attach), + DEVMETHOD(device_detach, vcxgbe_detach), + { 0, 0 } +}; +static driver_t vcxgbe_driver = { + "vcxgbe", + vcxgbe_methods, + sizeof(struct vi_info) +}; + static d_ioctl_t t4_ioctl; static d_open_t t4_open; static d_close_t t4_close; @@ -143,6 +159,13 @@ static driver_t cxl_driver = { sizeof(struct port_info) }; +/* T5 VI (vcxl) interface */ +static driver_t vcxl_driver = { + "vcxl", + vcxgbe_methods, + sizeof(struct vi_info) +}; + static struct cdevsw t5_cdevsw = { .d_version = D_VERSION, .d_flags = 0, @@ -203,6 +226,14 @@ TUNABLE_INT("hw.cxgbe.ntxq1g", &t4_ntxq1 static int t4_nrxq1g = -1; TUNABLE_INT("hw.cxgbe.nrxq1g", &t4_nrxq1g); +#define NTXQ_VI 1 +static int t4_ntxq_vi = -1; +TUNABLE_INT("hw.cxgbe.ntxq_vi", &t4_ntxq_vi); + +#define NRXQ_VI 1 +static int t4_nrxq_vi = -1; +TUNABLE_INT("hw.cxgbe.nrxq_vi", &t4_nrxq_vi); + static int t4_rsrv_noflowq = 0; TUNABLE_INT("hw.cxgbe.rsrv_noflowq", &t4_rsrv_noflowq); @@ -222,24 +253,24 @@ TUNABLE_INT("hw.cxgbe.nofldtxq1g", &t4_n #define NOFLDRXQ_1G 1 static int t4_nofldrxq1g = -1; TUNABLE_INT("hw.cxgbe.nofldrxq1g", &t4_nofldrxq1g); + +#define NOFLDTXQ_VI 1 +static int t4_nofldtxq_vi = -1; +TUNABLE_INT("hw.cxgbe.nofldtxq_vi", &t4_nofldtxq_vi); + +#define NOFLDRXQ_VI 1 +static int t4_nofldrxq_vi = -1; +TUNABLE_INT("hw.cxgbe.nofldrxq_vi", &t4_nofldrxq_vi); #endif #ifdef DEV_NETMAP -#define NNMTXQ_10G 2 -static int t4_nnmtxq10g = -1; -TUNABLE_INT("hw.cxgbe.nnmtxq10g", &t4_nnmtxq10g); - -#define NNMRXQ_10G 2 -static int t4_nnmrxq10g = -1; -TUNABLE_INT("hw.cxgbe.nnmrxq10g", &t4_nnmrxq10g); - -#define NNMTXQ_1G 1 -static int t4_nnmtxq1g = -1; -TUNABLE_INT("hw.cxgbe.nnmtxq1g", &t4_nnmtxq1g); - -#define NNMRXQ_1G 1 -static int t4_nnmrxq1g = -1; -TUNABLE_INT("hw.cxgbe.nnmrxq1g", &t4_nnmrxq1g); +#define NNMTXQ_VI 2 +static int t4_nnmtxq_vi = -1; +TUNABLE_INT("hw.cxgbe.nnmtxq_vi", &t4_nnmtxq_vi); + +#define NNMRXQ_VI 2 +static int t4_nnmrxq_vi = -1; +TUNABLE_INT("hw.cxgbe.nnmrxq_vi", &t4_nnmrxq_vi); #endif /* @@ -327,6 +358,19 @@ TUNABLE_INT("hw.cxgbe.fcoecaps_allowed", static int t5_write_combine = 0; TUNABLE_INT("hw.cxl.write_combine", &t5_write_combine); +static int t4_num_vis = 1; +TUNABLE_INT("hw.cxgbe.num_vis", &t4_num_vis); + +/* Functions used by extra VIs to obtain unique MAC addresses for each VI. */ +static int vi_mac_funcs[] = { + FW_VI_FUNC_OFLD, + FW_VI_FUNC_IWARP, + FW_VI_FUNC_OPENISCSI, + FW_VI_FUNC_OPENFCOE, + FW_VI_FUNC_FOISCSI, + FW_VI_FUNC_FOFCOE, +}; + struct intrs_and_queues { uint16_t intr_type; /* INTx, MSI, or MSI-X */ uint16_t nirq; /* Total # of vectors */ @@ -337,18 +381,18 @@ struct intrs_and_queues { uint16_t ntxq1g; /* # of NIC txq's for each 1G port */ uint16_t nrxq1g; /* # of NIC rxq's for each 1G port */ uint16_t rsrv_noflowq; /* Flag whether to reserve queue 0 */ -#ifdef TCP_OFFLOAD uint16_t nofldtxq10g; /* # of TOE txq's for each 10G port */ uint16_t nofldrxq10g; /* # of TOE rxq's for each 10G port */ uint16_t nofldtxq1g; /* # of TOE txq's for each 1G port */ uint16_t nofldrxq1g; /* # of TOE rxq's for each 1G port */ -#endif -#ifdef DEV_NETMAP - uint16_t nnmtxq10g; /* # of netmap txq's for each 10G port */ - uint16_t nnmrxq10g; /* # of netmap rxq's for each 10G port */ - uint16_t nnmtxq1g; /* # of netmap txq's for each 1G port */ - uint16_t nnmrxq1g; /* # of netmap rxq's for each 1G port */ -#endif + + /* The vcxgbe/vcxl interfaces use these and not the ones above. */ + uint16_t ntxq_vi; /* # of NIC txq's */ + uint16_t nrxq_vi; /* # of NIC rxq's */ + uint16_t nofldtxq_vi; /* # of TOE txq's */ + uint16_t nofldrxq_vi; /* # of TOE rxq's */ + uint16_t nnmtxq_vi; /* # of netmap txq's */ + uint16_t nnmrxq_vi; /* # of netmap rxq's */ }; struct filter_entry { @@ -370,7 +414,7 @@ static int validate_mt_off_len(struct ad uint32_t *); static void memwin_info(struct adapter *, int, uint32_t *, uint32_t *); static uint32_t position_memwin(struct adapter *, int, uint32_t); -static int cfg_itype_and_nqueues(struct adapter *, int, int, +static int cfg_itype_and_nqueues(struct adapter *, int, int, int, struct intrs_and_queues *); static int prep_firmware(struct adapter *); static int partition_resources(struct adapter *, const struct firmware *, @@ -380,8 +424,8 @@ static int get_params__post_init(struct static int set_params__post_init(struct adapter *); static void t4_set_desc(struct adapter *); static void build_medialist(struct port_info *, struct ifmedia *); -static int cxgbe_init_synchronized(struct port_info *); -static int cxgbe_uninit_synchronized(struct port_info *); +static int cxgbe_init_synchronized(struct vi_info *); +static int cxgbe_uninit_synchronized(struct vi_info *); static int setup_intr_handlers(struct adapter *); static void quiesce_txq(struct adapter *, struct sge_txq *); static void quiesce_wrq(struct adapter *, struct sge_wrq *); @@ -393,6 +437,7 @@ static int t4_free_irq(struct adapter *, static void reg_block_dump(struct adapter *, uint8_t *, unsigned int, unsigned int); static void t4_get_regs(struct adapter *, struct t4_regdump *, uint8_t *); +static void vi_refresh_stats(struct adapter *, struct vi_info *); static void cxgbe_refresh_stats(struct adapter *, struct port_info *); static void cxgbe_tick(void *); static void cxgbe_vlan_config(void *, struct ifnet *, uint16_t); @@ -400,8 +445,8 @@ static int cpl_not_handled(struct sge_iq struct mbuf *); static int an_not_handled(struct sge_iq *, const struct rsp_ctrl *); static int fw_msg_not_handled(struct adapter *, const __be64 *); -static int t4_sysctls(struct adapter *); -static int cxgbe_sysctls(struct port_info *); +static void t4_sysctls(struct adapter *); +static void cxgbe_sysctls(struct port_info *); static int sysctl_int_array(SYSCTL_HANDLER_ARGS); static int sysctl_bitfield(SYSCTL_HANDLER_ARGS); static int sysctl_btphy(SYSCTL_HANDLER_ARGS); @@ -459,7 +504,7 @@ static int read_i2c(struct adapter *, st static int set_sched_class(struct adapter *, struct t4_sched_params *); static int set_sched_queue(struct adapter *, struct t4_sched_queue *); #ifdef TCP_OFFLOAD -static int toe_capability(struct port_info *, int); +static int toe_capability(struct vi_info *, int); #endif static int mod_event(module_t, int, void *); @@ -604,7 +649,7 @@ static int t4_attach(device_t dev) { struct adapter *sc; - int rc = 0, i, n10g, n1g, rqidx, tqidx; + int rc = 0, i, j, n10g, n1g, rqidx, tqidx; struct intrs_and_queues iaq; struct sge *s; #ifdef TCP_OFFLOAD @@ -613,6 +658,7 @@ t4_attach(device_t dev) #ifdef DEV_NETMAP int nm_rqidx, nm_tqidx; #endif + int num_vis; sc = device_get_softc(dev); sc->dev = dev; @@ -646,7 +692,7 @@ t4_attach(device_t dev) mtx_init(&sc->sfl_lock, "starving freelists", 0, MTX_DEF); TAILQ_INIT(&sc->sfl); - callout_init(&sc->sfl_callout, CALLOUT_MPSAFE); + callout_init_mtx(&sc->sfl_callout, &sc->sfl_lock, 0); mtx_init(&sc->regwin_lock, "register and memory window", 0, MTX_DEF); @@ -731,6 +777,24 @@ t4_attach(device_t dev) goto done; /* error message displayed already */ /* + * Number of VIs to create per-port. The first VI is the "main" regular + * VI for the port. The rest are additional virtual interfaces on the + * same physical port. Note that the main VI does not have native + * netmap support but the extra VIs do. + * + * Limit the number of VIs per port to the number of available + * MAC addresses per port. + */ + if (t4_num_vis >= 1) + num_vis = t4_num_vis; + else + num_vis = 1; + if (num_vis > nitems(vi_mac_funcs)) { + num_vis = nitems(vi_mac_funcs); + device_printf(dev, "Number of VIs limited to %d\n", num_vis); + } + + /* * First pass over all the ports - allocate VIs and initialize some * basic parameters like mac address, port type, etc. We also figure * out whether a port is 10G or 1G and use that information when @@ -746,12 +810,22 @@ t4_attach(device_t dev) /* These must be set before t4_port_init */ pi->adapter = sc; pi->port_id = i; + /* + * XXX: vi[0] is special so we can't delay this allocation until + * pi->nvi's final value is known. + */ + pi->vi = malloc(sizeof(struct vi_info) * num_vis, M_CXGBE, + M_ZERO | M_WAITOK); - /* Allocate the vi and initialize parameters like mac addr */ + /* + * Allocate the "main" VI and initialize parameters + * like mac addr. + */ rc = -t4_port_init(pi, sc->mbox, sc->pf, 0); if (rc != 0) { device_printf(dev, "unable to initialize port %d: %d\n", i, rc); + free(pi->vi, M_CXGBE); free(pi, M_CXGBE); sc->port[i] = NULL; goto done; @@ -765,6 +839,7 @@ t4_attach(device_t dev) rc = -t4_link_start(sc, sc->mbox, pi->tx_chan, &pi->link_cfg); if (rc != 0) { device_printf(dev, "port %d l1cfg failed: %d\n", i, rc); + free(pi->vi, M_CXGBE); free(pi, M_CXGBE); sc->port[i] = NULL; goto done; @@ -777,20 +852,12 @@ t4_attach(device_t dev) if (is_10G_port(pi) || is_40G_port(pi)) { n10g++; - pi->tmr_idx = t4_tmr_idx_10g; - pi->pktc_idx = t4_pktc_idx_10g; } else { n1g++; - pi->tmr_idx = t4_tmr_idx_1g; - pi->pktc_idx = t4_pktc_idx_1g; } - pi->xact_addr_filt = -1; pi->linkdnrc = -1; - pi->qsize_rxq = t4_qsize_rxq; - pi->qsize_txq = t4_qsize_txq; - pi->dev = device_add_child(dev, is_t4(sc) ? "cxgbe" : "cxl", -1); if (pi->dev == NULL) { device_printf(dev, @@ -798,15 +865,18 @@ t4_attach(device_t dev) rc = ENXIO; goto done; } + pi->vi[0].dev = pi->dev; device_set_softc(pi->dev, pi); } /* * Interrupt type, # of interrupts, # of rx/tx queues, etc. */ - rc = cfg_itype_and_nqueues(sc, n10g, n1g, &iaq); + rc = cfg_itype_and_nqueues(sc, n10g, n1g, num_vis, &iaq); if (rc != 0) goto done; /* error message displayed already */ + if (iaq.nrxq_vi + iaq.nofldrxq_vi + iaq.nnmrxq_vi == 0) + num_vis = 1; sc->intr_type = iaq.intr_type; sc->intr_count = iaq.nirq; @@ -814,6 +884,10 @@ t4_attach(device_t dev) s = &sc->sge; s->nrxq = n10g * iaq.nrxq10g + n1g * iaq.nrxq1g; s->ntxq = n10g * iaq.ntxq10g + n1g * iaq.ntxq1g; + if (num_vis > 1) { + s->nrxq += (n10g + n1g) * (num_vis - 1) * iaq.nrxq_vi; + s->ntxq += (n10g + n1g) * (num_vis - 1) * iaq.ntxq_vi; + } s->neq = s->ntxq + s->nrxq; /* the free list in an rxq is an eq */ s->neq += sc->params.nports + 1;/* ctrl queues: 1 per port + 1 mgmt */ s->niq = s->nrxq + 1; /* 1 extra for firmware event queue */ @@ -821,6 +895,12 @@ t4_attach(device_t dev) if (is_offload(sc)) { s->nofldrxq = n10g * iaq.nofldrxq10g + n1g * iaq.nofldrxq1g; s->nofldtxq = n10g * iaq.nofldtxq10g + n1g * iaq.nofldtxq1g; + if (num_vis > 1) { + s->nofldrxq += (n10g + n1g) * (num_vis - 1) * + iaq.nofldrxq_vi; + s->nofldtxq += (n10g + n1g) * (num_vis - 1) * + iaq.nofldtxq_vi; + } s->neq += s->nofldtxq + s->nofldrxq; s->niq += s->nofldrxq; @@ -831,8 +911,10 @@ t4_attach(device_t dev) } #endif #ifdef DEV_NETMAP - s->nnmrxq = n10g * iaq.nnmrxq10g + n1g * iaq.nnmrxq1g; - s->nnmtxq = n10g * iaq.nnmtxq10g + n1g * iaq.nnmtxq1g; + if (num_vis > 1) { + s->nnmrxq = (n10g + n1g) * (num_vis - 1) * iaq.nnmrxq_vi; + s->nnmtxq = (n10g + n1g) * (num_vis - 1) * iaq.nnmtxq_vi; + } s->neq += s->nnmtxq + s->nnmrxq; s->niq += s->nnmrxq; @@ -871,57 +953,70 @@ t4_attach(device_t dev) #endif for_each_port(sc, i) { struct port_info *pi = sc->port[i]; + struct vi_info *vi; if (pi == NULL) continue; - pi->first_rxq = rqidx; - pi->first_txq = tqidx; - if (is_10G_port(pi) || is_40G_port(pi)) { - pi->flags |= iaq.intr_flags_10g; - pi->nrxq = iaq.nrxq10g; - pi->ntxq = iaq.ntxq10g; - } else { - pi->flags |= iaq.intr_flags_1g; - pi->nrxq = iaq.nrxq1g; - pi->ntxq = iaq.ntxq1g; - } + pi->nvi = num_vis; + for_each_vi(pi, j, vi) { + vi->pi = pi; + vi->qsize_rxq = t4_qsize_rxq; + vi->qsize_txq = t4_qsize_txq; - if (pi->ntxq > 1) - pi->rsrv_noflowq = iaq.rsrv_noflowq ? 1 : 0; - else - pi->rsrv_noflowq = 0; + vi->first_rxq = rqidx; + vi->first_txq = tqidx; + if (is_10G_port(pi) || is_40G_port(pi)) { + vi->tmr_idx = t4_tmr_idx_10g; + vi->pktc_idx = t4_pktc_idx_10g; + vi->flags |= iaq.intr_flags_10g & INTR_RXQ; + vi->nrxq = j == 0 ? iaq.nrxq10g : iaq.nrxq_vi; + vi->ntxq = j == 0 ? iaq.ntxq10g : iaq.ntxq_vi; + } else { + vi->tmr_idx = t4_tmr_idx_1g; + vi->pktc_idx = t4_pktc_idx_1g; + vi->flags |= iaq.intr_flags_1g & INTR_RXQ; + vi->nrxq = j == 0 ? iaq.nrxq1g : iaq.nrxq_vi; + vi->ntxq = j == 0 ? iaq.ntxq1g : iaq.ntxq_vi; + } + rqidx += vi->nrxq; + tqidx += vi->ntxq; + + if (j == 0 && vi->ntxq > 1) + vi->rsrv_noflowq = iaq.rsrv_noflowq ? 1 : 0; + else + vi->rsrv_noflowq = 0; - rqidx += pi->nrxq; - tqidx += pi->ntxq; #ifdef TCP_OFFLOAD - if (is_offload(sc)) { - pi->first_ofld_rxq = ofld_rqidx; - pi->first_ofld_txq = ofld_tqidx; + vi->first_ofld_rxq = ofld_rqidx; + vi->first_ofld_txq = ofld_tqidx; if (is_10G_port(pi) || is_40G_port(pi)) { - pi->nofldrxq = iaq.nofldrxq10g; - pi->nofldtxq = iaq.nofldtxq10g; + vi->flags |= iaq.intr_flags_10g & INTR_OFLD_RXQ; + vi->nofldrxq = j == 0 ? iaq.nofldrxq10g : + iaq.nofldrxq_vi; + vi->nofldtxq = j == 0 ? iaq.nofldtxq10g : + iaq.nofldtxq_vi; } else { - pi->nofldrxq = iaq.nofldrxq1g; - pi->nofldtxq = iaq.nofldtxq1g; + vi->flags |= iaq.intr_flags_1g & INTR_OFLD_RXQ; + vi->nofldrxq = j == 0 ? iaq.nofldrxq1g : + iaq.nofldrxq_vi; + vi->nofldtxq = j == 0 ? iaq.nofldtxq1g : + iaq.nofldtxq_vi; } - ofld_rqidx += pi->nofldrxq; - ofld_tqidx += pi->nofldtxq; - } + ofld_rqidx += vi->nofldrxq; + ofld_tqidx += vi->nofldtxq; #endif #ifdef DEV_NETMAP - pi->first_nm_rxq = nm_rqidx; - pi->first_nm_txq = nm_tqidx; - if (is_10G_port(pi) || is_40G_port(pi)) { - pi->nnmrxq = iaq.nnmrxq10g; - pi->nnmtxq = iaq.nnmtxq10g; - } else { - pi->nnmrxq = iaq.nnmrxq1g; - pi->nnmtxq = iaq.nnmtxq1g; - } - nm_rqidx += pi->nnmrxq; - nm_tqidx += pi->nnmtxq; + if (j > 0) { + vi->first_nm_rxq = nm_rqidx; + vi->first_nm_txq = nm_tqidx; + vi->nnmrxq = iaq.nnmrxq_vi; + vi->nnmtxq = iaq.nnmtxq_vi; + nm_rqidx += vi->nnmrxq; + nm_tqidx += vi->nnmtxq; + } #endif + } } rc = setup_intr_handlers(sc); @@ -996,11 +1091,12 @@ t4_detach(device_t dev) for (i = 0; i < MAX_NPORTS; i++) { pi = sc->port[i]; if (pi) { - t4_free_vi(sc, sc->mbox, sc->pf, 0, pi->viid); + t4_free_vi(sc, sc->mbox, sc->pf, 0, pi->vi[0].viid); if (pi->dev) device_delete_child(dev, pi->dev); mtx_destroy(&pi->pi_lock); + free(pi->vi, M_CXGBE); free(pi, M_CXGBE); } } @@ -1052,6 +1148,7 @@ t4_detach(device_t dev) mtx_destroy(&sc->sc_lock); } + callout_drain(&sc->sfl_callout); if (mtx_initialized(&sc->tids.ftid_lock)) mtx_destroy(&sc->tids.ftid_lock); if (mtx_initialized(&sc->sfl_lock)) @@ -1084,12 +1181,13 @@ cxgbe_probe(device_t dev) #define T4_CAP_ENABLE (T4_CAP) static int -cxgbe_attach(device_t dev) +cxgbe_vi_attach(device_t dev, struct vi_info *vi) { - struct port_info *pi = device_get_softc(dev); struct ifnet *ifp; - char *s; - int n, o; + struct sbuf *sb; + + vi->xact_addr_filt = -1; + callout_init(&vi->tick, 1); /* Allocate an ifnet and set it up */ ifp = if_alloc(IFT_ETHER); @@ -1097,10 +1195,8 @@ cxgbe_attach(device_t dev) device_printf(dev, "Cannot allocate ifnet\n"); *** DIFF OUTPUT TRUNCATED AT 1000 LINES ***
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?201610312245.u9VMjCTB019847>