1 /*- 2 * Copyright (c) 2014-2018, Matthew Macy <mmacy@mattmacy.io> 3 * All rights reserved. 4 * 5 * Redistribution and use in source and binary forms, with or without 6 * modification, are permitted provided that the following conditions are met: 7 * 8 * 1. Redistributions of source code must retain the above copyright notice, 9 * this list of conditions and the following disclaimer. 10 * 11 * 2. Neither the name of Matthew Macy nor the names of its 12 * contributors may be used to endorse or promote products derived from 13 * this software without specific prior written permission. 14 * 15 * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" 16 * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 17 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 18 * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE 19 * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR 20 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF 21 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS 22 * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN 23 * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) 24 * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE 25 * POSSIBILITY OF SUCH DAMAGE. 26 */ 27 28 #include <sys/cdefs.h> 29 #include "opt_inet.h" 30 #include "opt_inet6.h" 31 #include "opt_acpi.h" 32 #include "opt_sched.h" 33 34 #include <sys/param.h> 35 #include <sys/types.h> 36 #include <sys/bus.h> 37 #include <sys/eventhandler.h> 38 #include <sys/kernel.h> 39 #include <sys/lock.h> 40 #include <sys/mutex.h> 41 #include <sys/module.h> 42 #include <sys/kobj.h> 43 #include <sys/rman.h> 44 #include <sys/sbuf.h> 45 #include <sys/smp.h> 46 #include <sys/socket.h> 47 #include <sys/sockio.h> 48 #include <sys/sysctl.h> 49 #include <sys/syslog.h> 50 #include <sys/taskqueue.h> 51 #include <sys/limits.h> 52 53 #include <net/if.h> 54 #include <net/if_var.h> 55 #include <net/if_private.h> 56 #include <net/if_types.h> 57 #include <net/if_media.h> 58 #include <net/bpf.h> 59 #include <net/ethernet.h> 60 #include <net/mp_ring.h> 61 #include <net/debugnet.h> 62 #include <net/pfil.h> 63 #include <net/vnet.h> 64 65 #include <netinet/in.h> 66 #include <netinet/in_pcb.h> 67 #include <netinet/tcp_lro.h> 68 #include <netinet/in_systm.h> 69 #include <netinet/if_ether.h> 70 #include <netinet/ip.h> 71 #include <netinet/ip6.h> 72 #include <netinet/tcp.h> 73 #include <netinet/ip_var.h> 74 #include <netinet6/ip6_var.h> 75 76 #include <machine/bus.h> 77 #include <machine/in_cksum.h> 78 79 #include <vm/vm.h> 80 #include <vm/pmap.h> 81 82 #include <dev/led/led.h> 83 #include <dev/pci/pcireg.h> 84 #include <dev/pci/pcivar.h> 85 #include <dev/pci/pci_private.h> 86 87 #include <net/iflib.h> 88 89 #include "ifdi_if.h" 90 91 #ifdef PCI_IOV 92 #include <dev/pci/pci_iov.h> 93 #endif 94 95 #include <sys/bitstring.h> 96 /* 97 * enable accounting of every mbuf as it comes in to and goes out of 98 * iflib's software descriptor references 99 */ 100 #define MEMORY_LOGGING 0 101 /* 102 * Enable mbuf vectors for compressing long mbuf chains 103 */ 104 105 /* 106 * NB: 107 * - Prefetching in tx cleaning should perhaps be a tunable. The distance ahead 108 * we prefetch needs to be determined by the time spent in m_free vis a vis 109 * the cost of a prefetch. This will of course vary based on the workload: 110 * - NFLX's m_free path is dominated by vm-based M_EXT manipulation which 111 * is quite expensive, thus suggesting very little prefetch. 112 * - small packet forwarding which is just returning a single mbuf to 113 * UMA will typically be very fast vis a vis the cost of a memory 114 * access. 115 */ 116 117 /* 118 * File organization: 119 * - private structures 120 * - iflib private utility functions 121 * - ifnet functions 122 * - vlan registry and other exported functions 123 * - iflib public core functions 124 * 125 * 126 */ 127 static MALLOC_DEFINE(M_IFLIB, "iflib", "ifnet library"); 128 129 #define IFLIB_RXEOF_MORE (1U << 0) 130 #define IFLIB_RXEOF_EMPTY (2U << 0) 131 132 struct iflib_txq; 133 typedef struct iflib_txq *iflib_txq_t; 134 struct iflib_rxq; 135 typedef struct iflib_rxq *iflib_rxq_t; 136 struct iflib_fl; 137 typedef struct iflib_fl *iflib_fl_t; 138 139 struct iflib_ctx; 140 141 static void iru_init(if_rxd_update_t iru, iflib_rxq_t rxq, uint8_t flid); 142 static void iflib_timer(void *arg); 143 static void iflib_tqg_detach(if_ctx_t ctx); 144 145 typedef struct iflib_filter_info { 146 driver_filter_t *ifi_filter; 147 void *ifi_filter_arg; 148 struct grouptask *ifi_task; 149 void *ifi_ctx; 150 } *iflib_filter_info_t; 151 152 struct iflib_ctx { 153 KOBJ_FIELDS; 154 /* 155 * Pointer to hardware driver's softc 156 */ 157 void *ifc_softc; 158 device_t ifc_dev; 159 if_t ifc_ifp; 160 161 cpuset_t ifc_cpus; 162 if_shared_ctx_t ifc_sctx; 163 struct if_softc_ctx ifc_softc_ctx; 164 165 struct sx ifc_ctx_sx; 166 struct mtx ifc_state_mtx; 167 168 iflib_txq_t ifc_txqs; 169 iflib_rxq_t ifc_rxqs; 170 uint32_t ifc_if_flags; 171 uint32_t ifc_flags; 172 uint32_t ifc_max_fl_buf_size; 173 uint32_t ifc_rx_mbuf_sz; 174 175 int ifc_link_state; 176 int ifc_watchdog_events; 177 struct cdev *ifc_led_dev; 178 struct resource *ifc_msix_mem; 179 180 struct if_irq ifc_legacy_irq; 181 struct grouptask ifc_admin_task; 182 struct grouptask ifc_vflr_task; 183 struct iflib_filter_info ifc_filter_info; 184 struct ifmedia ifc_media; 185 struct ifmedia *ifc_mediap; 186 187 struct sysctl_oid *ifc_sysctl_node; 188 uint16_t ifc_sysctl_ntxqs; 189 uint16_t ifc_sysctl_nrxqs; 190 uint16_t ifc_sysctl_qs_eq_override; 191 uint16_t ifc_sysctl_rx_budget; 192 uint16_t ifc_sysctl_tx_abdicate; 193 uint16_t ifc_sysctl_core_offset; 194 #define CORE_OFFSET_UNSPECIFIED 0xffff 195 uint8_t ifc_sysctl_separate_txrx; 196 uint8_t ifc_sysctl_use_logical_cores; 197 uint16_t ifc_sysctl_extra_msix_vectors; 198 bool ifc_cpus_are_physical_cores; 199 200 qidx_t ifc_sysctl_ntxds[8]; 201 qidx_t ifc_sysctl_nrxds[8]; 202 struct if_txrx ifc_txrx; 203 #define isc_txd_encap ifc_txrx.ift_txd_encap 204 #define isc_txd_flush ifc_txrx.ift_txd_flush 205 #define isc_txd_credits_update ifc_txrx.ift_txd_credits_update 206 #define isc_rxd_available ifc_txrx.ift_rxd_available 207 #define isc_rxd_pkt_get ifc_txrx.ift_rxd_pkt_get 208 #define isc_rxd_refill ifc_txrx.ift_rxd_refill 209 #define isc_rxd_flush ifc_txrx.ift_rxd_flush 210 #define isc_legacy_intr ifc_txrx.ift_legacy_intr 211 #define isc_txq_select ifc_txrx.ift_txq_select 212 #define isc_txq_select_v2 ifc_txrx.ift_txq_select_v2 213 214 eventhandler_tag ifc_vlan_attach_event; 215 eventhandler_tag ifc_vlan_detach_event; 216 struct ether_addr ifc_mac; 217 }; 218 219 void * 220 iflib_get_softc(if_ctx_t ctx) 221 { 222 223 return (ctx->ifc_softc); 224 } 225 226 device_t 227 iflib_get_dev(if_ctx_t ctx) 228 { 229 230 return (ctx->ifc_dev); 231 } 232 233 if_t 234 iflib_get_ifp(if_ctx_t ctx) 235 { 236 237 return (ctx->ifc_ifp); 238 } 239 240 struct ifmedia * 241 iflib_get_media(if_ctx_t ctx) 242 { 243 244 return (ctx->ifc_mediap); 245 } 246 247 void 248 iflib_set_mac(if_ctx_t ctx, uint8_t mac[ETHER_ADDR_LEN]) 249 { 250 251 bcopy(mac, ctx->ifc_mac.octet, ETHER_ADDR_LEN); 252 } 253 254 if_softc_ctx_t 255 iflib_get_softc_ctx(if_ctx_t ctx) 256 { 257 258 return (&ctx->ifc_softc_ctx); 259 } 260 261 if_shared_ctx_t 262 iflib_get_sctx(if_ctx_t ctx) 263 { 264 265 return (ctx->ifc_sctx); 266 } 267 268 uint16_t 269 iflib_get_extra_msix_vectors_sysctl(if_ctx_t ctx) 270 { 271 272 return (ctx->ifc_sysctl_extra_msix_vectors); 273 } 274 275 #define IP_ALIGNED(m) ((((uintptr_t)(m)->m_data) & 0x3) == 0x2) 276 #define CACHE_PTR_INCREMENT (CACHE_LINE_SIZE / sizeof(void *)) 277 #define CACHE_PTR_NEXT(ptr) ((void *)(((uintptr_t)(ptr) + CACHE_LINE_SIZE - 1) & (CACHE_LINE_SIZE - 1))) 278 279 #define LINK_ACTIVE(ctx) ((ctx)->ifc_link_state == LINK_STATE_UP) 280 #define CTX_IS_VF(ctx) ((ctx)->ifc_sctx->isc_flags & IFLIB_IS_VF) 281 282 typedef struct iflib_sw_rx_desc_array { 283 bus_dmamap_t *ifsd_map; /* bus_dma maps for packet */ 284 struct mbuf **ifsd_m; /* pkthdr mbufs */ 285 caddr_t *ifsd_cl; /* direct cluster pointer for rx */ 286 bus_addr_t *ifsd_ba; /* bus addr of cluster for rx */ 287 } iflib_rxsd_array_t; 288 289 typedef struct iflib_sw_tx_desc_array { 290 bus_dmamap_t *ifsd_map; /* bus_dma maps for packet */ 291 bus_dmamap_t *ifsd_tso_map; /* bus_dma maps for TSO packet */ 292 struct mbuf **ifsd_m; /* pkthdr mbufs */ 293 } if_txsd_vec_t; 294 295 /* magic number that should be high enough for any hardware */ 296 #define IFLIB_MAX_TX_SEGS 128 297 #define IFLIB_RX_COPY_THRESH 128 298 #define IFLIB_MAX_RX_REFRESH 32 299 /* The minimum descriptors per second before we start coalescing */ 300 #define IFLIB_MIN_DESC_SEC 16384 301 #define IFLIB_DEFAULT_TX_UPDATE_FREQ 16 302 #define IFLIB_QUEUE_IDLE 0 303 #define IFLIB_QUEUE_HUNG 1 304 #define IFLIB_QUEUE_WORKING 2 305 /* maximum number of txqs that can share an rx interrupt */ 306 #define IFLIB_MAX_TX_SHARED_INTR 4 307 308 /* this should really scale with ring size - this is a fairly arbitrary value */ 309 #define TX_BATCH_SIZE 32 310 311 #define IFLIB_RESTART_BUDGET 8 312 313 #define IFC_LEGACY 0x001 314 #define IFC_QFLUSH 0x002 315 #define IFC_MULTISEG 0x004 316 #define IFC_SPARE1 0x008 317 #define IFC_SC_ALLOCATED 0x010 318 #define IFC_INIT_DONE 0x020 319 #define IFC_PREFETCH 0x040 320 #define IFC_DO_RESET 0x080 321 #define IFC_DO_WATCHDOG 0x100 322 #define IFC_SPARE0 0x200 323 #define IFC_SPARE2 0x400 324 #define IFC_IN_DETACH 0x800 325 326 #define IFC_NETMAP_TX_IRQ 0x80000000 327 328 #define CSUM_OFFLOAD (CSUM_IP_TSO | CSUM_IP6_TSO | CSUM_IP | \ 329 CSUM_IP_UDP | CSUM_IP_TCP | CSUM_IP_SCTP | \ 330 CSUM_IP6_UDP | CSUM_IP6_TCP | CSUM_IP6_SCTP) 331 332 struct iflib_txq { 333 qidx_t ift_in_use; 334 qidx_t ift_cidx; 335 qidx_t ift_cidx_processed; 336 qidx_t ift_pidx; 337 uint8_t ift_gen; 338 uint8_t ift_br_offset; 339 uint16_t ift_npending; 340 uint16_t ift_db_pending; 341 uint16_t ift_rs_pending; 342 /* implicit pad */ 343 uint8_t ift_txd_size[8]; 344 uint64_t ift_processed; 345 uint64_t ift_cleaned; 346 uint64_t ift_cleaned_prev; 347 #if MEMORY_LOGGING 348 uint64_t ift_enqueued; 349 uint64_t ift_dequeued; 350 #endif 351 uint64_t ift_no_tx_dma_setup; 352 uint64_t ift_no_desc_avail; 353 uint64_t ift_mbuf_defrag_failed; 354 uint64_t ift_mbuf_defrag; 355 uint64_t ift_map_failed; 356 uint64_t ift_txd_encap_efbig; 357 uint64_t ift_pullups; 358 uint64_t ift_last_timer_tick; 359 360 struct mtx ift_mtx; 361 struct mtx ift_db_mtx; 362 363 /* constant values */ 364 if_ctx_t ift_ctx; 365 struct ifmp_ring *ift_br; 366 struct grouptask ift_task; 367 qidx_t ift_size; 368 uint16_t ift_id; 369 struct callout ift_timer; 370 #ifdef DEV_NETMAP 371 struct callout ift_netmap_timer; 372 #endif /* DEV_NETMAP */ 373 374 if_txsd_vec_t ift_sds; 375 uint8_t ift_qstatus; 376 uint8_t ift_closed; 377 uint8_t ift_update_freq; 378 struct iflib_filter_info ift_filter_info; 379 bus_dma_tag_t ift_buf_tag; 380 bus_dma_tag_t ift_tso_buf_tag; 381 iflib_dma_info_t ift_ifdi; 382 #define MTX_NAME_LEN 32 383 char ift_mtx_name[MTX_NAME_LEN]; 384 bus_dma_segment_t ift_segs[IFLIB_MAX_TX_SEGS] __aligned(CACHE_LINE_SIZE); 385 #ifdef IFLIB_DIAGNOSTICS 386 uint64_t ift_cpu_exec_count[256]; 387 #endif 388 } __aligned(CACHE_LINE_SIZE); 389 390 struct iflib_fl { 391 qidx_t ifl_cidx; 392 qidx_t ifl_pidx; 393 qidx_t ifl_credits; 394 uint8_t ifl_gen; 395 uint8_t ifl_rxd_size; 396 #if MEMORY_LOGGING 397 uint64_t ifl_m_enqueued; 398 uint64_t ifl_m_dequeued; 399 uint64_t ifl_cl_enqueued; 400 uint64_t ifl_cl_dequeued; 401 #endif 402 /* implicit pad */ 403 bitstr_t *ifl_rx_bitmap; 404 qidx_t ifl_fragidx; 405 /* constant */ 406 qidx_t ifl_size; 407 uint16_t ifl_buf_size; 408 uint16_t ifl_cltype; 409 uma_zone_t ifl_zone; 410 iflib_rxsd_array_t ifl_sds; 411 iflib_rxq_t ifl_rxq; 412 uint8_t ifl_id; 413 bus_dma_tag_t ifl_buf_tag; 414 iflib_dma_info_t ifl_ifdi; 415 uint64_t ifl_bus_addrs[IFLIB_MAX_RX_REFRESH] __aligned(CACHE_LINE_SIZE); 416 qidx_t ifl_rxd_idxs[IFLIB_MAX_RX_REFRESH]; 417 } __aligned(CACHE_LINE_SIZE); 418 419 static inline qidx_t 420 get_inuse(int size, qidx_t cidx, qidx_t pidx, uint8_t gen) 421 { 422 qidx_t used; 423 424 if (pidx > cidx) 425 used = pidx - cidx; 426 else if (pidx < cidx) 427 used = size - cidx + pidx; 428 else if (gen == 0 && pidx == cidx) 429 used = 0; 430 else if (gen == 1 && pidx == cidx) 431 used = size; 432 else 433 panic("bad state"); 434 435 return (used); 436 } 437 438 #define TXQ_AVAIL(txq) (txq->ift_size - get_inuse(txq->ift_size, txq->ift_cidx, txq->ift_pidx, txq->ift_gen)) 439 440 #define IDXDIFF(head, tail, wrap) \ 441 ((head) >= (tail) ? (head) - (tail) : (wrap) - (tail) + (head)) 442 443 struct iflib_rxq { 444 if_ctx_t ifr_ctx; 445 iflib_fl_t ifr_fl; 446 uint64_t ifr_rx_irq; 447 struct pfil_head *pfil; 448 /* 449 * If there is a separate completion queue (IFLIB_HAS_RXCQ), this is 450 * the completion queue consumer index. Otherwise it's unused. 451 */ 452 qidx_t ifr_cq_cidx; 453 uint16_t ifr_id; 454 uint8_t ifr_nfl; 455 uint8_t ifr_ntxqirq; 456 uint8_t ifr_txqid[IFLIB_MAX_TX_SHARED_INTR]; 457 uint8_t ifr_fl_offset; 458 struct lro_ctrl ifr_lc; 459 struct grouptask ifr_task; 460 struct callout ifr_watchdog; 461 struct iflib_filter_info ifr_filter_info; 462 iflib_dma_info_t ifr_ifdi; 463 464 /* dynamically allocate if any drivers need a value substantially larger than this */ 465 struct if_rxd_frag ifr_frags[IFLIB_MAX_RX_SEGS] __aligned(CACHE_LINE_SIZE); 466 #ifdef IFLIB_DIAGNOSTICS 467 uint64_t ifr_cpu_exec_count[256]; 468 #endif 469 } __aligned(CACHE_LINE_SIZE); 470 471 typedef struct if_rxsd { 472 caddr_t *ifsd_cl; 473 iflib_fl_t ifsd_fl; 474 } *if_rxsd_t; 475 476 /* multiple of word size */ 477 #ifdef __LP64__ 478 #define PKT_INFO_SIZE 6 479 #define RXD_INFO_SIZE 5 480 #define PKT_TYPE uint64_t 481 #else 482 #define PKT_INFO_SIZE 11 483 #define RXD_INFO_SIZE 8 484 #define PKT_TYPE uint32_t 485 #endif 486 #define PKT_LOOP_BOUND ((PKT_INFO_SIZE / 3) * 3) 487 #define RXD_LOOP_BOUND ((RXD_INFO_SIZE / 4) * 4) 488 489 typedef struct if_pkt_info_pad { 490 PKT_TYPE pkt_val[PKT_INFO_SIZE]; 491 } *if_pkt_info_pad_t; 492 typedef struct if_rxd_info_pad { 493 PKT_TYPE rxd_val[RXD_INFO_SIZE]; 494 } *if_rxd_info_pad_t; 495 496 CTASSERT(sizeof(struct if_pkt_info_pad) == sizeof(struct if_pkt_info)); 497 CTASSERT(sizeof(struct if_rxd_info_pad) == sizeof(struct if_rxd_info)); 498 499 static inline void 500 pkt_info_zero(if_pkt_info_t pi) 501 { 502 if_pkt_info_pad_t pi_pad; 503 504 pi_pad = (if_pkt_info_pad_t)pi; 505 pi_pad->pkt_val[0] = 0; pi_pad->pkt_val[1] = 0; pi_pad->pkt_val[2] = 0; 506 pi_pad->pkt_val[3] = 0; pi_pad->pkt_val[4] = 0; pi_pad->pkt_val[5] = 0; 507 #ifndef __LP64__ 508 pi_pad->pkt_val[6] = 0; pi_pad->pkt_val[7] = 0; pi_pad->pkt_val[8] = 0; 509 pi_pad->pkt_val[9] = 0; pi_pad->pkt_val[10] = 0; 510 #endif 511 } 512 513 static inline void 514 rxd_info_zero(if_rxd_info_t ri) 515 { 516 if_rxd_info_pad_t ri_pad; 517 int i; 518 519 ri_pad = (if_rxd_info_pad_t)ri; 520 for (i = 0; i < RXD_LOOP_BOUND; i += 4) { 521 ri_pad->rxd_val[i] = 0; 522 ri_pad->rxd_val[i + 1] = 0; 523 ri_pad->rxd_val[i + 2] = 0; 524 ri_pad->rxd_val[i + 3] = 0; 525 } 526 #ifdef __LP64__ 527 ri_pad->rxd_val[RXD_INFO_SIZE - 1] = 0; 528 #endif 529 } 530 531 /* 532 * Only allow a single packet to take up most 1/nth of the tx ring 533 */ 534 #define MAX_SINGLE_PACKET_FRACTION 12 535 #define IF_BAD_DMA ((bus_addr_t)-1) 536 537 #define CTX_ACTIVE(ctx) ((if_getdrvflags((ctx)->ifc_ifp) & IFF_DRV_RUNNING)) 538 539 #define CTX_LOCK_INIT(_sc) sx_init(&(_sc)->ifc_ctx_sx, "iflib ctx lock") 540 #define CTX_LOCK(ctx) sx_xlock(&(ctx)->ifc_ctx_sx) 541 #define CTX_UNLOCK(ctx) sx_xunlock(&(ctx)->ifc_ctx_sx) 542 #define CTX_LOCK_DESTROY(ctx) sx_destroy(&(ctx)->ifc_ctx_sx) 543 544 #define STATE_LOCK_INIT(_sc, _name) mtx_init(&(_sc)->ifc_state_mtx, _name, "iflib state lock", MTX_DEF) 545 #define STATE_LOCK(ctx) mtx_lock(&(ctx)->ifc_state_mtx) 546 #define STATE_UNLOCK(ctx) mtx_unlock(&(ctx)->ifc_state_mtx) 547 #define STATE_LOCK_DESTROY(ctx) mtx_destroy(&(ctx)->ifc_state_mtx) 548 549 #define CALLOUT_LOCK(txq) mtx_lock(&txq->ift_mtx) 550 #define CALLOUT_UNLOCK(txq) mtx_unlock(&txq->ift_mtx) 551 552 /* Our boot-time initialization hook */ 553 static int iflib_module_event_handler(module_t, int, void *); 554 555 static moduledata_t iflib_moduledata = { 556 "iflib", 557 iflib_module_event_handler, 558 NULL 559 }; 560 561 DECLARE_MODULE(iflib, iflib_moduledata, SI_SUB_INIT_IF, SI_ORDER_ANY); 562 MODULE_VERSION(iflib, 1); 563 564 MODULE_DEPEND(iflib, pci, 1, 1, 1); 565 MODULE_DEPEND(iflib, ether, 1, 1, 1); 566 567 TASKQGROUP_DEFINE(if_io_tqg, mp_ncpus, 1); 568 TASKQGROUP_DEFINE(if_config_tqg, 1, 1); 569 570 #ifndef IFLIB_DEBUG_COUNTERS 571 #ifdef INVARIANTS 572 #define IFLIB_DEBUG_COUNTERS 1 573 #else 574 #define IFLIB_DEBUG_COUNTERS 0 575 #endif /* !INVARIANTS */ 576 #endif 577 578 static SYSCTL_NODE(_net, OID_AUTO, iflib, CTLFLAG_RD | CTLFLAG_MPSAFE, 0, 579 "iflib driver parameters"); 580 581 /* 582 * XXX need to ensure that this can't accidentally cause the head to be moved backwards 583 */ 584 static int iflib_min_tx_latency = 0; 585 SYSCTL_INT(_net_iflib, OID_AUTO, min_tx_latency, CTLFLAG_RW, 586 &iflib_min_tx_latency, 0, 587 "minimize transmit latency at the possible expense of throughput"); 588 static int iflib_no_tx_batch = 0; 589 SYSCTL_INT(_net_iflib, OID_AUTO, no_tx_batch, CTLFLAG_RW, 590 &iflib_no_tx_batch, 0, 591 "minimize transmit latency at the possible expense of throughput"); 592 static int iflib_timer_default = 1000; 593 SYSCTL_INT(_net_iflib, OID_AUTO, timer_default, CTLFLAG_RW, 594 &iflib_timer_default, 0, "number of ticks between iflib_timer calls"); 595 596 597 #if IFLIB_DEBUG_COUNTERS 598 599 static int iflib_tx_seen; 600 static int iflib_tx_sent; 601 static int iflib_tx_encap; 602 static int iflib_rx_allocs; 603 static int iflib_fl_refills; 604 static int iflib_fl_refills_large; 605 static int iflib_tx_frees; 606 607 SYSCTL_INT(_net_iflib, OID_AUTO, tx_seen, CTLFLAG_RD, &iflib_tx_seen, 0, 608 "# TX mbufs seen"); 609 SYSCTL_INT(_net_iflib, OID_AUTO, tx_sent, CTLFLAG_RD, &iflib_tx_sent, 0, 610 "# TX mbufs sent"); 611 SYSCTL_INT(_net_iflib, OID_AUTO, tx_encap, CTLFLAG_RD, &iflib_tx_encap, 0, 612 "# TX mbufs encapped"); 613 SYSCTL_INT(_net_iflib, OID_AUTO, tx_frees, CTLFLAG_RD, &iflib_tx_frees, 0, 614 "# TX frees"); 615 SYSCTL_INT(_net_iflib, OID_AUTO, rx_allocs, CTLFLAG_RD, &iflib_rx_allocs, 0, 616 "# RX allocations"); 617 SYSCTL_INT(_net_iflib, OID_AUTO, fl_refills, CTLFLAG_RD, &iflib_fl_refills, 0, 618 "# refills"); 619 SYSCTL_INT(_net_iflib, OID_AUTO, fl_refills_large, CTLFLAG_RD, 620 &iflib_fl_refills_large, 0, "# large refills"); 621 622 static int iflib_txq_drain_flushing; 623 static int iflib_txq_drain_oactive; 624 static int iflib_txq_drain_notready; 625 626 SYSCTL_INT(_net_iflib, OID_AUTO, txq_drain_flushing, CTLFLAG_RD, 627 &iflib_txq_drain_flushing, 0, "# drain flushes"); 628 SYSCTL_INT(_net_iflib, OID_AUTO, txq_drain_oactive, CTLFLAG_RD, 629 &iflib_txq_drain_oactive, 0, "# drain oactives"); 630 SYSCTL_INT(_net_iflib, OID_AUTO, txq_drain_notready, CTLFLAG_RD, 631 &iflib_txq_drain_notready, 0, "# drain notready"); 632 633 static int iflib_encap_load_mbuf_fail; 634 static int iflib_encap_pad_mbuf_fail; 635 static int iflib_encap_txq_avail_fail; 636 static int iflib_encap_txd_encap_fail; 637 638 SYSCTL_INT(_net_iflib, OID_AUTO, encap_load_mbuf_fail, CTLFLAG_RD, 639 &iflib_encap_load_mbuf_fail, 0, "# busdma load failures"); 640 SYSCTL_INT(_net_iflib, OID_AUTO, encap_pad_mbuf_fail, CTLFLAG_RD, 641 &iflib_encap_pad_mbuf_fail, 0, "# runt frame pad failures"); 642 SYSCTL_INT(_net_iflib, OID_AUTO, encap_txq_avail_fail, CTLFLAG_RD, 643 &iflib_encap_txq_avail_fail, 0, "# txq avail failures"); 644 SYSCTL_INT(_net_iflib, OID_AUTO, encap_txd_encap_fail, CTLFLAG_RD, 645 &iflib_encap_txd_encap_fail, 0, "# driver encap failures"); 646 647 static int iflib_task_fn_rxs; 648 static int iflib_rx_intr_enables; 649 static int iflib_fast_intrs; 650 static int iflib_rx_unavail; 651 static int iflib_rx_ctx_inactive; 652 static int iflib_rx_if_input; 653 static int iflib_rxd_flush; 654 655 static int iflib_verbose_debug; 656 657 SYSCTL_INT(_net_iflib, OID_AUTO, task_fn_rx, CTLFLAG_RD, &iflib_task_fn_rxs, 0, 658 "# task_fn_rx calls"); 659 SYSCTL_INT(_net_iflib, OID_AUTO, rx_intr_enables, CTLFLAG_RD, 660 &iflib_rx_intr_enables, 0, "# RX intr enables"); 661 SYSCTL_INT(_net_iflib, OID_AUTO, fast_intrs, CTLFLAG_RD, &iflib_fast_intrs, 0, 662 "# fast_intr calls"); 663 SYSCTL_INT(_net_iflib, OID_AUTO, rx_unavail, CTLFLAG_RD, &iflib_rx_unavail, 0, 664 "# times rxeof called with no available data"); 665 SYSCTL_INT(_net_iflib, OID_AUTO, rx_ctx_inactive, CTLFLAG_RD, 666 &iflib_rx_ctx_inactive, 0, "# times rxeof called with inactive context"); 667 SYSCTL_INT(_net_iflib, OID_AUTO, rx_if_input, CTLFLAG_RD, &iflib_rx_if_input, 668 0, "# times rxeof called if_input"); 669 SYSCTL_INT(_net_iflib, OID_AUTO, rxd_flush, CTLFLAG_RD, &iflib_rxd_flush, 0, 670 "# times rxd_flush called"); 671 SYSCTL_INT(_net_iflib, OID_AUTO, verbose_debug, CTLFLAG_RW, 672 &iflib_verbose_debug, 0, "enable verbose debugging"); 673 674 #define DBG_COUNTER_INC(name) atomic_add_int(&(iflib_ ## name), 1) 675 static void 676 iflib_debug_reset(void) 677 { 678 iflib_tx_seen = iflib_tx_sent = iflib_tx_encap = iflib_rx_allocs = 679 iflib_fl_refills = iflib_fl_refills_large = iflib_tx_frees = 680 iflib_txq_drain_flushing = iflib_txq_drain_oactive = 681 iflib_txq_drain_notready = 682 iflib_encap_load_mbuf_fail = iflib_encap_pad_mbuf_fail = 683 iflib_encap_txq_avail_fail = iflib_encap_txd_encap_fail = 684 iflib_task_fn_rxs = iflib_rx_intr_enables = iflib_fast_intrs = 685 iflib_rx_unavail = 686 iflib_rx_ctx_inactive = iflib_rx_if_input = 687 iflib_rxd_flush = 0; 688 } 689 690 #else 691 #define DBG_COUNTER_INC(name) 692 static void iflib_debug_reset(void) {} 693 #endif 694 695 #define IFLIB_DEBUG 0 696 697 static void iflib_tx_structures_free(if_ctx_t ctx); 698 static void iflib_rx_structures_free(if_ctx_t ctx); 699 static int iflib_queues_alloc(if_ctx_t ctx); 700 static int iflib_tx_credits_update(if_ctx_t ctx, iflib_txq_t txq); 701 static int iflib_rxd_avail(if_ctx_t ctx, iflib_rxq_t rxq, qidx_t cidx, qidx_t budget); 702 static int iflib_qset_structures_setup(if_ctx_t ctx); 703 static int iflib_msix_init(if_ctx_t ctx); 704 static int iflib_legacy_setup(if_ctx_t ctx, driver_filter_t filter, void *filterarg, int *rid, const char *str); 705 static void iflib_txq_check_drain(iflib_txq_t txq, int budget); 706 static uint32_t iflib_txq_can_drain(struct ifmp_ring *); 707 #ifdef ALTQ 708 static void iflib_altq_if_start(if_t ifp); 709 static int iflib_altq_if_transmit(if_t ifp, struct mbuf *m); 710 #endif 711 static int iflib_register(if_ctx_t); 712 static void iflib_deregister(if_ctx_t); 713 static void iflib_unregister_vlan_handlers(if_ctx_t ctx); 714 static uint16_t iflib_get_mbuf_size_for(unsigned int size); 715 static void iflib_init_locked(if_ctx_t ctx); 716 static void iflib_add_device_sysctl_pre(if_ctx_t ctx); 717 static void iflib_add_device_sysctl_post(if_ctx_t ctx); 718 static void iflib_ifmp_purge(iflib_txq_t txq); 719 static void _iflib_pre_assert(if_softc_ctx_t scctx); 720 static void iflib_stop(if_ctx_t ctx); 721 static void iflib_if_init_locked(if_ctx_t ctx); 722 static void iflib_free_intr_mem(if_ctx_t ctx); 723 #ifndef __NO_STRICT_ALIGNMENT 724 static struct mbuf *iflib_fixup_rx(struct mbuf *m); 725 #endif 726 727 static SLIST_HEAD(cpu_offset_list, cpu_offset) cpu_offsets = 728 SLIST_HEAD_INITIALIZER(cpu_offsets); 729 struct cpu_offset { 730 SLIST_ENTRY(cpu_offset) entries; 731 cpuset_t set; 732 unsigned int refcount; 733 uint16_t next_cpuid; 734 }; 735 static struct mtx cpu_offset_mtx; 736 MTX_SYSINIT(iflib_cpu_offset, &cpu_offset_mtx, "iflib_cpu_offset lock", 737 MTX_DEF); 738 739 DEBUGNET_DEFINE(iflib); 740 741 static int 742 iflib_num_rx_descs(if_ctx_t ctx) 743 { 744 if_softc_ctx_t scctx = &ctx->ifc_softc_ctx; 745 if_shared_ctx_t sctx = ctx->ifc_sctx; 746 uint16_t first_rxq = (sctx->isc_flags & IFLIB_HAS_RXCQ) ? 1 : 0; 747 748 return (scctx->isc_nrxd[first_rxq]); 749 } 750 751 static int 752 iflib_num_tx_descs(if_ctx_t ctx) 753 { 754 if_softc_ctx_t scctx = &ctx->ifc_softc_ctx; 755 if_shared_ctx_t sctx = ctx->ifc_sctx; 756 uint16_t first_txq = (sctx->isc_flags & IFLIB_HAS_TXCQ) ? 1 : 0; 757 758 return (scctx->isc_ntxd[first_txq]); 759 } 760 761 #ifdef DEV_NETMAP 762 #include <sys/selinfo.h> 763 #include <net/netmap.h> 764 #include <dev/netmap/netmap_kern.h> 765 766 MODULE_DEPEND(iflib, netmap, 1, 1, 1); 767 768 static int netmap_fl_refill(iflib_rxq_t rxq, struct netmap_kring *kring, bool init); 769 static void iflib_netmap_timer(void *arg); 770 771 /* 772 * device-specific sysctl variables: 773 * 774 * iflib_crcstrip: 0: keep CRC in rx frames (default), 1: strip it. 775 * During regular operations the CRC is stripped, but on some 776 * hardware reception of frames not multiple of 64 is slower, 777 * so using crcstrip=0 helps in benchmarks. 778 * 779 * iflib_rx_miss, iflib_rx_miss_bufs: 780 * count packets that might be missed due to lost interrupts. 781 */ 782 SYSCTL_DECL(_dev_netmap); 783 /* 784 * The xl driver by default strips CRCs and we do not override it. 785 */ 786 787 int iflib_crcstrip = 1; 788 SYSCTL_INT(_dev_netmap, OID_AUTO, iflib_crcstrip, 789 CTLFLAG_RW, &iflib_crcstrip, 1, "strip CRC on RX frames"); 790 791 int iflib_rx_miss, iflib_rx_miss_bufs; 792 SYSCTL_INT(_dev_netmap, OID_AUTO, iflib_rx_miss, 793 CTLFLAG_RW, &iflib_rx_miss, 0, "potentially missed RX intr"); 794 SYSCTL_INT(_dev_netmap, OID_AUTO, iflib_rx_miss_bufs, 795 CTLFLAG_RW, &iflib_rx_miss_bufs, 0, "potentially missed RX intr bufs"); 796 797 /* 798 * Register/unregister. We are already under netmap lock. 799 * Only called on the first register or the last unregister. 800 */ 801 static int 802 iflib_netmap_register(struct netmap_adapter *na, int onoff) 803 { 804 if_t ifp = na->ifp; 805 if_ctx_t ctx = if_getsoftc(ifp); 806 int status; 807 808 CTX_LOCK(ctx); 809 if (!CTX_IS_VF(ctx)) 810 IFDI_CRCSTRIP_SET(ctx, onoff, iflib_crcstrip); 811 812 iflib_stop(ctx); 813 814 /* 815 * Enable (or disable) netmap flags, and intercept (or restore) 816 * ifp->if_transmit. This is done once the device has been stopped 817 * to prevent race conditions. Also, this must be done after 818 * calling netmap_disable_all_rings() and before calling 819 * netmap_enable_all_rings(), so that these two functions see the 820 * updated state of the NAF_NETMAP_ON bit. 821 */ 822 if (onoff) { 823 nm_set_native_flags(na); 824 } else { 825 nm_clear_native_flags(na); 826 } 827 828 iflib_init_locked(ctx); 829 IFDI_CRCSTRIP_SET(ctx, onoff, iflib_crcstrip); // XXX why twice ? 830 status = if_getdrvflags(ifp) & IFF_DRV_RUNNING ? 0 : 1; 831 if (status) 832 nm_clear_native_flags(na); 833 CTX_UNLOCK(ctx); 834 return (status); 835 } 836 837 static int 838 iflib_netmap_config(struct netmap_adapter *na, struct nm_config_info *info) 839 { 840 if_t ifp = na->ifp; 841 if_ctx_t ctx = if_getsoftc(ifp); 842 iflib_rxq_t rxq = &ctx->ifc_rxqs[0]; 843 iflib_fl_t fl = &rxq->ifr_fl[0]; 844 845 info->num_tx_rings = ctx->ifc_softc_ctx.isc_ntxqsets; 846 info->num_rx_rings = ctx->ifc_softc_ctx.isc_nrxqsets; 847 info->num_tx_descs = iflib_num_tx_descs(ctx); 848 info->num_rx_descs = iflib_num_rx_descs(ctx); 849 info->rx_buf_maxsize = fl->ifl_buf_size; 850 nm_prinf("txr %u rxr %u txd %u rxd %u rbufsz %u", 851 info->num_tx_rings, info->num_rx_rings, info->num_tx_descs, 852 info->num_rx_descs, info->rx_buf_maxsize); 853 854 return (0); 855 } 856 857 static int 858 netmap_fl_refill(iflib_rxq_t rxq, struct netmap_kring *kring, bool init) 859 { 860 struct netmap_adapter *na = kring->na; 861 u_int const lim = kring->nkr_num_slots - 1; 862 struct netmap_ring *ring = kring->ring; 863 bus_dmamap_t *map; 864 struct if_rxd_update iru; 865 if_ctx_t ctx = rxq->ifr_ctx; 866 iflib_fl_t fl = &rxq->ifr_fl[0]; 867 u_int nic_i_first, nic_i; 868 u_int nm_i; 869 int i, n; 870 #if IFLIB_DEBUG_COUNTERS 871 int rf_count = 0; 872 #endif 873 874 /* 875 * This function is used both at initialization and in rxsync. 876 * At initialization we need to prepare (with isc_rxd_refill()) 877 * all the netmap buffers currently owned by the kernel, in 878 * such a way to keep fl->ifl_pidx and kring->nr_hwcur in sync 879 * (except for kring->nkr_hwofs). These may be less than 880 * kring->nkr_num_slots if netmap_reset() was called while 881 * an application using the kring that still owned some 882 * buffers. 883 * At rxsync time, both indexes point to the next buffer to be 884 * refilled. 885 * In any case we publish (with isc_rxd_flush()) up to 886 * (fl->ifl_pidx - 1) % N (included), to avoid the NIC tail/prod 887 * pointer to overrun the head/cons pointer, although this is 888 * not necessary for some NICs (e.g. vmx). 889 */ 890 if (__predict_false(init)) { 891 n = kring->nkr_num_slots - nm_kr_rxspace(kring); 892 } else { 893 n = kring->rhead - kring->nr_hwcur; 894 if (n == 0) 895 return (0); /* Nothing to do. */ 896 if (n < 0) 897 n += kring->nkr_num_slots; 898 } 899 900 iru_init(&iru, rxq, 0 /* flid */); 901 map = fl->ifl_sds.ifsd_map; 902 nic_i = fl->ifl_pidx; 903 nm_i = netmap_idx_n2k(kring, nic_i); 904 if (__predict_false(init)) { 905 /* 906 * On init/reset, nic_i must be 0, and we must 907 * start to refill from hwtail (see netmap_reset()). 908 */ 909 MPASS(nic_i == 0); 910 MPASS(nm_i == kring->nr_hwtail); 911 } else 912 MPASS(nm_i == kring->nr_hwcur); 913 DBG_COUNTER_INC(fl_refills); 914 while (n > 0) { 915 #if IFLIB_DEBUG_COUNTERS 916 if (++rf_count == 9) 917 DBG_COUNTER_INC(fl_refills_large); 918 #endif 919 nic_i_first = nic_i; 920 for (i = 0; n > 0 && i < IFLIB_MAX_RX_REFRESH; n--, i++) { 921 struct netmap_slot *slot = &ring->slot[nm_i]; 922 uint64_t paddr; 923 void *addr = PNMB(na, slot, &paddr); 924 925 MPASS(i < IFLIB_MAX_RX_REFRESH); 926 927 if (addr == NETMAP_BUF_BASE(na)) /* bad buf */ 928 return (netmap_ring_reinit(kring)); 929 930 fl->ifl_bus_addrs[i] = paddr + 931 nm_get_offset(kring, slot); 932 fl->ifl_rxd_idxs[i] = nic_i; 933 934 if (__predict_false(init)) { 935 netmap_load_map(na, fl->ifl_buf_tag, 936 map[nic_i], addr); 937 } else if (slot->flags & NS_BUF_CHANGED) { 938 /* buffer has changed, reload map */ 939 netmap_reload_map(na, fl->ifl_buf_tag, 940 map[nic_i], addr); 941 } 942 bus_dmamap_sync(fl->ifl_buf_tag, map[nic_i], 943 BUS_DMASYNC_PREREAD); 944 slot->flags &= ~NS_BUF_CHANGED; 945 946 nm_i = nm_next(nm_i, lim); 947 nic_i = nm_next(nic_i, lim); 948 } 949 950 iru.iru_pidx = nic_i_first; 951 iru.iru_count = i; 952 ctx->isc_rxd_refill(ctx->ifc_softc, &iru); 953 } 954 fl->ifl_pidx = nic_i; 955 /* 956 * At the end of the loop we must have refilled everything 957 * we could possibly refill. 958 */ 959 MPASS(nm_i == kring->rhead); 960 kring->nr_hwcur = nm_i; 961 962 bus_dmamap_sync(fl->ifl_ifdi->idi_tag, fl->ifl_ifdi->idi_map, 963 BUS_DMASYNC_PREREAD | BUS_DMASYNC_PREWRITE); 964 ctx->isc_rxd_flush(ctx->ifc_softc, rxq->ifr_id, fl->ifl_id, 965 nm_prev(nic_i, lim)); 966 DBG_COUNTER_INC(rxd_flush); 967 968 return (0); 969 } 970 971 #define NETMAP_TX_TIMER_US 90 972 973 /* 974 * Reconcile kernel and user view of the transmit ring. 975 * 976 * All information is in the kring. 977 * Userspace wants to send packets up to the one before kring->rhead, 978 * kernel knows kring->nr_hwcur is the first unsent packet. 979 * 980 * Here we push packets out (as many as possible), and possibly 981 * reclaim buffers from previously completed transmission. 982 * 983 * The caller (netmap) guarantees that there is only one instance 984 * running at any time. Any interference with other driver 985 * methods should be handled by the individual drivers. 986 */ 987 static int 988 iflib_netmap_txsync(struct netmap_kring *kring, int flags) 989 { 990 struct netmap_adapter *na = kring->na; 991 if_t ifp = na->ifp; 992 struct netmap_ring *ring = kring->ring; 993 u_int nm_i; /* index into the netmap kring */ 994 u_int nic_i; /* index into the NIC ring */ 995 u_int const lim = kring->nkr_num_slots - 1; 996 u_int const head = kring->rhead; 997 struct if_pkt_info pi; 998 int tx_pkts = 0, tx_bytes = 0; 999 1000 /* 1001 * interrupts on every tx packet are expensive so request 1002 * them every half ring, or where NS_REPORT is set 1003 */ 1004 u_int report_frequency = kring->nkr_num_slots >> 1; 1005 /* device-specific */ 1006 if_ctx_t ctx = if_getsoftc(ifp); 1007 iflib_txq_t txq = &ctx->ifc_txqs[kring->ring_id]; 1008 1009 bus_dmamap_sync(txq->ift_ifdi->idi_tag, txq->ift_ifdi->idi_map, 1010 BUS_DMASYNC_POSTREAD | BUS_DMASYNC_POSTWRITE); 1011 1012 /* 1013 * First part: process new packets to send. 1014 * nm_i is the current index in the netmap kring, 1015 * nic_i is the corresponding index in the NIC ring. 1016 * 1017 * If we have packets to send (nm_i != head) 1018 * iterate over the netmap ring, fetch length and update 1019 * the corresponding slot in the NIC ring. Some drivers also 1020 * need to update the buffer's physical address in the NIC slot 1021 * even NS_BUF_CHANGED is not set (PNMB computes the addresses). 1022 * 1023 * The netmap_reload_map() calls is especially expensive, 1024 * even when (as in this case) the tag is 0, so do only 1025 * when the buffer has actually changed. 1026 * 1027 * If possible do not set the report/intr bit on all slots, 1028 * but only a few times per ring or when NS_REPORT is set. 1029 * 1030 * Finally, on 10G and faster drivers, it might be useful 1031 * to prefetch the next slot and txr entry. 1032 */ 1033 1034 nm_i = kring->nr_hwcur; 1035 if (nm_i != head) { /* we have new packets to send */ 1036 uint32_t pkt_len = 0, seg_idx = 0; 1037 int nic_i_start = -1, flags = 0; 1038 pkt_info_zero(&pi); 1039 pi.ipi_segs = txq->ift_segs; 1040 pi.ipi_qsidx = kring->ring_id; 1041 nic_i = netmap_idx_k2n(kring, nm_i); 1042 1043 __builtin_prefetch(&ring->slot[nm_i]); 1044 __builtin_prefetch(&txq->ift_sds.ifsd_m[nic_i]); 1045 __builtin_prefetch(&txq->ift_sds.ifsd_map[nic_i]); 1046 1047 while (nm_i != head) { 1048 struct netmap_slot *slot = &ring->slot[nm_i]; 1049 uint64_t offset = nm_get_offset(kring, slot); 1050 u_int len = slot->len; 1051 uint64_t paddr; 1052 void *addr = PNMB(na, slot, &paddr); 1053 1054 flags |= (slot->flags & NS_REPORT || 1055 nic_i == 0 || nic_i == report_frequency) ? 1056 IPI_TX_INTR : 0; 1057 1058 /* 1059 * If this is the first packet fragment, save the 1060 * index of the first NIC slot for later. 1061 */ 1062 if (nic_i_start < 0) 1063 nic_i_start = nic_i; 1064 1065 pi.ipi_segs[seg_idx].ds_addr = paddr + offset; 1066 pi.ipi_segs[seg_idx].ds_len = len; 1067 if (len) { 1068 pkt_len += len; 1069 seg_idx++; 1070 } 1071 1072 if (!(slot->flags & NS_MOREFRAG)) { 1073 pi.ipi_len = pkt_len; 1074 pi.ipi_nsegs = seg_idx; 1075 pi.ipi_pidx = nic_i_start; 1076 pi.ipi_ndescs = 0; 1077 pi.ipi_flags = flags; 1078 1079 /* Prepare the NIC TX ring. */ 1080 ctx->isc_txd_encap(ctx->ifc_softc, &pi); 1081 DBG_COUNTER_INC(tx_encap); 1082 1083 /* Update transmit counters */ 1084 tx_bytes += pi.ipi_len; 1085 tx_pkts++; 1086 1087 /* Reinit per-packet info for the next one. */ 1088 flags = seg_idx = pkt_len = 0; 1089 nic_i_start = -1; 1090 } 1091 1092 /* prefetch for next round */ 1093 __builtin_prefetch(&ring->slot[nm_i + 1]); 1094 __builtin_prefetch(&txq->ift_sds.ifsd_m[nic_i + 1]); 1095 __builtin_prefetch(&txq->ift_sds.ifsd_map[nic_i + 1]); 1096 1097 NM_CHECK_ADDR_LEN_OFF(na, len, offset); 1098 1099 if (slot->flags & NS_BUF_CHANGED) { 1100 /* buffer has changed, reload map */ 1101 netmap_reload_map(na, txq->ift_buf_tag, 1102 txq->ift_sds.ifsd_map[nic_i], addr); 1103 } 1104 /* make sure changes to the buffer are synced */ 1105 bus_dmamap_sync(txq->ift_buf_tag, 1106 txq->ift_sds.ifsd_map[nic_i], 1107 BUS_DMASYNC_PREWRITE); 1108 1109 slot->flags &= ~(NS_REPORT | NS_BUF_CHANGED | NS_MOREFRAG); 1110 nm_i = nm_next(nm_i, lim); 1111 nic_i = nm_next(nic_i, lim); 1112 } 1113 kring->nr_hwcur = nm_i; 1114 1115 /* synchronize the NIC ring */ 1116 bus_dmamap_sync(txq->ift_ifdi->idi_tag, txq->ift_ifdi->idi_map, 1117 BUS_DMASYNC_PREREAD | BUS_DMASYNC_PREWRITE); 1118 1119 /* (re)start the tx unit up to slot nic_i (excluded) */ 1120 ctx->isc_txd_flush(ctx->ifc_softc, txq->ift_id, nic_i); 1121 } 1122 1123 /* 1124 * Second part: reclaim buffers for completed transmissions. 1125 * 1126 * If there are unclaimed buffers, attempt to reclaim them. 1127 * If we don't manage to reclaim them all, and TX IRQs are not in use, 1128 * trigger a per-tx-queue timer to try again later. 1129 */ 1130 if (kring->nr_hwtail != nm_prev(kring->nr_hwcur, lim)) { 1131 if (iflib_tx_credits_update(ctx, txq)) { 1132 /* some tx completed, increment avail */ 1133 nic_i = txq->ift_cidx_processed; 1134 kring->nr_hwtail = nm_prev(netmap_idx_n2k(kring, nic_i), lim); 1135 } 1136 } 1137 1138 if (!(ctx->ifc_flags & IFC_NETMAP_TX_IRQ)) 1139 if (kring->nr_hwtail != nm_prev(kring->nr_hwcur, lim)) { 1140 callout_reset_sbt_on(&txq->ift_netmap_timer, 1141 NETMAP_TX_TIMER_US * SBT_1US, SBT_1US, 1142 iflib_netmap_timer, txq, 1143 txq->ift_netmap_timer.c_cpu, 0); 1144 } 1145 1146 if_inc_counter(ifp, IFCOUNTER_OBYTES, tx_bytes); 1147 if_inc_counter(ifp, IFCOUNTER_OPACKETS, tx_pkts); 1148 1149 return (0); 1150 } 1151 1152 /* 1153 * Reconcile kernel and user view of the receive ring. 1154 * Same as for the txsync, this routine must be efficient. 1155 * The caller guarantees a single invocations, but races against 1156 * the rest of the driver should be handled here. 1157 * 1158 * On call, kring->rhead is the first packet that userspace wants 1159 * to keep, and kring->rcur is the wakeup point. 1160 * The kernel has previously reported packets up to kring->rtail. 1161 * 1162 * If (flags & NAF_FORCE_READ) also check for incoming packets irrespective 1163 * of whether or not we received an interrupt. 1164 */ 1165 static int 1166 iflib_netmap_rxsync(struct netmap_kring *kring, int flags) 1167 { 1168 struct netmap_adapter *na = kring->na; 1169 struct netmap_ring *ring = kring->ring; 1170 if_t ifp = na->ifp; 1171 uint32_t nm_i; /* index into the netmap ring */ 1172 uint32_t nic_i; /* index into the NIC ring */ 1173 u_int n; 1174 u_int const lim = kring->nkr_num_slots - 1; 1175 int force_update = (flags & NAF_FORCE_READ) || kring->nr_kflags & NKR_PENDINTR; 1176 int i = 0, rx_bytes = 0, rx_pkts = 0; 1177 1178 if_ctx_t ctx = if_getsoftc(ifp); 1179 if_shared_ctx_t sctx = ctx->ifc_sctx; 1180 if_softc_ctx_t scctx = &ctx->ifc_softc_ctx; 1181 iflib_rxq_t rxq = &ctx->ifc_rxqs[kring->ring_id]; 1182 iflib_fl_t fl = &rxq->ifr_fl[0]; 1183 struct if_rxd_info ri; 1184 qidx_t *cidxp; 1185 1186 /* 1187 * netmap only uses free list 0, to avoid out of order consumption 1188 * of receive buffers 1189 */ 1190 1191 bus_dmamap_sync(fl->ifl_ifdi->idi_tag, fl->ifl_ifdi->idi_map, 1192 BUS_DMASYNC_POSTREAD | BUS_DMASYNC_POSTWRITE); 1193 1194 /* 1195 * First part: import newly received packets. 1196 * 1197 * nm_i is the index of the next free slot in the netmap ring, 1198 * nic_i is the index of the next received packet in the NIC ring 1199 * (or in the free list 0 if IFLIB_HAS_RXCQ is set), and they may 1200 * differ in case if_init() has been called while 1201 * in netmap mode. For the receive ring we have 1202 * 1203 * nic_i = fl->ifl_cidx; 1204 * nm_i = kring->nr_hwtail (previous) 1205 * and 1206 * nm_i == (nic_i + kring->nkr_hwofs) % ring_size 1207 * 1208 * fl->ifl_cidx is set to 0 on a ring reinit 1209 */ 1210 if (netmap_no_pendintr || force_update) { 1211 uint32_t hwtail_lim = nm_prev(kring->nr_hwcur, lim); 1212 bool have_rxcq = sctx->isc_flags & IFLIB_HAS_RXCQ; 1213 int crclen = iflib_crcstrip ? 0 : 4; 1214 int error, avail; 1215 1216 /* 1217 * For the free list consumer index, we use the same 1218 * logic as in iflib_rxeof(). 1219 */ 1220 if (have_rxcq) 1221 cidxp = &rxq->ifr_cq_cidx; 1222 else 1223 cidxp = &fl->ifl_cidx; 1224 avail = ctx->isc_rxd_available(ctx->ifc_softc, 1225 rxq->ifr_id, *cidxp, USHRT_MAX); 1226 1227 nic_i = fl->ifl_cidx; 1228 nm_i = netmap_idx_n2k(kring, nic_i); 1229 MPASS(nm_i == kring->nr_hwtail); 1230 for (n = 0; avail > 0 && nm_i != hwtail_lim; n++, avail--) { 1231 rxd_info_zero(&ri); 1232 ri.iri_frags = rxq->ifr_frags; 1233 ri.iri_qsidx = kring->ring_id; 1234 ri.iri_ifp = ctx->ifc_ifp; 1235 ri.iri_cidx = *cidxp; 1236 1237 error = ctx->isc_rxd_pkt_get(ctx->ifc_softc, &ri); 1238 for (i = 0; i < ri.iri_nfrags; i++) { 1239 if (error) { 1240 ring->slot[nm_i].len = 0; 1241 ring->slot[nm_i].flags = 0; 1242 } else { 1243 ring->slot[nm_i].len = ri.iri_frags[i].irf_len; 1244 if (i == (ri.iri_nfrags - 1)) { 1245 ring->slot[nm_i].len -= crclen; 1246 ring->slot[nm_i].flags = 0; 1247 1248 /* Update receive counters */ 1249 rx_bytes += ri.iri_len; 1250 rx_pkts++; 1251 } else 1252 ring->slot[nm_i].flags = NS_MOREFRAG; 1253 } 1254 1255 bus_dmamap_sync(fl->ifl_buf_tag, 1256 fl->ifl_sds.ifsd_map[nic_i], BUS_DMASYNC_POSTREAD); 1257 nm_i = nm_next(nm_i, lim); 1258 fl->ifl_cidx = nic_i = nm_next(nic_i, lim); 1259 } 1260 1261 if (have_rxcq) { 1262 *cidxp = ri.iri_cidx; 1263 while (*cidxp >= scctx->isc_nrxd[0]) 1264 *cidxp -= scctx->isc_nrxd[0]; 1265 } 1266 1267 } 1268 if (n) { /* update the state variables */ 1269 if (netmap_no_pendintr && !force_update) { 1270 /* diagnostics */ 1271 iflib_rx_miss++; 1272 iflib_rx_miss_bufs += n; 1273 } 1274 kring->nr_hwtail = nm_i; 1275 } 1276 kring->nr_kflags &= ~NKR_PENDINTR; 1277 } 1278 /* 1279 * Second part: skip past packets that userspace has released. 1280 * (kring->nr_hwcur to head excluded), 1281 * and make the buffers available for reception. 1282 * As usual nm_i is the index in the netmap ring, 1283 * nic_i is the index in the NIC ring, and 1284 * nm_i == (nic_i + kring->nkr_hwofs) % ring_size 1285 */ 1286 netmap_fl_refill(rxq, kring, false); 1287 1288 if_inc_counter(ifp, IFCOUNTER_IBYTES, rx_bytes); 1289 if_inc_counter(ifp, IFCOUNTER_IPACKETS, rx_pkts); 1290 1291 return (0); 1292 } 1293 1294 static void 1295 iflib_netmap_intr(struct netmap_adapter *na, int onoff) 1296 { 1297 if_ctx_t ctx = if_getsoftc(na->ifp); 1298 1299 CTX_LOCK(ctx); 1300 if (onoff) { 1301 IFDI_INTR_ENABLE(ctx); 1302 } else { 1303 IFDI_INTR_DISABLE(ctx); 1304 } 1305 CTX_UNLOCK(ctx); 1306 } 1307 1308 static int 1309 iflib_netmap_attach(if_ctx_t ctx) 1310 { 1311 struct netmap_adapter na; 1312 1313 bzero(&na, sizeof(na)); 1314 1315 na.ifp = ctx->ifc_ifp; 1316 na.na_flags = NAF_BDG_MAYSLEEP | NAF_MOREFRAG | NAF_OFFSETS; 1317 MPASS(ctx->ifc_softc_ctx.isc_ntxqsets); 1318 MPASS(ctx->ifc_softc_ctx.isc_nrxqsets); 1319 1320 na.num_tx_desc = iflib_num_tx_descs(ctx); 1321 na.num_rx_desc = iflib_num_rx_descs(ctx); 1322 na.nm_txsync = iflib_netmap_txsync; 1323 na.nm_rxsync = iflib_netmap_rxsync; 1324 na.nm_register = iflib_netmap_register; 1325 na.nm_intr = iflib_netmap_intr; 1326 na.nm_config = iflib_netmap_config; 1327 na.num_tx_rings = ctx->ifc_softc_ctx.isc_ntxqsets; 1328 na.num_rx_rings = ctx->ifc_softc_ctx.isc_nrxqsets; 1329 return (netmap_attach(&na)); 1330 } 1331 1332 static int 1333 iflib_netmap_txq_init(if_ctx_t ctx, iflib_txq_t txq) 1334 { 1335 struct netmap_adapter *na = NA(ctx->ifc_ifp); 1336 struct netmap_slot *slot; 1337 1338 slot = netmap_reset(na, NR_TX, txq->ift_id, 0); 1339 if (slot == NULL) 1340 return (0); 1341 for (int i = 0; i < ctx->ifc_softc_ctx.isc_ntxd[0]; i++) { 1342 /* 1343 * In netmap mode, set the map for the packet buffer. 1344 * NOTE: Some drivers (not this one) also need to set 1345 * the physical buffer address in the NIC ring. 1346 * netmap_idx_n2k() maps a nic index, i, into the corresponding 1347 * netmap slot index, si 1348 */ 1349 int si = netmap_idx_n2k(na->tx_rings[txq->ift_id], i); 1350 netmap_load_map(na, txq->ift_buf_tag, txq->ift_sds.ifsd_map[i], 1351 NMB(na, slot + si)); 1352 } 1353 return (1); 1354 } 1355 1356 static int 1357 iflib_netmap_rxq_init(if_ctx_t ctx, iflib_rxq_t rxq) 1358 { 1359 struct netmap_adapter *na = NA(ctx->ifc_ifp); 1360 struct netmap_kring *kring; 1361 struct netmap_slot *slot; 1362 1363 slot = netmap_reset(na, NR_RX, rxq->ifr_id, 0); 1364 if (slot == NULL) 1365 return (0); 1366 kring = na->rx_rings[rxq->ifr_id]; 1367 netmap_fl_refill(rxq, kring, true); 1368 return (1); 1369 } 1370 1371 static void 1372 iflib_netmap_timer(void *arg) 1373 { 1374 iflib_txq_t txq = arg; 1375 if_ctx_t ctx = txq->ift_ctx; 1376 1377 /* 1378 * Wake up the netmap application, to give it a chance to 1379 * call txsync and reclaim more completed TX buffers. 1380 */ 1381 netmap_tx_irq(ctx->ifc_ifp, txq->ift_id); 1382 } 1383 1384 #define iflib_netmap_detach(ifp) netmap_detach(ifp) 1385 1386 #else 1387 #define iflib_netmap_txq_init(ctx, txq) (0) 1388 #define iflib_netmap_rxq_init(ctx, rxq) (0) 1389 #define iflib_netmap_detach(ifp) 1390 #define netmap_enable_all_rings(ifp) 1391 #define netmap_disable_all_rings(ifp) 1392 1393 #define iflib_netmap_attach(ctx) (0) 1394 #define netmap_rx_irq(ifp, qid, budget) (0) 1395 #endif 1396 1397 #if defined(__i386__) || defined(__amd64__) 1398 static __inline void 1399 prefetch(void *x) 1400 { 1401 __asm volatile("prefetcht0 %0" :: "m" (*(unsigned long *)x)); 1402 } 1403 1404 static __inline void 1405 prefetch2cachelines(void *x) 1406 { 1407 __asm volatile("prefetcht0 %0" :: "m" (*(unsigned long *)x)); 1408 #if (CACHE_LINE_SIZE < 128) 1409 __asm volatile("prefetcht0 %0" :: "m" (*(((unsigned long *)x) + CACHE_LINE_SIZE / (sizeof(unsigned long))))); 1410 #endif 1411 } 1412 #else 1413 static __inline void 1414 prefetch(void *x) 1415 { 1416 } 1417 1418 static __inline void 1419 prefetch2cachelines(void *x) 1420 { 1421 } 1422 #endif 1423 1424 static void 1425 iru_init(if_rxd_update_t iru, iflib_rxq_t rxq, uint8_t flid) 1426 { 1427 iflib_fl_t fl; 1428 1429 fl = &rxq->ifr_fl[flid]; 1430 iru->iru_paddrs = fl->ifl_bus_addrs; 1431 iru->iru_idxs = fl->ifl_rxd_idxs; 1432 iru->iru_qsidx = rxq->ifr_id; 1433 iru->iru_buf_size = fl->ifl_buf_size; 1434 iru->iru_flidx = fl->ifl_id; 1435 } 1436 1437 static void 1438 _iflib_dmamap_cb(void *arg, bus_dma_segment_t *segs, int nseg, int err) 1439 { 1440 if (err) 1441 return; 1442 *(bus_addr_t *) arg = segs[0].ds_addr; 1443 } 1444 1445 #define DMA_WIDTH_TO_BUS_LOWADDR(width) \ 1446 (((width) == 0) || (width) == flsll(BUS_SPACE_MAXADDR) ? \ 1447 BUS_SPACE_MAXADDR : (1ULL << (width)) - 1ULL) 1448 1449 int 1450 iflib_dma_alloc_align(if_ctx_t ctx, int size, int align, iflib_dma_info_t dma, int mapflags) 1451 { 1452 int err; 1453 device_t dev = ctx->ifc_dev; 1454 bus_addr_t lowaddr; 1455 1456 lowaddr = DMA_WIDTH_TO_BUS_LOWADDR(ctx->ifc_softc_ctx.isc_dma_width); 1457 1458 err = bus_dma_tag_create(bus_get_dma_tag(dev), /* parent */ 1459 align, 0, /* alignment, bounds */ 1460 lowaddr, /* lowaddr */ 1461 BUS_SPACE_MAXADDR, /* highaddr */ 1462 NULL, NULL, /* filter, filterarg */ 1463 size, /* maxsize */ 1464 1, /* nsegments */ 1465 size, /* maxsegsize */ 1466 BUS_DMA_ALLOCNOW, /* flags */ 1467 NULL, /* lockfunc */ 1468 NULL, /* lockarg */ 1469 &dma->idi_tag); 1470 if (err) { 1471 device_printf(dev, 1472 "%s: bus_dma_tag_create failed: %d (size=%d, align=%d)\n", 1473 __func__, err, size, align); 1474 goto fail_0; 1475 } 1476 1477 err = bus_dmamem_alloc(dma->idi_tag, (void **)&dma->idi_vaddr, 1478 BUS_DMA_NOWAIT | BUS_DMA_COHERENT | BUS_DMA_ZERO, &dma->idi_map); 1479 if (err) { 1480 device_printf(dev, 1481 "%s: bus_dmamem_alloc(%ju) failed: %d\n", 1482 __func__, (uintmax_t)size, err); 1483 goto fail_1; 1484 } 1485 1486 dma->idi_paddr = IF_BAD_DMA; 1487 err = bus_dmamap_load(dma->idi_tag, dma->idi_map, dma->idi_vaddr, 1488 size, _iflib_dmamap_cb, &dma->idi_paddr, mapflags | BUS_DMA_NOWAIT); 1489 if (err || dma->idi_paddr == IF_BAD_DMA) { 1490 device_printf(dev, 1491 "%s: bus_dmamap_load failed: %d\n", 1492 __func__, err); 1493 goto fail_2; 1494 } 1495 1496 dma->idi_size = size; 1497 return (0); 1498 1499 fail_2: 1500 bus_dmamem_free(dma->idi_tag, dma->idi_vaddr, dma->idi_map); 1501 fail_1: 1502 bus_dma_tag_destroy(dma->idi_tag); 1503 fail_0: 1504 dma->idi_tag = NULL; 1505 1506 return (err); 1507 } 1508 1509 int 1510 iflib_dma_alloc(if_ctx_t ctx, int size, iflib_dma_info_t dma, int mapflags) 1511 { 1512 if_shared_ctx_t sctx = ctx->ifc_sctx; 1513 1514 KASSERT(sctx->isc_q_align != 0, ("alignment value not initialized")); 1515 1516 return (iflib_dma_alloc_align(ctx, size, sctx->isc_q_align, dma, mapflags)); 1517 } 1518 1519 int 1520 iflib_dma_alloc_multi(if_ctx_t ctx, int *sizes, iflib_dma_info_t *dmalist, int mapflags, int count) 1521 { 1522 int i, err; 1523 iflib_dma_info_t *dmaiter; 1524 1525 dmaiter = dmalist; 1526 for (i = 0; i < count; i++, dmaiter++) { 1527 if ((err = iflib_dma_alloc(ctx, sizes[i], *dmaiter, mapflags)) != 0) 1528 break; 1529 } 1530 if (err) 1531 iflib_dma_free_multi(dmalist, i); 1532 return (err); 1533 } 1534 1535 void 1536 iflib_dma_free(iflib_dma_info_t dma) 1537 { 1538 if (dma->idi_tag == NULL) 1539 return; 1540 if (dma->idi_paddr != IF_BAD_DMA) { 1541 bus_dmamap_sync(dma->idi_tag, dma->idi_map, 1542 BUS_DMASYNC_POSTREAD | BUS_DMASYNC_POSTWRITE); 1543 bus_dmamap_unload(dma->idi_tag, dma->idi_map); 1544 dma->idi_paddr = IF_BAD_DMA; 1545 } 1546 if (dma->idi_vaddr != NULL) { 1547 bus_dmamem_free(dma->idi_tag, dma->idi_vaddr, dma->idi_map); 1548 dma->idi_vaddr = NULL; 1549 } 1550 bus_dma_tag_destroy(dma->idi_tag); 1551 dma->idi_tag = NULL; 1552 } 1553 1554 void 1555 iflib_dma_free_multi(iflib_dma_info_t *dmalist, int count) 1556 { 1557 int i; 1558 iflib_dma_info_t *dmaiter = dmalist; 1559 1560 for (i = 0; i < count; i++, dmaiter++) 1561 iflib_dma_free(*dmaiter); 1562 } 1563 1564 static int 1565 iflib_fast_intr(void *arg) 1566 { 1567 iflib_filter_info_t info = arg; 1568 struct grouptask *gtask = info->ifi_task; 1569 int result; 1570 1571 DBG_COUNTER_INC(fast_intrs); 1572 if (info->ifi_filter != NULL) { 1573 result = info->ifi_filter(info->ifi_filter_arg); 1574 if ((result & FILTER_SCHEDULE_THREAD) == 0) 1575 return (result); 1576 } 1577 1578 GROUPTASK_ENQUEUE(gtask); 1579 return (FILTER_HANDLED); 1580 } 1581 1582 static int 1583 iflib_fast_intr_rxtx(void *arg) 1584 { 1585 iflib_filter_info_t info = arg; 1586 struct grouptask *gtask = info->ifi_task; 1587 if_ctx_t ctx; 1588 iflib_rxq_t rxq = (iflib_rxq_t)info->ifi_ctx; 1589 iflib_txq_t txq; 1590 void *sc; 1591 int i, cidx, result; 1592 qidx_t txqid; 1593 bool intr_enable, intr_legacy; 1594 1595 DBG_COUNTER_INC(fast_intrs); 1596 if (info->ifi_filter != NULL) { 1597 result = info->ifi_filter(info->ifi_filter_arg); 1598 if ((result & FILTER_SCHEDULE_THREAD) == 0) 1599 return (result); 1600 } 1601 1602 ctx = rxq->ifr_ctx; 1603 sc = ctx->ifc_softc; 1604 intr_enable = false; 1605 intr_legacy = !!(ctx->ifc_flags & IFC_LEGACY); 1606 MPASS(rxq->ifr_ntxqirq); 1607 for (i = 0; i < rxq->ifr_ntxqirq; i++) { 1608 txqid = rxq->ifr_txqid[i]; 1609 txq = &ctx->ifc_txqs[txqid]; 1610 bus_dmamap_sync(txq->ift_ifdi->idi_tag, txq->ift_ifdi->idi_map, 1611 BUS_DMASYNC_POSTREAD); 1612 if (!ctx->isc_txd_credits_update(sc, txqid, false)) { 1613 if (intr_legacy) 1614 intr_enable = true; 1615 else 1616 IFDI_TX_QUEUE_INTR_ENABLE(ctx, txqid); 1617 continue; 1618 } 1619 GROUPTASK_ENQUEUE(&txq->ift_task); 1620 } 1621 if (ctx->ifc_sctx->isc_flags & IFLIB_HAS_RXCQ) 1622 cidx = rxq->ifr_cq_cidx; 1623 else 1624 cidx = rxq->ifr_fl[0].ifl_cidx; 1625 if (iflib_rxd_avail(ctx, rxq, cidx, 1)) 1626 GROUPTASK_ENQUEUE(gtask); 1627 else { 1628 if (intr_legacy) 1629 intr_enable = true; 1630 else 1631 IFDI_RX_QUEUE_INTR_ENABLE(ctx, rxq->ifr_id); 1632 DBG_COUNTER_INC(rx_intr_enables); 1633 } 1634 if (intr_enable) 1635 IFDI_INTR_ENABLE(ctx); 1636 return (FILTER_HANDLED); 1637 } 1638 1639 static int 1640 iflib_fast_intr_ctx(void *arg) 1641 { 1642 iflib_filter_info_t info = arg; 1643 struct grouptask *gtask = info->ifi_task; 1644 int result; 1645 1646 DBG_COUNTER_INC(fast_intrs); 1647 if (info->ifi_filter != NULL) { 1648 result = info->ifi_filter(info->ifi_filter_arg); 1649 if ((result & FILTER_SCHEDULE_THREAD) == 0) 1650 return (result); 1651 } 1652 1653 if (gtask->gt_taskqueue != NULL) 1654 GROUPTASK_ENQUEUE(gtask); 1655 return (FILTER_HANDLED); 1656 } 1657 1658 static int 1659 _iflib_irq_alloc(if_ctx_t ctx, if_irq_t irq, int rid, 1660 driver_filter_t filter, driver_intr_t handler, void *arg, 1661 const char *name) 1662 { 1663 struct resource *res; 1664 void *tag = NULL; 1665 device_t dev = ctx->ifc_dev; 1666 int flags, i, rc; 1667 1668 flags = RF_ACTIVE; 1669 if (ctx->ifc_flags & IFC_LEGACY) 1670 flags |= RF_SHAREABLE; 1671 MPASS(rid < 512); 1672 i = rid; 1673 res = bus_alloc_resource_any(dev, SYS_RES_IRQ, &i, flags); 1674 if (res == NULL) { 1675 device_printf(dev, 1676 "failed to allocate IRQ for rid %d, name %s.\n", rid, name); 1677 return (ENOMEM); 1678 } 1679 irq->ii_res = res; 1680 KASSERT(filter == NULL || handler == NULL, ("filter and handler can't both be non-NULL")); 1681 rc = bus_setup_intr(dev, res, INTR_MPSAFE | INTR_TYPE_NET, 1682 filter, handler, arg, &tag); 1683 if (rc != 0) { 1684 device_printf(dev, 1685 "failed to setup interrupt for rid %d, name %s: %d\n", 1686 rid, name ? name : "unknown", rc); 1687 return (rc); 1688 } else if (name) 1689 bus_describe_intr(dev, res, tag, "%s", name); 1690 1691 irq->ii_tag = tag; 1692 return (0); 1693 } 1694 1695 /********************************************************************* 1696 * 1697 * Allocate DMA resources for TX buffers as well as memory for the TX 1698 * mbuf map. TX DMA maps (non-TSO/TSO) and TX mbuf map are kept in a 1699 * iflib_sw_tx_desc_array structure, storing all the information that 1700 * is needed to transmit a packet on the wire. This is called only 1701 * once at attach, setup is done every reset. 1702 * 1703 **********************************************************************/ 1704 static int 1705 iflib_txsd_alloc(iflib_txq_t txq) 1706 { 1707 if_ctx_t ctx = txq->ift_ctx; 1708 if_shared_ctx_t sctx = ctx->ifc_sctx; 1709 if_softc_ctx_t scctx = &ctx->ifc_softc_ctx; 1710 device_t dev = ctx->ifc_dev; 1711 bus_size_t tsomaxsize; 1712 bus_addr_t lowaddr; 1713 int err, nsegments, ntsosegments; 1714 bool tso; 1715 1716 nsegments = scctx->isc_tx_nsegments; 1717 ntsosegments = scctx->isc_tx_tso_segments_max; 1718 tsomaxsize = scctx->isc_tx_tso_size_max; 1719 if (if_getcapabilities(ctx->ifc_ifp) & IFCAP_VLAN_MTU) 1720 tsomaxsize += sizeof(struct ether_vlan_header); 1721 MPASS(scctx->isc_ntxd[0] > 0); 1722 MPASS(scctx->isc_ntxd[txq->ift_br_offset] > 0); 1723 MPASS(nsegments > 0); 1724 if (if_getcapabilities(ctx->ifc_ifp) & IFCAP_TSO) { 1725 MPASS(ntsosegments > 0); 1726 MPASS(sctx->isc_tso_maxsize >= tsomaxsize); 1727 } 1728 1729 lowaddr = DMA_WIDTH_TO_BUS_LOWADDR(scctx->isc_dma_width); 1730 1731 /* 1732 * Set up DMA tags for TX buffers. 1733 */ 1734 if ((err = bus_dma_tag_create(bus_get_dma_tag(dev), 1735 1, 0, /* alignment, bounds */ 1736 lowaddr, /* lowaddr */ 1737 BUS_SPACE_MAXADDR, /* highaddr */ 1738 NULL, NULL, /* filter, filterarg */ 1739 sctx->isc_tx_maxsize, /* maxsize */ 1740 nsegments, /* nsegments */ 1741 sctx->isc_tx_maxsegsize, /* maxsegsize */ 1742 0, /* flags */ 1743 NULL, /* lockfunc */ 1744 NULL, /* lockfuncarg */ 1745 &txq->ift_buf_tag))) { 1746 device_printf(dev, "Unable to allocate TX DMA tag: %d\n", err); 1747 device_printf(dev, "maxsize: %ju nsegments: %d maxsegsize: %ju\n", 1748 (uintmax_t)sctx->isc_tx_maxsize, nsegments, (uintmax_t)sctx->isc_tx_maxsegsize); 1749 goto fail; 1750 } 1751 tso = (if_getcapabilities(ctx->ifc_ifp) & IFCAP_TSO) != 0; 1752 if (tso && (err = bus_dma_tag_create(bus_get_dma_tag(dev), 1753 1, 0, /* alignment, bounds */ 1754 lowaddr, /* lowaddr */ 1755 BUS_SPACE_MAXADDR, /* highaddr */ 1756 NULL, NULL, /* filter, filterarg */ 1757 tsomaxsize, /* maxsize */ 1758 ntsosegments, /* nsegments */ 1759 sctx->isc_tso_maxsegsize,/* maxsegsize */ 1760 0, /* flags */ 1761 NULL, /* lockfunc */ 1762 NULL, /* lockfuncarg */ 1763 &txq->ift_tso_buf_tag))) { 1764 device_printf(dev, "Unable to allocate TSO TX DMA tag: %d\n", 1765 err); 1766 goto fail; 1767 } 1768 1769 /* Allocate memory for the TX mbuf map. */ 1770 if (!(txq->ift_sds.ifsd_m = 1771 (struct mbuf **) malloc(sizeof(struct mbuf *) * 1772 scctx->isc_ntxd[txq->ift_br_offset], M_IFLIB, M_NOWAIT | M_ZERO))) { 1773 device_printf(dev, "Unable to allocate TX mbuf map memory\n"); 1774 err = ENOMEM; 1775 goto fail; 1776 } 1777 1778 /* 1779 * Create the DMA maps for TX buffers. 1780 */ 1781 if ((txq->ift_sds.ifsd_map = (bus_dmamap_t *)malloc( 1782 sizeof(bus_dmamap_t) * scctx->isc_ntxd[txq->ift_br_offset], 1783 M_IFLIB, M_NOWAIT | M_ZERO)) == NULL) { 1784 device_printf(dev, 1785 "Unable to allocate TX buffer DMA map memory\n"); 1786 err = ENOMEM; 1787 goto fail; 1788 } 1789 if (tso && (txq->ift_sds.ifsd_tso_map = (bus_dmamap_t *)malloc( 1790 sizeof(bus_dmamap_t) * scctx->isc_ntxd[txq->ift_br_offset], 1791 M_IFLIB, M_NOWAIT | M_ZERO)) == NULL) { 1792 device_printf(dev, 1793 "Unable to allocate TSO TX buffer map memory\n"); 1794 err = ENOMEM; 1795 goto fail; 1796 } 1797 for (int i = 0; i < scctx->isc_ntxd[txq->ift_br_offset]; i++) { 1798 err = bus_dmamap_create(txq->ift_buf_tag, 0, 1799 &txq->ift_sds.ifsd_map[i]); 1800 if (err != 0) { 1801 device_printf(dev, "Unable to create TX DMA map\n"); 1802 goto fail; 1803 } 1804 if (!tso) 1805 continue; 1806 err = bus_dmamap_create(txq->ift_tso_buf_tag, 0, 1807 &txq->ift_sds.ifsd_tso_map[i]); 1808 if (err != 0) { 1809 device_printf(dev, "Unable to create TSO TX DMA map\n"); 1810 goto fail; 1811 } 1812 } 1813 return (0); 1814 fail: 1815 /* We free all, it handles case where we are in the middle */ 1816 iflib_tx_structures_free(ctx); 1817 return (err); 1818 } 1819 1820 static void 1821 iflib_txsd_destroy(if_ctx_t ctx, iflib_txq_t txq, int i) 1822 { 1823 bus_dmamap_t map; 1824 1825 if (txq->ift_sds.ifsd_map != NULL) { 1826 map = txq->ift_sds.ifsd_map[i]; 1827 bus_dmamap_sync(txq->ift_buf_tag, map, BUS_DMASYNC_POSTWRITE); 1828 bus_dmamap_unload(txq->ift_buf_tag, map); 1829 bus_dmamap_destroy(txq->ift_buf_tag, map); 1830 txq->ift_sds.ifsd_map[i] = NULL; 1831 } 1832 1833 if (txq->ift_sds.ifsd_tso_map != NULL) { 1834 map = txq->ift_sds.ifsd_tso_map[i]; 1835 bus_dmamap_sync(txq->ift_tso_buf_tag, map, 1836 BUS_DMASYNC_POSTWRITE); 1837 bus_dmamap_unload(txq->ift_tso_buf_tag, map); 1838 bus_dmamap_destroy(txq->ift_tso_buf_tag, map); 1839 txq->ift_sds.ifsd_tso_map[i] = NULL; 1840 } 1841 } 1842 1843 static void 1844 iflib_txq_destroy(iflib_txq_t txq) 1845 { 1846 if_ctx_t ctx = txq->ift_ctx; 1847 1848 for (int i = 0; i < txq->ift_size; i++) 1849 iflib_txsd_destroy(ctx, txq, i); 1850 1851 if (txq->ift_br != NULL) { 1852 ifmp_ring_free(txq->ift_br); 1853 txq->ift_br = NULL; 1854 } 1855 1856 mtx_destroy(&txq->ift_mtx); 1857 1858 if (txq->ift_sds.ifsd_map != NULL) { 1859 free(txq->ift_sds.ifsd_map, M_IFLIB); 1860 txq->ift_sds.ifsd_map = NULL; 1861 } 1862 if (txq->ift_sds.ifsd_tso_map != NULL) { 1863 free(txq->ift_sds.ifsd_tso_map, M_IFLIB); 1864 txq->ift_sds.ifsd_tso_map = NULL; 1865 } 1866 if (txq->ift_sds.ifsd_m != NULL) { 1867 free(txq->ift_sds.ifsd_m, M_IFLIB); 1868 txq->ift_sds.ifsd_m = NULL; 1869 } 1870 if (txq->ift_buf_tag != NULL) { 1871 bus_dma_tag_destroy(txq->ift_buf_tag); 1872 txq->ift_buf_tag = NULL; 1873 } 1874 if (txq->ift_tso_buf_tag != NULL) { 1875 bus_dma_tag_destroy(txq->ift_tso_buf_tag); 1876 txq->ift_tso_buf_tag = NULL; 1877 } 1878 if (txq->ift_ifdi != NULL) { 1879 free(txq->ift_ifdi, M_IFLIB); 1880 } 1881 } 1882 1883 static void 1884 iflib_txsd_free(if_ctx_t ctx, iflib_txq_t txq, int i) 1885 { 1886 struct mbuf **mp; 1887 1888 mp = &txq->ift_sds.ifsd_m[i]; 1889 if (*mp == NULL) 1890 return; 1891 1892 if (txq->ift_sds.ifsd_map != NULL) { 1893 bus_dmamap_sync(txq->ift_buf_tag, 1894 txq->ift_sds.ifsd_map[i], BUS_DMASYNC_POSTWRITE); 1895 bus_dmamap_unload(txq->ift_buf_tag, txq->ift_sds.ifsd_map[i]); 1896 } 1897 if (txq->ift_sds.ifsd_tso_map != NULL) { 1898 bus_dmamap_sync(txq->ift_tso_buf_tag, 1899 txq->ift_sds.ifsd_tso_map[i], BUS_DMASYNC_POSTWRITE); 1900 bus_dmamap_unload(txq->ift_tso_buf_tag, 1901 txq->ift_sds.ifsd_tso_map[i]); 1902 } 1903 m_freem(*mp); 1904 DBG_COUNTER_INC(tx_frees); 1905 *mp = NULL; 1906 } 1907 1908 static int 1909 iflib_txq_setup(iflib_txq_t txq) 1910 { 1911 if_ctx_t ctx = txq->ift_ctx; 1912 if_softc_ctx_t scctx = &ctx->ifc_softc_ctx; 1913 if_shared_ctx_t sctx = ctx->ifc_sctx; 1914 iflib_dma_info_t di; 1915 int i; 1916 1917 /* Set number of descriptors available */ 1918 txq->ift_qstatus = IFLIB_QUEUE_IDLE; 1919 /* XXX make configurable */ 1920 txq->ift_update_freq = IFLIB_DEFAULT_TX_UPDATE_FREQ; 1921 1922 /* Reset indices */ 1923 txq->ift_cidx_processed = 0; 1924 txq->ift_pidx = txq->ift_cidx = txq->ift_npending = 0; 1925 txq->ift_size = scctx->isc_ntxd[txq->ift_br_offset]; 1926 1927 for (i = 0, di = txq->ift_ifdi; i < sctx->isc_ntxqs; i++, di++) 1928 bzero((void *)di->idi_vaddr, di->idi_size); 1929 1930 IFDI_TXQ_SETUP(ctx, txq->ift_id); 1931 for (i = 0, di = txq->ift_ifdi; i < sctx->isc_ntxqs; i++, di++) 1932 bus_dmamap_sync(di->idi_tag, di->idi_map, 1933 BUS_DMASYNC_PREREAD | BUS_DMASYNC_PREWRITE); 1934 return (0); 1935 } 1936 1937 /********************************************************************* 1938 * 1939 * Allocate DMA resources for RX buffers as well as memory for the RX 1940 * mbuf map, direct RX cluster pointer map and RX cluster bus address 1941 * map. RX DMA map, RX mbuf map, direct RX cluster pointer map and 1942 * RX cluster map are kept in a iflib_sw_rx_desc_array structure. 1943 * Since we use use one entry in iflib_sw_rx_desc_array per received 1944 * packet, the maximum number of entries we'll need is equal to the 1945 * number of hardware receive descriptors that we've allocated. 1946 * 1947 **********************************************************************/ 1948 static int 1949 iflib_rxsd_alloc(iflib_rxq_t rxq) 1950 { 1951 if_ctx_t ctx = rxq->ifr_ctx; 1952 if_shared_ctx_t sctx = ctx->ifc_sctx; 1953 if_softc_ctx_t scctx = &ctx->ifc_softc_ctx; 1954 device_t dev = ctx->ifc_dev; 1955 iflib_fl_t fl; 1956 bus_addr_t lowaddr; 1957 int err; 1958 1959 MPASS(scctx->isc_nrxd[0] > 0); 1960 MPASS(scctx->isc_nrxd[rxq->ifr_fl_offset] > 0); 1961 1962 lowaddr = DMA_WIDTH_TO_BUS_LOWADDR(scctx->isc_dma_width); 1963 1964 fl = rxq->ifr_fl; 1965 for (int i = 0; i < rxq->ifr_nfl; i++, fl++) { 1966 fl->ifl_size = scctx->isc_nrxd[rxq->ifr_fl_offset]; /* this isn't necessarily the same */ 1967 /* Set up DMA tag for RX buffers. */ 1968 err = bus_dma_tag_create(bus_get_dma_tag(dev), /* parent */ 1969 1, 0, /* alignment, bounds */ 1970 lowaddr, /* lowaddr */ 1971 BUS_SPACE_MAXADDR, /* highaddr */ 1972 NULL, NULL, /* filter, filterarg */ 1973 sctx->isc_rx_maxsize, /* maxsize */ 1974 sctx->isc_rx_nsegments, /* nsegments */ 1975 sctx->isc_rx_maxsegsize, /* maxsegsize */ 1976 0, /* flags */ 1977 NULL, /* lockfunc */ 1978 NULL, /* lockarg */ 1979 &fl->ifl_buf_tag); 1980 if (err) { 1981 device_printf(dev, 1982 "Unable to allocate RX DMA tag: %d\n", err); 1983 goto fail; 1984 } 1985 1986 /* Allocate memory for the RX mbuf map. */ 1987 if (!(fl->ifl_sds.ifsd_m = 1988 (struct mbuf **) malloc(sizeof(struct mbuf *) * 1989 scctx->isc_nrxd[rxq->ifr_fl_offset], M_IFLIB, M_NOWAIT | M_ZERO))) { 1990 device_printf(dev, 1991 "Unable to allocate RX mbuf map memory\n"); 1992 err = ENOMEM; 1993 goto fail; 1994 } 1995 1996 /* Allocate memory for the direct RX cluster pointer map. */ 1997 if (!(fl->ifl_sds.ifsd_cl = 1998 (caddr_t *) malloc(sizeof(caddr_t) * 1999 scctx->isc_nrxd[rxq->ifr_fl_offset], M_IFLIB, M_NOWAIT | M_ZERO))) { 2000 device_printf(dev, 2001 "Unable to allocate RX cluster map memory\n"); 2002 err = ENOMEM; 2003 goto fail; 2004 } 2005 2006 /* Allocate memory for the RX cluster bus address map. */ 2007 if (!(fl->ifl_sds.ifsd_ba = 2008 (bus_addr_t *) malloc(sizeof(bus_addr_t) * 2009 scctx->isc_nrxd[rxq->ifr_fl_offset], M_IFLIB, M_NOWAIT | M_ZERO))) { 2010 device_printf(dev, 2011 "Unable to allocate RX bus address map memory\n"); 2012 err = ENOMEM; 2013 goto fail; 2014 } 2015 2016 /* 2017 * Create the DMA maps for RX buffers. 2018 */ 2019 if (!(fl->ifl_sds.ifsd_map = 2020 (bus_dmamap_t *) malloc(sizeof(bus_dmamap_t) * scctx->isc_nrxd[rxq->ifr_fl_offset], M_IFLIB, M_NOWAIT | M_ZERO))) { 2021 device_printf(dev, 2022 "Unable to allocate RX buffer DMA map memory\n"); 2023 err = ENOMEM; 2024 goto fail; 2025 } 2026 for (int i = 0; i < scctx->isc_nrxd[rxq->ifr_fl_offset]; i++) { 2027 err = bus_dmamap_create(fl->ifl_buf_tag, 0, 2028 &fl->ifl_sds.ifsd_map[i]); 2029 if (err != 0) { 2030 device_printf(dev, "Unable to create RX buffer DMA map\n"); 2031 goto fail; 2032 } 2033 } 2034 } 2035 return (0); 2036 2037 fail: 2038 iflib_rx_structures_free(ctx); 2039 return (err); 2040 } 2041 2042 /* 2043 * Internal service routines 2044 */ 2045 2046 struct rxq_refill_cb_arg { 2047 int error; 2048 bus_dma_segment_t seg; 2049 int nseg; 2050 }; 2051 2052 static void 2053 _rxq_refill_cb(void *arg, bus_dma_segment_t *segs, int nseg, int error) 2054 { 2055 struct rxq_refill_cb_arg *cb_arg = arg; 2056 2057 cb_arg->error = error; 2058 cb_arg->seg = segs[0]; 2059 cb_arg->nseg = nseg; 2060 } 2061 2062 /** 2063 * iflib_fl_refill - refill an rxq free-buffer list 2064 * @ctx: the iflib context 2065 * @fl: the free list to refill 2066 * @count: the number of new buffers to allocate 2067 * 2068 * (Re)populate an rxq free-buffer list with up to @count new packet buffers. 2069 * The caller must assure that @count does not exceed the queue's capacity 2070 * minus one (since we always leave a descriptor unavailable). 2071 */ 2072 static uint8_t 2073 iflib_fl_refill(if_ctx_t ctx, iflib_fl_t fl, int count) 2074 { 2075 struct if_rxd_update iru; 2076 struct rxq_refill_cb_arg cb_arg; 2077 struct mbuf *m; 2078 caddr_t cl, *sd_cl; 2079 struct mbuf **sd_m; 2080 bus_dmamap_t *sd_map; 2081 bus_addr_t bus_addr, *sd_ba; 2082 int err, frag_idx, i, idx, n, pidx; 2083 qidx_t credits; 2084 2085 MPASS(count <= fl->ifl_size - fl->ifl_credits - 1); 2086 2087 sd_m = fl->ifl_sds.ifsd_m; 2088 sd_map = fl->ifl_sds.ifsd_map; 2089 sd_cl = fl->ifl_sds.ifsd_cl; 2090 sd_ba = fl->ifl_sds.ifsd_ba; 2091 pidx = fl->ifl_pidx; 2092 idx = pidx; 2093 frag_idx = fl->ifl_fragidx; 2094 credits = fl->ifl_credits; 2095 2096 i = 0; 2097 n = count; 2098 MPASS(n > 0); 2099 MPASS(credits + n <= fl->ifl_size); 2100 2101 if (pidx < fl->ifl_cidx) 2102 MPASS(pidx + n <= fl->ifl_cidx); 2103 if (pidx == fl->ifl_cidx && (credits < fl->ifl_size)) 2104 MPASS(fl->ifl_gen == 0); 2105 if (pidx > fl->ifl_cidx) 2106 MPASS(n <= fl->ifl_size - pidx + fl->ifl_cidx); 2107 2108 DBG_COUNTER_INC(fl_refills); 2109 if (n > 8) 2110 DBG_COUNTER_INC(fl_refills_large); 2111 iru_init(&iru, fl->ifl_rxq, fl->ifl_id); 2112 while (n-- > 0) { 2113 /* 2114 * We allocate an uninitialized mbuf + cluster, mbuf is 2115 * initialized after rx. 2116 * 2117 * If the cluster is still set then we know a minimum sized 2118 * packet was received 2119 */ 2120 bit_ffc_at(fl->ifl_rx_bitmap, frag_idx, fl->ifl_size, 2121 &frag_idx); 2122 if (frag_idx < 0) 2123 bit_ffc(fl->ifl_rx_bitmap, fl->ifl_size, &frag_idx); 2124 MPASS(frag_idx >= 0); 2125 if ((cl = sd_cl[frag_idx]) == NULL) { 2126 cl = uma_zalloc(fl->ifl_zone, M_NOWAIT); 2127 if (__predict_false(cl == NULL)) 2128 break; 2129 2130 cb_arg.error = 0; 2131 MPASS(sd_map != NULL); 2132 err = bus_dmamap_load(fl->ifl_buf_tag, sd_map[frag_idx], 2133 cl, fl->ifl_buf_size, _rxq_refill_cb, &cb_arg, 2134 BUS_DMA_NOWAIT); 2135 if (__predict_false(err != 0 || cb_arg.error)) { 2136 uma_zfree(fl->ifl_zone, cl); 2137 break; 2138 } 2139 2140 sd_ba[frag_idx] = bus_addr = cb_arg.seg.ds_addr; 2141 sd_cl[frag_idx] = cl; 2142 #if MEMORY_LOGGING 2143 fl->ifl_cl_enqueued++; 2144 #endif 2145 } else { 2146 bus_addr = sd_ba[frag_idx]; 2147 } 2148 bus_dmamap_sync(fl->ifl_buf_tag, sd_map[frag_idx], 2149 BUS_DMASYNC_PREREAD); 2150 2151 if (sd_m[frag_idx] == NULL) { 2152 m = m_gethdr_raw(M_NOWAIT, 0); 2153 if (__predict_false(m == NULL)) 2154 break; 2155 sd_m[frag_idx] = m; 2156 } 2157 bit_set(fl->ifl_rx_bitmap, frag_idx); 2158 #if MEMORY_LOGGING 2159 fl->ifl_m_enqueued++; 2160 #endif 2161 2162 DBG_COUNTER_INC(rx_allocs); 2163 fl->ifl_rxd_idxs[i] = frag_idx; 2164 fl->ifl_bus_addrs[i] = bus_addr; 2165 credits++; 2166 i++; 2167 MPASS(credits <= fl->ifl_size); 2168 if (++idx == fl->ifl_size) { 2169 #ifdef INVARIANTS 2170 fl->ifl_gen = 1; 2171 #endif 2172 idx = 0; 2173 } 2174 if (n == 0 || i == IFLIB_MAX_RX_REFRESH) { 2175 iru.iru_pidx = pidx; 2176 iru.iru_count = i; 2177 ctx->isc_rxd_refill(ctx->ifc_softc, &iru); 2178 fl->ifl_pidx = idx; 2179 fl->ifl_credits = credits; 2180 pidx = idx; 2181 i = 0; 2182 } 2183 } 2184 2185 if (n < count - 1) { 2186 if (i != 0) { 2187 iru.iru_pidx = pidx; 2188 iru.iru_count = i; 2189 ctx->isc_rxd_refill(ctx->ifc_softc, &iru); 2190 fl->ifl_pidx = idx; 2191 fl->ifl_credits = credits; 2192 } 2193 DBG_COUNTER_INC(rxd_flush); 2194 bus_dmamap_sync(fl->ifl_ifdi->idi_tag, fl->ifl_ifdi->idi_map, 2195 BUS_DMASYNC_PREREAD | BUS_DMASYNC_PREWRITE); 2196 ctx->isc_rxd_flush(ctx->ifc_softc, fl->ifl_rxq->ifr_id, 2197 fl->ifl_id, fl->ifl_pidx); 2198 if (__predict_true(bit_test(fl->ifl_rx_bitmap, frag_idx))) { 2199 fl->ifl_fragidx = frag_idx + 1; 2200 if (fl->ifl_fragidx == fl->ifl_size) 2201 fl->ifl_fragidx = 0; 2202 } else { 2203 fl->ifl_fragidx = frag_idx; 2204 } 2205 } 2206 2207 return (n == -1 ? 0 : IFLIB_RXEOF_EMPTY); 2208 } 2209 2210 static inline uint8_t 2211 iflib_fl_refill_all(if_ctx_t ctx, iflib_fl_t fl) 2212 { 2213 /* 2214 * We leave an unused descriptor to avoid pidx to catch up with cidx. 2215 * This is important as it confuses most NICs. For instance, 2216 * Intel NICs have (per receive ring) RDH and RDT registers, where 2217 * RDH points to the next receive descriptor to be used by the NIC, 2218 * and RDT for the next receive descriptor to be published by the 2219 * driver to the NIC (RDT - 1 is thus the last valid one). 2220 * The condition RDH == RDT means no descriptors are available to 2221 * the NIC, and thus it would be ambiguous if it also meant that 2222 * all the descriptors are available to the NIC. 2223 */ 2224 int32_t reclaimable = fl->ifl_size - fl->ifl_credits - 1; 2225 #ifdef INVARIANTS 2226 int32_t delta = fl->ifl_size - get_inuse(fl->ifl_size, fl->ifl_cidx, fl->ifl_pidx, fl->ifl_gen) - 1; 2227 #endif 2228 2229 MPASS(fl->ifl_credits <= fl->ifl_size); 2230 MPASS(reclaimable == delta); 2231 2232 if (reclaimable > 0) 2233 return (iflib_fl_refill(ctx, fl, reclaimable)); 2234 return (0); 2235 } 2236 2237 uint8_t 2238 iflib_in_detach(if_ctx_t ctx) 2239 { 2240 bool in_detach; 2241 2242 STATE_LOCK(ctx); 2243 in_detach = !!(ctx->ifc_flags & IFC_IN_DETACH); 2244 STATE_UNLOCK(ctx); 2245 return (in_detach); 2246 } 2247 2248 static void 2249 iflib_fl_bufs_free(iflib_fl_t fl) 2250 { 2251 iflib_dma_info_t idi = fl->ifl_ifdi; 2252 bus_dmamap_t sd_map; 2253 uint32_t i; 2254 2255 for (i = 0; i < fl->ifl_size; i++) { 2256 struct mbuf **sd_m = &fl->ifl_sds.ifsd_m[i]; 2257 caddr_t *sd_cl = &fl->ifl_sds.ifsd_cl[i]; 2258 2259 if (*sd_cl != NULL) { 2260 sd_map = fl->ifl_sds.ifsd_map[i]; 2261 bus_dmamap_sync(fl->ifl_buf_tag, sd_map, 2262 BUS_DMASYNC_POSTREAD); 2263 bus_dmamap_unload(fl->ifl_buf_tag, sd_map); 2264 uma_zfree(fl->ifl_zone, *sd_cl); 2265 *sd_cl = NULL; 2266 if (*sd_m != NULL) { 2267 m_init(*sd_m, M_NOWAIT, MT_DATA, 0); 2268 m_free_raw(*sd_m); 2269 *sd_m = NULL; 2270 } 2271 } else { 2272 MPASS(*sd_m == NULL); 2273 } 2274 #if MEMORY_LOGGING 2275 fl->ifl_m_dequeued++; 2276 fl->ifl_cl_dequeued++; 2277 #endif 2278 } 2279 #ifdef INVARIANTS 2280 for (i = 0; i < fl->ifl_size; i++) { 2281 MPASS(fl->ifl_sds.ifsd_cl[i] == NULL); 2282 MPASS(fl->ifl_sds.ifsd_m[i] == NULL); 2283 } 2284 #endif 2285 /* 2286 * Reset free list values 2287 */ 2288 fl->ifl_credits = fl->ifl_cidx = fl->ifl_pidx = fl->ifl_gen = fl->ifl_fragidx = 0; 2289 bzero(idi->idi_vaddr, idi->idi_size); 2290 } 2291 2292 /********************************************************************* 2293 * 2294 * Initialize a free list and its buffers. 2295 * 2296 **********************************************************************/ 2297 static int 2298 iflib_fl_setup(iflib_fl_t fl) 2299 { 2300 iflib_rxq_t rxq = fl->ifl_rxq; 2301 if_ctx_t ctx = rxq->ifr_ctx; 2302 if_softc_ctx_t scctx = &ctx->ifc_softc_ctx; 2303 int qidx; 2304 2305 bit_nclear(fl->ifl_rx_bitmap, 0, fl->ifl_size - 1); 2306 /* 2307 * Free current RX buffer structs and their mbufs 2308 */ 2309 iflib_fl_bufs_free(fl); 2310 /* Now replenish the mbufs */ 2311 MPASS(fl->ifl_credits == 0); 2312 qidx = rxq->ifr_fl_offset + fl->ifl_id; 2313 if (scctx->isc_rxd_buf_size[qidx] != 0) 2314 fl->ifl_buf_size = scctx->isc_rxd_buf_size[qidx]; 2315 else 2316 fl->ifl_buf_size = ctx->ifc_rx_mbuf_sz; 2317 /* 2318 * ifl_buf_size may be a driver-supplied value, so pull it up 2319 * to the selected mbuf size. 2320 */ 2321 fl->ifl_buf_size = iflib_get_mbuf_size_for(fl->ifl_buf_size); 2322 if (fl->ifl_buf_size > ctx->ifc_max_fl_buf_size) 2323 ctx->ifc_max_fl_buf_size = fl->ifl_buf_size; 2324 fl->ifl_cltype = m_gettype(fl->ifl_buf_size); 2325 fl->ifl_zone = m_getzone(fl->ifl_buf_size); 2326 2327 /* 2328 * Avoid pre-allocating zillions of clusters to an idle card 2329 * potentially speeding up attach. In any case make sure 2330 * to leave a descriptor unavailable. See the comment in 2331 * iflib_fl_refill_all(). 2332 */ 2333 MPASS(fl->ifl_size > 0); 2334 (void)iflib_fl_refill(ctx, fl, min(128, fl->ifl_size - 1)); 2335 if (min(128, fl->ifl_size - 1) != fl->ifl_credits) 2336 return (ENOBUFS); 2337 /* 2338 * handle failure 2339 */ 2340 MPASS(rxq != NULL); 2341 MPASS(fl->ifl_ifdi != NULL); 2342 bus_dmamap_sync(fl->ifl_ifdi->idi_tag, fl->ifl_ifdi->idi_map, 2343 BUS_DMASYNC_PREREAD | BUS_DMASYNC_PREWRITE); 2344 return (0); 2345 } 2346 2347 /********************************************************************* 2348 * 2349 * Free receive ring data structures 2350 * 2351 **********************************************************************/ 2352 static void 2353 iflib_rx_sds_free(iflib_rxq_t rxq) 2354 { 2355 iflib_fl_t fl; 2356 int i, j; 2357 2358 if (rxq->ifr_fl != NULL) { 2359 for (i = 0; i < rxq->ifr_nfl; i++) { 2360 fl = &rxq->ifr_fl[i]; 2361 if (fl->ifl_buf_tag != NULL) { 2362 if (fl->ifl_sds.ifsd_map != NULL) { 2363 for (j = 0; j < fl->ifl_size; j++) { 2364 bus_dmamap_sync( 2365 fl->ifl_buf_tag, 2366 fl->ifl_sds.ifsd_map[j], 2367 BUS_DMASYNC_POSTREAD); 2368 bus_dmamap_unload( 2369 fl->ifl_buf_tag, 2370 fl->ifl_sds.ifsd_map[j]); 2371 bus_dmamap_destroy( 2372 fl->ifl_buf_tag, 2373 fl->ifl_sds.ifsd_map[j]); 2374 } 2375 } 2376 bus_dma_tag_destroy(fl->ifl_buf_tag); 2377 fl->ifl_buf_tag = NULL; 2378 } 2379 free(fl->ifl_sds.ifsd_m, M_IFLIB); 2380 free(fl->ifl_sds.ifsd_cl, M_IFLIB); 2381 free(fl->ifl_sds.ifsd_ba, M_IFLIB); 2382 free(fl->ifl_sds.ifsd_map, M_IFLIB); 2383 free(fl->ifl_rx_bitmap, M_IFLIB); 2384 fl->ifl_sds.ifsd_m = NULL; 2385 fl->ifl_sds.ifsd_cl = NULL; 2386 fl->ifl_sds.ifsd_ba = NULL; 2387 fl->ifl_sds.ifsd_map = NULL; 2388 fl->ifl_rx_bitmap = NULL; 2389 } 2390 free(rxq->ifr_fl, M_IFLIB); 2391 rxq->ifr_fl = NULL; 2392 free(rxq->ifr_ifdi, M_IFLIB); 2393 rxq->ifr_ifdi = NULL; 2394 rxq->ifr_cq_cidx = 0; 2395 } 2396 } 2397 2398 /* 2399 * Timer routine 2400 */ 2401 static void 2402 iflib_timer(void *arg) 2403 { 2404 iflib_txq_t txq = arg; 2405 if_ctx_t ctx = txq->ift_ctx; 2406 if_softc_ctx_t sctx = &ctx->ifc_softc_ctx; 2407 uint64_t this_tick = ticks; 2408 2409 if (!(if_getdrvflags(ctx->ifc_ifp) & IFF_DRV_RUNNING)) 2410 return; 2411 2412 /* 2413 ** Check on the state of the TX queue(s), this 2414 ** can be done without the lock because its RO 2415 ** and the HUNG state will be static if set. 2416 */ 2417 if (this_tick - txq->ift_last_timer_tick >= iflib_timer_default) { 2418 txq->ift_last_timer_tick = this_tick; 2419 IFDI_TIMER(ctx, txq->ift_id); 2420 if ((txq->ift_qstatus == IFLIB_QUEUE_HUNG) && 2421 ((txq->ift_cleaned_prev == txq->ift_cleaned) || 2422 (sctx->isc_pause_frames == 0))) 2423 goto hung; 2424 2425 if (txq->ift_qstatus != IFLIB_QUEUE_IDLE && 2426 ifmp_ring_is_stalled(txq->ift_br)) { 2427 KASSERT(ctx->ifc_link_state == LINK_STATE_UP, 2428 ("queue can't be marked as hung if interface is down")); 2429 txq->ift_qstatus = IFLIB_QUEUE_HUNG; 2430 } 2431 txq->ift_cleaned_prev = txq->ift_cleaned; 2432 } 2433 /* handle any laggards */ 2434 if (txq->ift_db_pending) 2435 GROUPTASK_ENQUEUE(&txq->ift_task); 2436 2437 sctx->isc_pause_frames = 0; 2438 if (if_getdrvflags(ctx->ifc_ifp) & IFF_DRV_RUNNING) 2439 callout_reset_on(&txq->ift_timer, iflib_timer_default, iflib_timer, 2440 txq, txq->ift_timer.c_cpu); 2441 return; 2442 2443 hung: 2444 device_printf(ctx->ifc_dev, 2445 "Watchdog timeout (TX: %d desc avail: %d pidx: %d) -- resetting\n", 2446 txq->ift_id, TXQ_AVAIL(txq), txq->ift_pidx); 2447 STATE_LOCK(ctx); 2448 if_setdrvflagbits(ctx->ifc_ifp, IFF_DRV_OACTIVE, IFF_DRV_RUNNING); 2449 ctx->ifc_flags |= (IFC_DO_WATCHDOG | IFC_DO_RESET); 2450 iflib_admin_intr_deferred(ctx); 2451 STATE_UNLOCK(ctx); 2452 } 2453 2454 static uint16_t 2455 iflib_get_mbuf_size_for(unsigned int size) 2456 { 2457 2458 if (size <= MCLBYTES) 2459 return (MCLBYTES); 2460 else 2461 return (MJUMPAGESIZE); 2462 } 2463 2464 static void 2465 iflib_calc_rx_mbuf_sz(if_ctx_t ctx) 2466 { 2467 if_softc_ctx_t sctx = &ctx->ifc_softc_ctx; 2468 2469 /* 2470 * XXX don't set the max_frame_size to larger 2471 * than the hardware can handle 2472 */ 2473 ctx->ifc_rx_mbuf_sz = 2474 iflib_get_mbuf_size_for(sctx->isc_max_frame_size); 2475 } 2476 2477 uint32_t 2478 iflib_get_rx_mbuf_sz(if_ctx_t ctx) 2479 { 2480 2481 return (ctx->ifc_rx_mbuf_sz); 2482 } 2483 2484 static void 2485 iflib_init_locked(if_ctx_t ctx) 2486 { 2487 if_softc_ctx_t scctx = &ctx->ifc_softc_ctx; 2488 if_t ifp = ctx->ifc_ifp; 2489 iflib_fl_t fl; 2490 iflib_txq_t txq; 2491 iflib_rxq_t rxq; 2492 int i, j, tx_ip_csum_flags, tx_ip6_csum_flags; 2493 2494 if_setdrvflagbits(ifp, IFF_DRV_OACTIVE, IFF_DRV_RUNNING); 2495 IFDI_INTR_DISABLE(ctx); 2496 2497 /* 2498 * See iflib_stop(). Useful in case iflib_init_locked() is 2499 * called without first calling iflib_stop(). 2500 */ 2501 netmap_disable_all_rings(ifp); 2502 2503 tx_ip_csum_flags = scctx->isc_tx_csum_flags & (CSUM_IP | CSUM_TCP | CSUM_UDP | CSUM_SCTP); 2504 tx_ip6_csum_flags = scctx->isc_tx_csum_flags & (CSUM_IP6_TCP | CSUM_IP6_UDP | CSUM_IP6_SCTP); 2505 /* Set hardware offload abilities */ 2506 if_clearhwassist(ifp); 2507 if (if_getcapenable(ifp) & IFCAP_TXCSUM) 2508 if_sethwassistbits(ifp, tx_ip_csum_flags, 0); 2509 if (if_getcapenable(ifp) & IFCAP_TXCSUM_IPV6) 2510 if_sethwassistbits(ifp, tx_ip6_csum_flags, 0); 2511 if (if_getcapenable(ifp) & IFCAP_TSO4) 2512 if_sethwassistbits(ifp, CSUM_IP_TSO, 0); 2513 if (if_getcapenable(ifp) & IFCAP_TSO6) 2514 if_sethwassistbits(ifp, CSUM_IP6_TSO, 0); 2515 2516 for (i = 0, txq = ctx->ifc_txqs; i < scctx->isc_ntxqsets; i++, txq++) { 2517 CALLOUT_LOCK(txq); 2518 callout_stop(&txq->ift_timer); 2519 #ifdef DEV_NETMAP 2520 callout_stop(&txq->ift_netmap_timer); 2521 #endif /* DEV_NETMAP */ 2522 CALLOUT_UNLOCK(txq); 2523 (void)iflib_netmap_txq_init(ctx, txq); 2524 } 2525 2526 /* 2527 * Calculate a suitable Rx mbuf size prior to calling IFDI_INIT, so 2528 * that drivers can use the value when setting up the hardware receive 2529 * buffers. 2530 */ 2531 iflib_calc_rx_mbuf_sz(ctx); 2532 2533 #ifdef INVARIANTS 2534 i = if_getdrvflags(ifp); 2535 #endif 2536 IFDI_INIT(ctx); 2537 MPASS(if_getdrvflags(ifp) == i); 2538 for (i = 0, rxq = ctx->ifc_rxqs; i < scctx->isc_nrxqsets; i++, rxq++) { 2539 if (iflib_netmap_rxq_init(ctx, rxq) > 0) { 2540 /* This rxq is in netmap mode. Skip normal init. */ 2541 continue; 2542 } 2543 for (j = 0, fl = rxq->ifr_fl; j < rxq->ifr_nfl; j++, fl++) { 2544 if (iflib_fl_setup(fl)) { 2545 device_printf(ctx->ifc_dev, 2546 "setting up free list %d failed - " 2547 "check cluster settings\n", j); 2548 goto done; 2549 } 2550 } 2551 } 2552 done: 2553 if_setdrvflagbits(ctx->ifc_ifp, IFF_DRV_RUNNING, IFF_DRV_OACTIVE); 2554 IFDI_INTR_ENABLE(ctx); 2555 txq = ctx->ifc_txqs; 2556 for (i = 0; i < scctx->isc_ntxqsets; i++, txq++) 2557 callout_reset_on(&txq->ift_timer, iflib_timer_default, iflib_timer, txq, 2558 txq->ift_timer.c_cpu); 2559 2560 /* Re-enable txsync/rxsync. */ 2561 netmap_enable_all_rings(ifp); 2562 } 2563 2564 static int 2565 iflib_media_change(if_t ifp) 2566 { 2567 if_ctx_t ctx = if_getsoftc(ifp); 2568 int err; 2569 2570 CTX_LOCK(ctx); 2571 if ((err = IFDI_MEDIA_CHANGE(ctx)) == 0) 2572 iflib_if_init_locked(ctx); 2573 CTX_UNLOCK(ctx); 2574 return (err); 2575 } 2576 2577 static void 2578 iflib_media_status(if_t ifp, struct ifmediareq *ifmr) 2579 { 2580 if_ctx_t ctx = if_getsoftc(ifp); 2581 2582 CTX_LOCK(ctx); 2583 IFDI_UPDATE_ADMIN_STATUS(ctx); 2584 IFDI_MEDIA_STATUS(ctx, ifmr); 2585 CTX_UNLOCK(ctx); 2586 } 2587 2588 static void 2589 iflib_stop(if_ctx_t ctx) 2590 { 2591 iflib_txq_t txq = ctx->ifc_txqs; 2592 iflib_rxq_t rxq = ctx->ifc_rxqs; 2593 if_softc_ctx_t scctx = &ctx->ifc_softc_ctx; 2594 if_shared_ctx_t sctx = ctx->ifc_sctx; 2595 iflib_dma_info_t di; 2596 iflib_fl_t fl; 2597 int i, j; 2598 2599 /* Tell the stack that the interface is no longer active */ 2600 if_setdrvflagbits(ctx->ifc_ifp, IFF_DRV_OACTIVE, IFF_DRV_RUNNING); 2601 2602 IFDI_INTR_DISABLE(ctx); 2603 DELAY(1000); 2604 IFDI_STOP(ctx); 2605 DELAY(1000); 2606 2607 /* 2608 * Stop any pending txsync/rxsync and prevent new ones 2609 * form starting. Processes blocked in poll() will get 2610 * POLLERR. 2611 */ 2612 netmap_disable_all_rings(ctx->ifc_ifp); 2613 2614 iflib_debug_reset(); 2615 /* Wait for current tx queue users to exit to disarm watchdog timer. */ 2616 for (i = 0; i < scctx->isc_ntxqsets; i++, txq++) { 2617 /* make sure all transmitters have completed before proceeding XXX */ 2618 2619 CALLOUT_LOCK(txq); 2620 callout_stop(&txq->ift_timer); 2621 #ifdef DEV_NETMAP 2622 callout_stop(&txq->ift_netmap_timer); 2623 #endif /* DEV_NETMAP */ 2624 CALLOUT_UNLOCK(txq); 2625 2626 /* clean any enqueued buffers */ 2627 iflib_ifmp_purge(txq); 2628 /* Free any existing tx buffers. */ 2629 for (j = 0; j < txq->ift_size; j++) { 2630 iflib_txsd_free(ctx, txq, j); 2631 } 2632 txq->ift_processed = txq->ift_cleaned = txq->ift_cidx_processed = 0; 2633 txq->ift_in_use = txq->ift_gen = txq->ift_no_desc_avail = 0; 2634 if (sctx->isc_flags & IFLIB_PRESERVE_TX_INDICES) 2635 txq->ift_cidx = txq->ift_pidx; 2636 else 2637 txq->ift_cidx = txq->ift_pidx = 0; 2638 2639 txq->ift_closed = txq->ift_mbuf_defrag = txq->ift_mbuf_defrag_failed = 0; 2640 txq->ift_no_tx_dma_setup = txq->ift_txd_encap_efbig = txq->ift_map_failed = 0; 2641 txq->ift_pullups = 0; 2642 ifmp_ring_reset_stats(txq->ift_br); 2643 for (j = 0, di = txq->ift_ifdi; j < sctx->isc_ntxqs; j++, di++) 2644 bzero((void *)di->idi_vaddr, di->idi_size); 2645 } 2646 for (i = 0; i < scctx->isc_nrxqsets; i++, rxq++) { 2647 if (rxq->ifr_task.gt_taskqueue != NULL) 2648 gtaskqueue_drain(rxq->ifr_task.gt_taskqueue, 2649 &rxq->ifr_task.gt_task); 2650 2651 rxq->ifr_cq_cidx = 0; 2652 for (j = 0, di = rxq->ifr_ifdi; j < sctx->isc_nrxqs; j++, di++) 2653 bzero((void *)di->idi_vaddr, di->idi_size); 2654 /* also resets the free lists pidx/cidx */ 2655 for (j = 0, fl = rxq->ifr_fl; j < rxq->ifr_nfl; j++, fl++) 2656 iflib_fl_bufs_free(fl); 2657 } 2658 } 2659 2660 static inline caddr_t 2661 calc_next_rxd(iflib_fl_t fl, int cidx) 2662 { 2663 qidx_t size; 2664 int nrxd; 2665 caddr_t start, end, cur, next; 2666 2667 nrxd = fl->ifl_size; 2668 size = fl->ifl_rxd_size; 2669 start = fl->ifl_ifdi->idi_vaddr; 2670 2671 if (__predict_false(size == 0)) 2672 return (start); 2673 cur = start + size * cidx; 2674 end = start + size * nrxd; 2675 next = CACHE_PTR_NEXT(cur); 2676 return (next < end ? next : start); 2677 } 2678 2679 static inline void 2680 prefetch_pkts(iflib_fl_t fl, int cidx) 2681 { 2682 int nextptr; 2683 int nrxd = fl->ifl_size; 2684 caddr_t next_rxd; 2685 2686 nextptr = (cidx + CACHE_PTR_INCREMENT) & (nrxd - 1); 2687 prefetch(&fl->ifl_sds.ifsd_m[nextptr]); 2688 prefetch(&fl->ifl_sds.ifsd_cl[nextptr]); 2689 next_rxd = calc_next_rxd(fl, cidx); 2690 prefetch(next_rxd); 2691 prefetch(fl->ifl_sds.ifsd_m[(cidx + 1) & (nrxd - 1)]); 2692 prefetch(fl->ifl_sds.ifsd_m[(cidx + 2) & (nrxd - 1)]); 2693 prefetch(fl->ifl_sds.ifsd_m[(cidx + 3) & (nrxd - 1)]); 2694 prefetch(fl->ifl_sds.ifsd_m[(cidx + 4) & (nrxd - 1)]); 2695 prefetch(fl->ifl_sds.ifsd_cl[(cidx + 1) & (nrxd - 1)]); 2696 prefetch(fl->ifl_sds.ifsd_cl[(cidx + 2) & (nrxd - 1)]); 2697 prefetch(fl->ifl_sds.ifsd_cl[(cidx + 3) & (nrxd - 1)]); 2698 prefetch(fl->ifl_sds.ifsd_cl[(cidx + 4) & (nrxd - 1)]); 2699 } 2700 2701 static struct mbuf * 2702 rxd_frag_to_sd(iflib_rxq_t rxq, if_rxd_frag_t irf, bool unload, if_rxsd_t sd, 2703 int *pf_rv, if_rxd_info_t ri) 2704 { 2705 bus_dmamap_t map; 2706 iflib_fl_t fl; 2707 caddr_t payload; 2708 struct mbuf *m; 2709 int flid, cidx, len, next; 2710 2711 map = NULL; 2712 flid = irf->irf_flid; 2713 cidx = irf->irf_idx; 2714 fl = &rxq->ifr_fl[flid]; 2715 sd->ifsd_fl = fl; 2716 sd->ifsd_cl = &fl->ifl_sds.ifsd_cl[cidx]; 2717 fl->ifl_credits--; 2718 #if MEMORY_LOGGING 2719 fl->ifl_m_dequeued++; 2720 #endif 2721 if (rxq->ifr_ctx->ifc_flags & IFC_PREFETCH) 2722 prefetch_pkts(fl, cidx); 2723 next = (cidx + CACHE_PTR_INCREMENT) & (fl->ifl_size - 1); 2724 prefetch(&fl->ifl_sds.ifsd_map[next]); 2725 map = fl->ifl_sds.ifsd_map[cidx]; 2726 2727 bus_dmamap_sync(fl->ifl_buf_tag, map, BUS_DMASYNC_POSTREAD); 2728 2729 if (rxq->pfil != NULL && PFIL_HOOKED_IN(rxq->pfil) && pf_rv != NULL && 2730 irf->irf_len != 0) { 2731 payload = *sd->ifsd_cl; 2732 payload += ri->iri_pad; 2733 len = ri->iri_len - ri->iri_pad; 2734 *pf_rv = pfil_mem_in(rxq->pfil, payload, len, ri->iri_ifp, &m); 2735 switch (*pf_rv) { 2736 case PFIL_DROPPED: 2737 case PFIL_CONSUMED: 2738 /* 2739 * The filter ate it. Everything is recycled. 2740 */ 2741 m = NULL; 2742 unload = 0; 2743 break; 2744 case PFIL_REALLOCED: 2745 /* 2746 * The filter copied it. Everything is recycled. 2747 * 'm' points at new mbuf. 2748 */ 2749 unload = 0; 2750 break; 2751 case PFIL_PASS: 2752 /* 2753 * Filter said it was OK, so receive like 2754 * normal 2755 */ 2756 m = fl->ifl_sds.ifsd_m[cidx]; 2757 fl->ifl_sds.ifsd_m[cidx] = NULL; 2758 break; 2759 default: 2760 MPASS(0); 2761 } 2762 } else { 2763 m = fl->ifl_sds.ifsd_m[cidx]; 2764 fl->ifl_sds.ifsd_m[cidx] = NULL; 2765 if (pf_rv != NULL) 2766 *pf_rv = PFIL_PASS; 2767 } 2768 2769 if (unload && irf->irf_len != 0) 2770 bus_dmamap_unload(fl->ifl_buf_tag, map); 2771 fl->ifl_cidx = (fl->ifl_cidx + 1) & (fl->ifl_size - 1); 2772 if (__predict_false(fl->ifl_cidx == 0)) 2773 fl->ifl_gen = 0; 2774 bit_clear(fl->ifl_rx_bitmap, cidx); 2775 return (m); 2776 } 2777 2778 static struct mbuf * 2779 assemble_segments(iflib_rxq_t rxq, if_rxd_info_t ri, if_rxsd_t sd, int *pf_rv) 2780 { 2781 struct mbuf *m, *mh, *mt; 2782 caddr_t cl; 2783 int *pf_rv_ptr, flags, i, padlen; 2784 bool consumed; 2785 2786 i = 0; 2787 mh = NULL; 2788 consumed = false; 2789 *pf_rv = PFIL_PASS; 2790 pf_rv_ptr = pf_rv; 2791 do { 2792 m = rxd_frag_to_sd(rxq, &ri->iri_frags[i], !consumed, sd, 2793 pf_rv_ptr, ri); 2794 2795 MPASS(*sd->ifsd_cl != NULL); 2796 2797 /* 2798 * Exclude zero-length frags & frags from 2799 * packets the filter has consumed or dropped 2800 */ 2801 if (ri->iri_frags[i].irf_len == 0 || consumed || 2802 *pf_rv == PFIL_CONSUMED || *pf_rv == PFIL_DROPPED) { 2803 if (mh == NULL) { 2804 /* everything saved here */ 2805 consumed = true; 2806 pf_rv_ptr = NULL; 2807 continue; 2808 } 2809 /* XXX we can save the cluster here, but not the mbuf */ 2810 m_init(m, M_NOWAIT, MT_DATA, 0); 2811 m_free(m); 2812 continue; 2813 } 2814 if (mh == NULL) { 2815 flags = M_PKTHDR | M_EXT; 2816 mh = mt = m; 2817 padlen = ri->iri_pad; 2818 } else { 2819 flags = M_EXT; 2820 mt->m_next = m; 2821 mt = m; 2822 /* assuming padding is only on the first fragment */ 2823 padlen = 0; 2824 } 2825 cl = *sd->ifsd_cl; 2826 *sd->ifsd_cl = NULL; 2827 2828 /* Can these two be made one ? */ 2829 m_init(m, M_NOWAIT, MT_DATA, flags); 2830 m_cljset(m, cl, sd->ifsd_fl->ifl_cltype); 2831 /* 2832 * These must follow m_init and m_cljset 2833 */ 2834 m->m_data += padlen; 2835 ri->iri_len -= padlen; 2836 m->m_len = ri->iri_frags[i].irf_len; 2837 } while (++i < ri->iri_nfrags); 2838 2839 return (mh); 2840 } 2841 2842 /* 2843 * Process one software descriptor 2844 */ 2845 static struct mbuf * 2846 iflib_rxd_pkt_get(iflib_rxq_t rxq, if_rxd_info_t ri) 2847 { 2848 struct if_rxsd sd; 2849 struct mbuf *m; 2850 int pf_rv; 2851 2852 /* should I merge this back in now that the two paths are basically duplicated? */ 2853 if (ri->iri_nfrags == 1 && 2854 ri->iri_frags[0].irf_len != 0 && 2855 ri->iri_frags[0].irf_len <= MIN(IFLIB_RX_COPY_THRESH, MHLEN)) { 2856 m = rxd_frag_to_sd(rxq, &ri->iri_frags[0], false, &sd, 2857 &pf_rv, ri); 2858 if (pf_rv != PFIL_PASS && pf_rv != PFIL_REALLOCED) 2859 return (m); 2860 if (pf_rv == PFIL_PASS) { 2861 m_init(m, M_NOWAIT, MT_DATA, M_PKTHDR); 2862 #ifndef __NO_STRICT_ALIGNMENT 2863 if (!IP_ALIGNED(m) && ri->iri_pad == 0) 2864 m->m_data += 2; 2865 #endif 2866 memcpy(m->m_data, *sd.ifsd_cl, ri->iri_len); 2867 m->m_len = ri->iri_frags[0].irf_len; 2868 m->m_data += ri->iri_pad; 2869 ri->iri_len -= ri->iri_pad; 2870 } 2871 } else { 2872 m = assemble_segments(rxq, ri, &sd, &pf_rv); 2873 if (m == NULL) 2874 return (NULL); 2875 if (pf_rv != PFIL_PASS && pf_rv != PFIL_REALLOCED) 2876 return (m); 2877 } 2878 m->m_pkthdr.len = ri->iri_len; 2879 m->m_pkthdr.rcvif = ri->iri_ifp; 2880 m->m_flags |= ri->iri_flags; 2881 m->m_pkthdr.ether_vtag = ri->iri_vtag; 2882 m->m_pkthdr.flowid = ri->iri_flowid; 2883 M_HASHTYPE_SET(m, ri->iri_rsstype); 2884 m->m_pkthdr.csum_flags = ri->iri_csum_flags; 2885 m->m_pkthdr.csum_data = ri->iri_csum_data; 2886 return (m); 2887 } 2888 2889 #if defined(INET6) || defined(INET) 2890 static void 2891 iflib_get_ip_forwarding(struct lro_ctrl *lc, bool *v4, bool *v6) 2892 { 2893 CURVNET_SET(if_getvnet(lc->ifp)); 2894 #if defined(INET6) 2895 *v6 = V_ip6_forwarding; 2896 #endif 2897 #if defined(INET) 2898 *v4 = V_ipforwarding; 2899 #endif 2900 CURVNET_RESTORE(); 2901 } 2902 2903 /* 2904 * Returns true if it's possible this packet could be LROed. 2905 * if it returns false, it is guaranteed that tcp_lro_rx() 2906 * would not return zero. 2907 */ 2908 static bool 2909 iflib_check_lro_possible(struct mbuf *m, bool v4_forwarding, bool v6_forwarding) 2910 { 2911 struct ether_header *eh; 2912 2913 eh = mtod(m, struct ether_header *); 2914 switch (eh->ether_type) { 2915 #if defined(INET6) 2916 case htons(ETHERTYPE_IPV6): 2917 return (!v6_forwarding); 2918 #endif 2919 #if defined(INET) 2920 case htons(ETHERTYPE_IP): 2921 return (!v4_forwarding); 2922 #endif 2923 } 2924 2925 return (false); 2926 } 2927 #else 2928 static void 2929 iflib_get_ip_forwarding(struct lro_ctrl *lc __unused, bool *v4 __unused, bool *v6 __unused) 2930 { 2931 } 2932 #endif 2933 2934 static void 2935 _task_fn_rx_watchdog(void *context) 2936 { 2937 iflib_rxq_t rxq = context; 2938 2939 GROUPTASK_ENQUEUE(&rxq->ifr_task); 2940 } 2941 2942 static uint8_t 2943 iflib_rxeof(iflib_rxq_t rxq, qidx_t budget) 2944 { 2945 if_t ifp; 2946 if_ctx_t ctx = rxq->ifr_ctx; 2947 if_shared_ctx_t sctx = ctx->ifc_sctx; 2948 if_softc_ctx_t scctx = &ctx->ifc_softc_ctx; 2949 int avail, i; 2950 qidx_t *cidxp; 2951 struct if_rxd_info ri; 2952 int err, budget_left, rx_bytes, rx_pkts; 2953 iflib_fl_t fl; 2954 int lro_enabled; 2955 bool v4_forwarding, v6_forwarding, lro_possible; 2956 uint8_t retval = 0; 2957 2958 /* 2959 * XXX early demux data packets so that if_input processing only handles 2960 * acks in interrupt context 2961 */ 2962 struct mbuf *m, *mh, *mt, *mf; 2963 2964 NET_EPOCH_ASSERT(); 2965 2966 lro_possible = v4_forwarding = v6_forwarding = false; 2967 ifp = ctx->ifc_ifp; 2968 mh = mt = NULL; 2969 MPASS(budget > 0); 2970 rx_pkts = rx_bytes = 0; 2971 if (sctx->isc_flags & IFLIB_HAS_RXCQ) 2972 cidxp = &rxq->ifr_cq_cidx; 2973 else 2974 cidxp = &rxq->ifr_fl[0].ifl_cidx; 2975 if ((avail = iflib_rxd_avail(ctx, rxq, *cidxp, budget)) == 0) { 2976 for (i = 0, fl = &rxq->ifr_fl[0]; i < sctx->isc_nfl; i++, fl++) 2977 retval |= iflib_fl_refill_all(ctx, fl); 2978 DBG_COUNTER_INC(rx_unavail); 2979 return (retval); 2980 } 2981 2982 /* pfil needs the vnet to be set */ 2983 CURVNET_SET_QUIET(if_getvnet(ifp)); 2984 for (budget_left = budget; budget_left > 0 && avail > 0;) { 2985 if (__predict_false(!CTX_ACTIVE(ctx))) { 2986 DBG_COUNTER_INC(rx_ctx_inactive); 2987 break; 2988 } 2989 /* 2990 * Reset client set fields to their default values 2991 */ 2992 rxd_info_zero(&ri); 2993 ri.iri_qsidx = rxq->ifr_id; 2994 ri.iri_cidx = *cidxp; 2995 ri.iri_ifp = ifp; 2996 ri.iri_frags = rxq->ifr_frags; 2997 err = ctx->isc_rxd_pkt_get(ctx->ifc_softc, &ri); 2998 2999 if (err) 3000 goto err; 3001 rx_pkts += 1; 3002 rx_bytes += ri.iri_len; 3003 if (sctx->isc_flags & IFLIB_HAS_RXCQ) { 3004 *cidxp = ri.iri_cidx; 3005 /* Update our consumer index */ 3006 /* XXX NB: shurd - check if this is still safe */ 3007 while (rxq->ifr_cq_cidx >= scctx->isc_nrxd[0]) 3008 rxq->ifr_cq_cidx -= scctx->isc_nrxd[0]; 3009 /* was this only a completion queue message? */ 3010 if (__predict_false(ri.iri_nfrags == 0)) 3011 continue; 3012 } 3013 MPASS(ri.iri_nfrags != 0); 3014 MPASS(ri.iri_len != 0); 3015 3016 /* will advance the cidx on the corresponding free lists */ 3017 m = iflib_rxd_pkt_get(rxq, &ri); 3018 avail--; 3019 budget_left--; 3020 if (avail == 0 && budget_left) 3021 avail = iflib_rxd_avail(ctx, rxq, *cidxp, budget_left); 3022 3023 if (__predict_false(m == NULL)) 3024 continue; 3025 3026 /* imm_pkt: -- cxgb */ 3027 if (mh == NULL) 3028 mh = mt = m; 3029 else { 3030 mt->m_nextpkt = m; 3031 mt = m; 3032 } 3033 } 3034 CURVNET_RESTORE(); 3035 /* make sure that we can refill faster than drain */ 3036 for (i = 0, fl = &rxq->ifr_fl[0]; i < sctx->isc_nfl; i++, fl++) 3037 retval |= iflib_fl_refill_all(ctx, fl); 3038 3039 lro_enabled = (if_getcapenable(ifp) & IFCAP_LRO); 3040 if (lro_enabled) 3041 iflib_get_ip_forwarding(&rxq->ifr_lc, &v4_forwarding, &v6_forwarding); 3042 mt = mf = NULL; 3043 while (mh != NULL) { 3044 m = mh; 3045 mh = mh->m_nextpkt; 3046 m->m_nextpkt = NULL; 3047 #ifndef __NO_STRICT_ALIGNMENT 3048 if (!IP_ALIGNED(m) && (m = iflib_fixup_rx(m)) == NULL) 3049 continue; 3050 #endif 3051 #if defined(INET6) || defined(INET) 3052 if (lro_enabled) { 3053 if (!lro_possible) { 3054 lro_possible = iflib_check_lro_possible(m, v4_forwarding, v6_forwarding); 3055 if (lro_possible && mf != NULL) { 3056 if_input(ifp, mf); 3057 DBG_COUNTER_INC(rx_if_input); 3058 mt = mf = NULL; 3059 } 3060 } 3061 if ((m->m_pkthdr.csum_flags & (CSUM_L4_CALC | CSUM_L4_VALID)) == 3062 (CSUM_L4_CALC | CSUM_L4_VALID)) { 3063 if (lro_possible && tcp_lro_rx(&rxq->ifr_lc, m, 0) == 0) 3064 continue; 3065 } 3066 } 3067 #endif 3068 if (lro_possible) { 3069 if_input(ifp, m); 3070 DBG_COUNTER_INC(rx_if_input); 3071 continue; 3072 } 3073 3074 if (mf == NULL) 3075 mf = m; 3076 if (mt != NULL) 3077 mt->m_nextpkt = m; 3078 mt = m; 3079 } 3080 if (mf != NULL) { 3081 if_input(ifp, mf); 3082 DBG_COUNTER_INC(rx_if_input); 3083 } 3084 3085 if_inc_counter(ifp, IFCOUNTER_IBYTES, rx_bytes); 3086 if_inc_counter(ifp, IFCOUNTER_IPACKETS, rx_pkts); 3087 3088 /* 3089 * Flush any outstanding LRO work 3090 */ 3091 #if defined(INET6) || defined(INET) 3092 tcp_lro_flush_all(&rxq->ifr_lc); 3093 #endif 3094 if (avail != 0 || iflib_rxd_avail(ctx, rxq, *cidxp, 1) != 0) 3095 retval |= IFLIB_RXEOF_MORE; 3096 return (retval); 3097 err: 3098 STATE_LOCK(ctx); 3099 ctx->ifc_flags |= IFC_DO_RESET; 3100 iflib_admin_intr_deferred(ctx); 3101 STATE_UNLOCK(ctx); 3102 return (0); 3103 } 3104 3105 #define TXD_NOTIFY_COUNT(txq) (((txq)->ift_size / (txq)->ift_update_freq) - 1) 3106 static inline qidx_t 3107 txq_max_db_deferred(iflib_txq_t txq, qidx_t in_use) 3108 { 3109 qidx_t notify_count = TXD_NOTIFY_COUNT(txq); 3110 qidx_t minthresh = txq->ift_size / 8; 3111 if (in_use > 4 * minthresh) 3112 return (notify_count); 3113 if (in_use > 2 * minthresh) 3114 return (notify_count >> 1); 3115 if (in_use > minthresh) 3116 return (notify_count >> 3); 3117 return (0); 3118 } 3119 3120 static inline qidx_t 3121 txq_max_rs_deferred(iflib_txq_t txq) 3122 { 3123 qidx_t notify_count = TXD_NOTIFY_COUNT(txq); 3124 qidx_t minthresh = txq->ift_size / 8; 3125 if (txq->ift_in_use > 4 * minthresh) 3126 return (notify_count); 3127 if (txq->ift_in_use > 2 * minthresh) 3128 return (notify_count >> 1); 3129 if (txq->ift_in_use > minthresh) 3130 return (notify_count >> 2); 3131 return (2); 3132 } 3133 3134 #define M_CSUM_FLAGS(m) ((m)->m_pkthdr.csum_flags) 3135 #define M_HAS_VLANTAG(m) (m->m_flags & M_VLANTAG) 3136 3137 #define TXQ_MAX_DB_DEFERRED(txq, in_use) txq_max_db_deferred((txq), (in_use)) 3138 #define TXQ_MAX_RS_DEFERRED(txq) txq_max_rs_deferred(txq) 3139 #define TXQ_MAX_DB_CONSUMED(size) (size >> 4) 3140 3141 /* forward compatibility for cxgb */ 3142 #define FIRST_QSET(ctx) 0 3143 #define NTXQSETS(ctx) ((ctx)->ifc_softc_ctx.isc_ntxqsets) 3144 #define NRXQSETS(ctx) ((ctx)->ifc_softc_ctx.isc_nrxqsets) 3145 #define QIDX(ctx, m) ((((m)->m_pkthdr.flowid & ctx->ifc_softc_ctx.isc_rss_table_mask) % NTXQSETS(ctx)) + FIRST_QSET(ctx)) 3146 #define DESC_RECLAIMABLE(q) ((int)((q)->ift_processed - (q)->ift_cleaned - (q)->ift_ctx->ifc_softc_ctx.isc_tx_nsegments)) 3147 3148 /* XXX we should be setting this to something other than zero */ 3149 #define RECLAIM_THRESH(ctx) ((ctx)->ifc_sctx->isc_tx_reclaim_thresh) 3150 #define MAX_TX_DESC(ctx) MAX((ctx)->ifc_softc_ctx.isc_tx_tso_segments_max, \ 3151 (ctx)->ifc_softc_ctx.isc_tx_nsegments) 3152 3153 static inline bool 3154 iflib_txd_db_check(iflib_txq_t txq, int ring) 3155 { 3156 if_ctx_t ctx = txq->ift_ctx; 3157 qidx_t dbval, max; 3158 3159 max = TXQ_MAX_DB_DEFERRED(txq, txq->ift_in_use); 3160 3161 /* force || threshold exceeded || at the edge of the ring */ 3162 if (ring || (txq->ift_db_pending >= max) || (TXQ_AVAIL(txq) <= MAX_TX_DESC(ctx) + 2)) { 3163 3164 /* 3165 * 'npending' is used if the card's doorbell is in terms of the number of descriptors 3166 * pending flush (BRCM). 'pidx' is used in cases where the card's doorbeel uses the 3167 * producer index explicitly (INTC). 3168 */ 3169 dbval = txq->ift_npending ? txq->ift_npending : txq->ift_pidx; 3170 bus_dmamap_sync(txq->ift_ifdi->idi_tag, txq->ift_ifdi->idi_map, 3171 BUS_DMASYNC_PREREAD | BUS_DMASYNC_PREWRITE); 3172 ctx->isc_txd_flush(ctx->ifc_softc, txq->ift_id, dbval); 3173 3174 /* 3175 * Absent bugs there are zero packets pending so reset pending counts to zero. 3176 */ 3177 txq->ift_db_pending = txq->ift_npending = 0; 3178 return (true); 3179 } 3180 return (false); 3181 } 3182 3183 #ifdef PKT_DEBUG 3184 static void 3185 print_pkt(if_pkt_info_t pi) 3186 { 3187 printf("pi len: %d qsidx: %d nsegs: %d ndescs: %d flags: %x pidx: %d\n", 3188 pi->ipi_len, pi->ipi_qsidx, pi->ipi_nsegs, pi->ipi_ndescs, pi->ipi_flags, pi->ipi_pidx); 3189 printf("pi new_pidx: %d csum_flags: %lx tso_segsz: %d mflags: %x vtag: %d\n", 3190 pi->ipi_new_pidx, pi->ipi_csum_flags, pi->ipi_tso_segsz, pi->ipi_mflags, pi->ipi_vtag); 3191 printf("pi etype: %d ehdrlen: %d ip_hlen: %d ipproto: %d\n", 3192 pi->ipi_etype, pi->ipi_ehdrlen, pi->ipi_ip_hlen, pi->ipi_ipproto); 3193 } 3194 #endif 3195 3196 #define IS_TSO4(pi) ((pi)->ipi_csum_flags & CSUM_IP_TSO) 3197 #define IS_TX_OFFLOAD4(pi) ((pi)->ipi_csum_flags & (CSUM_IP_TCP | CSUM_IP_TSO)) 3198 #define IS_TSO6(pi) ((pi)->ipi_csum_flags & CSUM_IP6_TSO) 3199 #define IS_TX_OFFLOAD6(pi) ((pi)->ipi_csum_flags & (CSUM_IP6_TCP | CSUM_IP6_TSO)) 3200 3201 /** 3202 * Parses out ethernet header information in the given mbuf. 3203 * Returns in pi: ipi_etype (EtherType) and ipi_ehdrlen (Ethernet header length) 3204 * 3205 * This will account for the VLAN header if present. 3206 * 3207 * XXX: This doesn't handle QinQ, which could prevent TX offloads for those 3208 * types of packets. 3209 */ 3210 static int 3211 iflib_parse_ether_header(if_pkt_info_t pi, struct mbuf **mp, uint64_t *pullups) 3212 { 3213 struct ether_vlan_header *eh; 3214 struct mbuf *m; 3215 3216 m = *mp; 3217 if (__predict_false(m->m_len < sizeof(*eh))) { 3218 (*pullups)++; 3219 if (__predict_false((m = m_pullup(m, sizeof(*eh))) == NULL)) 3220 return (ENOMEM); 3221 } 3222 eh = mtod(m, struct ether_vlan_header *); 3223 if (eh->evl_encap_proto == htons(ETHERTYPE_VLAN)) { 3224 pi->ipi_etype = ntohs(eh->evl_proto); 3225 pi->ipi_ehdrlen = ETHER_HDR_LEN + ETHER_VLAN_ENCAP_LEN; 3226 } else { 3227 pi->ipi_etype = ntohs(eh->evl_encap_proto); 3228 pi->ipi_ehdrlen = ETHER_HDR_LEN; 3229 } 3230 *mp = m; 3231 3232 return (0); 3233 } 3234 3235 /** 3236 * Parse up to the L3 header and extract IPv4/IPv6 header information into pi. 3237 * Currently this information includes: IP ToS value, IP header version/presence 3238 * 3239 * This is missing some checks and doesn't edit the packet content as it goes, 3240 * unlike iflib_parse_header(), in order to keep the amount of code here minimal. 3241 */ 3242 static int 3243 iflib_parse_header_partial(if_pkt_info_t pi, struct mbuf **mp, uint64_t *pullups) 3244 { 3245 struct mbuf *m; 3246 int err; 3247 3248 *pullups = 0; 3249 m = *mp; 3250 if (!M_WRITABLE(m)) { 3251 if ((m = m_dup(m, M_NOWAIT)) == NULL) { 3252 return (ENOMEM); 3253 } else { 3254 m_freem(*mp); 3255 DBG_COUNTER_INC(tx_frees); 3256 *mp = m; 3257 } 3258 } 3259 3260 /* Fills out pi->ipi_etype */ 3261 err = iflib_parse_ether_header(pi, mp, pullups); 3262 if (err) 3263 return (err); 3264 m = *mp; 3265 3266 switch (pi->ipi_etype) { 3267 #ifdef INET 3268 case ETHERTYPE_IP: 3269 { 3270 struct mbuf *n; 3271 struct ip *ip = NULL; 3272 int miniplen; 3273 3274 miniplen = min(m->m_pkthdr.len, pi->ipi_ehdrlen + sizeof(*ip)); 3275 if (__predict_false(m->m_len < miniplen)) { 3276 /* 3277 * Check for common case where the first mbuf only contains 3278 * the Ethernet header 3279 */ 3280 if (m->m_len == pi->ipi_ehdrlen) { 3281 n = m->m_next; 3282 MPASS(n); 3283 /* If next mbuf contains at least the minimal IP header, then stop */ 3284 if (n->m_len >= sizeof(*ip)) { 3285 ip = (struct ip *)n->m_data; 3286 } else { 3287 (*pullups)++; 3288 if (__predict_false((m = m_pullup(m, miniplen)) == NULL)) 3289 return (ENOMEM); 3290 ip = (struct ip *)(m->m_data + pi->ipi_ehdrlen); 3291 } 3292 } else { 3293 (*pullups)++; 3294 if (__predict_false((m = m_pullup(m, miniplen)) == NULL)) 3295 return (ENOMEM); 3296 ip = (struct ip *)(m->m_data + pi->ipi_ehdrlen); 3297 } 3298 } else { 3299 ip = (struct ip *)(m->m_data + pi->ipi_ehdrlen); 3300 } 3301 3302 /* Have the IPv4 header w/ no options here */ 3303 pi->ipi_ip_hlen = ip->ip_hl << 2; 3304 pi->ipi_ipproto = ip->ip_p; 3305 pi->ipi_ip_tos = ip->ip_tos; 3306 pi->ipi_flags |= IPI_TX_IPV4; 3307 3308 break; 3309 } 3310 #endif 3311 #ifdef INET6 3312 case ETHERTYPE_IPV6: 3313 { 3314 struct ip6_hdr *ip6; 3315 3316 if (__predict_false(m->m_len < pi->ipi_ehdrlen + sizeof(struct ip6_hdr))) { 3317 (*pullups)++; 3318 if (__predict_false((m = m_pullup(m, pi->ipi_ehdrlen + sizeof(struct ip6_hdr))) == NULL)) 3319 return (ENOMEM); 3320 } 3321 ip6 = (struct ip6_hdr *)(m->m_data + pi->ipi_ehdrlen); 3322 3323 /* Have the IPv6 fixed header here */ 3324 pi->ipi_ip_hlen = sizeof(struct ip6_hdr); 3325 pi->ipi_ipproto = ip6->ip6_nxt; 3326 pi->ipi_ip_tos = IPV6_TRAFFIC_CLASS(ip6); 3327 pi->ipi_flags |= IPI_TX_IPV6; 3328 3329 break; 3330 } 3331 #endif 3332 default: 3333 pi->ipi_csum_flags &= ~CSUM_OFFLOAD; 3334 pi->ipi_ip_hlen = 0; 3335 break; 3336 } 3337 *mp = m; 3338 3339 return (0); 3340 3341 } 3342 3343 static int 3344 iflib_parse_header(iflib_txq_t txq, if_pkt_info_t pi, struct mbuf **mp) 3345 { 3346 if_shared_ctx_t sctx = txq->ift_ctx->ifc_sctx; 3347 struct mbuf *m; 3348 int err; 3349 3350 m = *mp; 3351 if ((sctx->isc_flags & IFLIB_NEED_SCRATCH) && 3352 M_WRITABLE(m) == 0) { 3353 if ((m = m_dup(m, M_NOWAIT)) == NULL) { 3354 return (ENOMEM); 3355 } else { 3356 m_freem(*mp); 3357 DBG_COUNTER_INC(tx_frees); 3358 *mp = m; 3359 } 3360 } 3361 3362 /* Fills out pi->ipi_etype */ 3363 err = iflib_parse_ether_header(pi, mp, &txq->ift_pullups); 3364 if (__predict_false(err)) 3365 return (err); 3366 m = *mp; 3367 3368 switch (pi->ipi_etype) { 3369 #ifdef INET 3370 case ETHERTYPE_IP: 3371 { 3372 struct mbuf *n; 3373 struct ip *ip = NULL; 3374 struct tcphdr *th = NULL; 3375 int minthlen; 3376 3377 minthlen = min(m->m_pkthdr.len, pi->ipi_ehdrlen + sizeof(*ip) + sizeof(*th)); 3378 if (__predict_false(m->m_len < minthlen)) { 3379 /* 3380 * if this code bloat is causing too much of a hit 3381 * move it to a separate function and mark it noinline 3382 */ 3383 if (m->m_len == pi->ipi_ehdrlen) { 3384 n = m->m_next; 3385 MPASS(n); 3386 if (n->m_len >= sizeof(*ip)) { 3387 ip = (struct ip *)n->m_data; 3388 if (n->m_len >= (ip->ip_hl << 2) + sizeof(*th)) 3389 th = (struct tcphdr *)((caddr_t)ip + (ip->ip_hl << 2)); 3390 } else { 3391 txq->ift_pullups++; 3392 if (__predict_false((m = m_pullup(m, minthlen)) == NULL)) 3393 return (ENOMEM); 3394 ip = (struct ip *)(m->m_data + pi->ipi_ehdrlen); 3395 } 3396 } else { 3397 txq->ift_pullups++; 3398 if (__predict_false((m = m_pullup(m, minthlen)) == NULL)) 3399 return (ENOMEM); 3400 ip = (struct ip *)(m->m_data + pi->ipi_ehdrlen); 3401 if (m->m_len >= (ip->ip_hl << 2) + sizeof(*th)) 3402 th = (struct tcphdr *)((caddr_t)ip + (ip->ip_hl << 2)); 3403 } 3404 } else { 3405 ip = (struct ip *)(m->m_data + pi->ipi_ehdrlen); 3406 if (m->m_len >= (ip->ip_hl << 2) + sizeof(*th)) 3407 th = (struct tcphdr *)((caddr_t)ip + (ip->ip_hl << 2)); 3408 } 3409 pi->ipi_ip_hlen = ip->ip_hl << 2; 3410 pi->ipi_ipproto = ip->ip_p; 3411 pi->ipi_ip_tos = ip->ip_tos; 3412 pi->ipi_flags |= IPI_TX_IPV4; 3413 3414 /* TCP checksum offload may require TCP header length */ 3415 if (IS_TX_OFFLOAD4(pi)) { 3416 if (__predict_true(pi->ipi_ipproto == IPPROTO_TCP)) { 3417 if (__predict_false(th == NULL)) { 3418 txq->ift_pullups++; 3419 if (__predict_false((m = m_pullup(m, (ip->ip_hl << 2) + sizeof(*th))) == NULL)) 3420 return (ENOMEM); 3421 th = (struct tcphdr *)((caddr_t)ip + pi->ipi_ip_hlen); 3422 } 3423 pi->ipi_tcp_hflags = th->th_flags; 3424 pi->ipi_tcp_hlen = th->th_off << 2; 3425 pi->ipi_tcp_seq = th->th_seq; 3426 } 3427 if (IS_TSO4(pi)) { 3428 if (__predict_false(ip->ip_p != IPPROTO_TCP)) 3429 return (ENXIO); 3430 /* 3431 * TSO always requires hardware checksum offload. 3432 */ 3433 pi->ipi_csum_flags |= (CSUM_IP_TCP | CSUM_IP); 3434 th->th_sum = in_pseudo(ip->ip_src.s_addr, 3435 ip->ip_dst.s_addr, htons(IPPROTO_TCP)); 3436 pi->ipi_tso_segsz = m->m_pkthdr.tso_segsz; 3437 if (sctx->isc_flags & IFLIB_TSO_INIT_IP) { 3438 ip->ip_sum = 0; 3439 ip->ip_len = htons(pi->ipi_ip_hlen + pi->ipi_tcp_hlen + pi->ipi_tso_segsz); 3440 } 3441 } 3442 } 3443 if ((sctx->isc_flags & IFLIB_NEED_ZERO_CSUM) && (pi->ipi_csum_flags & CSUM_IP)) 3444 ip->ip_sum = 0; 3445 3446 break; 3447 } 3448 #endif 3449 #ifdef INET6 3450 case ETHERTYPE_IPV6: 3451 { 3452 struct ip6_hdr *ip6 = (struct ip6_hdr *)(m->m_data + pi->ipi_ehdrlen); 3453 struct tcphdr *th; 3454 pi->ipi_ip_hlen = sizeof(struct ip6_hdr); 3455 3456 if (__predict_false(m->m_len < pi->ipi_ehdrlen + sizeof(struct ip6_hdr))) { 3457 txq->ift_pullups++; 3458 if (__predict_false((m = m_pullup(m, pi->ipi_ehdrlen + sizeof(struct ip6_hdr))) == NULL)) 3459 return (ENOMEM); 3460 } 3461 th = (struct tcphdr *)((caddr_t)ip6 + pi->ipi_ip_hlen); 3462 3463 /* XXX-BZ this will go badly in case of ext hdrs. */ 3464 pi->ipi_ipproto = ip6->ip6_nxt; 3465 pi->ipi_ip_tos = IPV6_TRAFFIC_CLASS(ip6); 3466 pi->ipi_flags |= IPI_TX_IPV6; 3467 3468 /* TCP checksum offload may require TCP header length */ 3469 if (IS_TX_OFFLOAD6(pi)) { 3470 if (pi->ipi_ipproto == IPPROTO_TCP) { 3471 if (__predict_false(m->m_len < pi->ipi_ehdrlen + sizeof(struct ip6_hdr) + sizeof(struct tcphdr))) { 3472 txq->ift_pullups++; 3473 if (__predict_false((m = m_pullup(m, pi->ipi_ehdrlen + sizeof(struct ip6_hdr) + sizeof(struct tcphdr))) == NULL)) 3474 return (ENOMEM); 3475 } 3476 pi->ipi_tcp_hflags = th->th_flags; 3477 pi->ipi_tcp_hlen = th->th_off << 2; 3478 pi->ipi_tcp_seq = th->th_seq; 3479 } 3480 if (IS_TSO6(pi)) { 3481 if (__predict_false(ip6->ip6_nxt != IPPROTO_TCP)) 3482 return (ENXIO); 3483 /* 3484 * TSO always requires hardware checksum offload. 3485 */ 3486 pi->ipi_csum_flags |= CSUM_IP6_TCP; 3487 th->th_sum = in6_cksum_pseudo(ip6, 0, IPPROTO_TCP, 0); 3488 pi->ipi_tso_segsz = m->m_pkthdr.tso_segsz; 3489 } 3490 } 3491 break; 3492 } 3493 #endif 3494 default: 3495 pi->ipi_csum_flags &= ~CSUM_OFFLOAD; 3496 pi->ipi_ip_hlen = 0; 3497 break; 3498 } 3499 *mp = m; 3500 3501 return (0); 3502 } 3503 3504 /* 3505 * If dodgy hardware rejects the scatter gather chain we've handed it 3506 * we'll need to remove the mbuf chain from ifsg_m[] before we can add the 3507 * m_defrag'd mbufs 3508 */ 3509 static __noinline struct mbuf * 3510 iflib_remove_mbuf(iflib_txq_t txq) 3511 { 3512 int ntxd, pidx; 3513 struct mbuf *m, **ifsd_m; 3514 3515 ifsd_m = txq->ift_sds.ifsd_m; 3516 ntxd = txq->ift_size; 3517 pidx = txq->ift_pidx & (ntxd - 1); 3518 ifsd_m = txq->ift_sds.ifsd_m; 3519 m = ifsd_m[pidx]; 3520 ifsd_m[pidx] = NULL; 3521 bus_dmamap_unload(txq->ift_buf_tag, txq->ift_sds.ifsd_map[pidx]); 3522 if (txq->ift_sds.ifsd_tso_map != NULL) 3523 bus_dmamap_unload(txq->ift_tso_buf_tag, 3524 txq->ift_sds.ifsd_tso_map[pidx]); 3525 #if MEMORY_LOGGING 3526 txq->ift_dequeued++; 3527 #endif 3528 return (m); 3529 } 3530 3531 static inline caddr_t 3532 calc_next_txd(iflib_txq_t txq, int cidx, uint8_t qid) 3533 { 3534 qidx_t size; 3535 int ntxd; 3536 caddr_t start, end, cur, next; 3537 3538 ntxd = txq->ift_size; 3539 size = txq->ift_txd_size[qid]; 3540 start = txq->ift_ifdi[qid].idi_vaddr; 3541 3542 if (__predict_false(size == 0)) 3543 return (start); 3544 cur = start + size * cidx; 3545 end = start + size * ntxd; 3546 next = CACHE_PTR_NEXT(cur); 3547 return (next < end ? next : start); 3548 } 3549 3550 /* 3551 * Pad an mbuf to ensure a minimum ethernet frame size. 3552 * min_frame_size is the frame size (less CRC) to pad the mbuf to 3553 */ 3554 static __noinline int 3555 iflib_ether_pad(device_t dev, struct mbuf **m_head, uint16_t min_frame_size) 3556 { 3557 /* 3558 * 18 is enough bytes to pad an ARP packet to 46 bytes, and 3559 * and ARP message is the smallest common payload I can think of 3560 */ 3561 static char pad[18]; /* just zeros */ 3562 int n; 3563 struct mbuf *new_head; 3564 3565 if (!M_WRITABLE(*m_head)) { 3566 new_head = m_dup(*m_head, M_NOWAIT); 3567 if (new_head == NULL) { 3568 m_freem(*m_head); 3569 device_printf(dev, "cannot pad short frame, m_dup() failed"); 3570 DBG_COUNTER_INC(encap_pad_mbuf_fail); 3571 DBG_COUNTER_INC(tx_frees); 3572 return (ENOMEM); 3573 } 3574 m_freem(*m_head); 3575 *m_head = new_head; 3576 } 3577 3578 for (n = min_frame_size - (*m_head)->m_pkthdr.len; 3579 n > 0; n -= sizeof(pad)) 3580 if (!m_append(*m_head, min(n, sizeof(pad)), pad)) 3581 break; 3582 3583 if (n > 0) { 3584 m_freem(*m_head); 3585 device_printf(dev, "cannot pad short frame\n"); 3586 DBG_COUNTER_INC(encap_pad_mbuf_fail); 3587 DBG_COUNTER_INC(tx_frees); 3588 return (ENOBUFS); 3589 } 3590 3591 return (0); 3592 } 3593 3594 static int 3595 iflib_encap(iflib_txq_t txq, struct mbuf **m_headp) 3596 { 3597 if_ctx_t ctx; 3598 if_shared_ctx_t sctx; 3599 if_softc_ctx_t scctx; 3600 bus_dma_tag_t buf_tag; 3601 bus_dma_segment_t *segs; 3602 struct mbuf *m_head, **ifsd_m; 3603 void *next_txd; 3604 bus_dmamap_t map; 3605 struct if_pkt_info pi; 3606 int remap = 0; 3607 int err, nsegs, ndesc, max_segs, pidx, cidx, next, ntxd; 3608 3609 ctx = txq->ift_ctx; 3610 sctx = ctx->ifc_sctx; 3611 scctx = &ctx->ifc_softc_ctx; 3612 segs = txq->ift_segs; 3613 ntxd = txq->ift_size; 3614 m_head = *m_headp; 3615 map = NULL; 3616 3617 /* 3618 * If we're doing TSO the next descriptor to clean may be quite far ahead 3619 */ 3620 cidx = txq->ift_cidx; 3621 pidx = txq->ift_pidx; 3622 if (ctx->ifc_flags & IFC_PREFETCH) { 3623 next = (cidx + CACHE_PTR_INCREMENT) & (ntxd - 1); 3624 if (!(ctx->ifc_flags & IFLIB_HAS_TXCQ)) { 3625 next_txd = calc_next_txd(txq, cidx, 0); 3626 prefetch(next_txd); 3627 } 3628 3629 /* prefetch the next cache line of mbuf pointers and flags */ 3630 prefetch(&txq->ift_sds.ifsd_m[next]); 3631 prefetch(&txq->ift_sds.ifsd_map[next]); 3632 next = (cidx + CACHE_LINE_SIZE) & (ntxd - 1); 3633 } 3634 map = txq->ift_sds.ifsd_map[pidx]; 3635 ifsd_m = txq->ift_sds.ifsd_m; 3636 3637 if (m_head->m_pkthdr.csum_flags & CSUM_TSO) { 3638 buf_tag = txq->ift_tso_buf_tag; 3639 max_segs = scctx->isc_tx_tso_segments_max; 3640 map = txq->ift_sds.ifsd_tso_map[pidx]; 3641 MPASS(buf_tag != NULL); 3642 MPASS(max_segs > 0); 3643 } else { 3644 buf_tag = txq->ift_buf_tag; 3645 max_segs = scctx->isc_tx_nsegments; 3646 map = txq->ift_sds.ifsd_map[pidx]; 3647 } 3648 if ((sctx->isc_flags & IFLIB_NEED_ETHER_PAD) && 3649 __predict_false(m_head->m_pkthdr.len < scctx->isc_min_frame_size)) { 3650 err = iflib_ether_pad(ctx->ifc_dev, m_headp, scctx->isc_min_frame_size); 3651 if (err) { 3652 DBG_COUNTER_INC(encap_txd_encap_fail); 3653 return (err); 3654 } 3655 } 3656 m_head = *m_headp; 3657 3658 pkt_info_zero(&pi); 3659 pi.ipi_mflags = (m_head->m_flags & (M_VLANTAG | M_BCAST | M_MCAST)); 3660 pi.ipi_pidx = pidx; 3661 pi.ipi_qsidx = txq->ift_id; 3662 pi.ipi_len = m_head->m_pkthdr.len; 3663 pi.ipi_csum_flags = m_head->m_pkthdr.csum_flags; 3664 pi.ipi_vtag = M_HAS_VLANTAG(m_head) ? m_head->m_pkthdr.ether_vtag : 0; 3665 3666 /* deliberate bitwise OR to make one condition */ 3667 if (__predict_true((pi.ipi_csum_flags | pi.ipi_vtag))) { 3668 if (__predict_false((err = iflib_parse_header(txq, &pi, m_headp)) != 0)) { 3669 DBG_COUNTER_INC(encap_txd_encap_fail); 3670 return (err); 3671 } 3672 m_head = *m_headp; 3673 } 3674 3675 retry: 3676 err = bus_dmamap_load_mbuf_sg(buf_tag, map, m_head, segs, &nsegs, 3677 BUS_DMA_NOWAIT); 3678 defrag: 3679 if (__predict_false(err)) { 3680 switch (err) { 3681 case EFBIG: 3682 /* try collapse once and defrag once */ 3683 if (remap == 0) { 3684 m_head = m_collapse(*m_headp, M_NOWAIT, max_segs); 3685 /* try defrag if collapsing fails */ 3686 if (m_head == NULL) 3687 remap++; 3688 } 3689 if (remap == 1) { 3690 txq->ift_mbuf_defrag++; 3691 m_head = m_defrag(*m_headp, M_NOWAIT); 3692 } 3693 /* 3694 * remap should never be >1 unless bus_dmamap_load_mbuf_sg 3695 * failed to map an mbuf that was run through m_defrag 3696 */ 3697 MPASS(remap <= 1); 3698 if (__predict_false(m_head == NULL || remap > 1)) 3699 goto defrag_failed; 3700 remap++; 3701 *m_headp = m_head; 3702 goto retry; 3703 break; 3704 case ENOMEM: 3705 txq->ift_no_tx_dma_setup++; 3706 break; 3707 default: 3708 txq->ift_no_tx_dma_setup++; 3709 m_freem(*m_headp); 3710 DBG_COUNTER_INC(tx_frees); 3711 *m_headp = NULL; 3712 break; 3713 } 3714 txq->ift_map_failed++; 3715 DBG_COUNTER_INC(encap_load_mbuf_fail); 3716 DBG_COUNTER_INC(encap_txd_encap_fail); 3717 return (err); 3718 } 3719 ifsd_m[pidx] = m_head; 3720 /* 3721 * XXX assumes a 1 to 1 relationship between segments and 3722 * descriptors - this does not hold true on all drivers, e.g. 3723 * cxgb 3724 */ 3725 if (__predict_false(nsegs + 2 > TXQ_AVAIL(txq))) { 3726 txq->ift_no_desc_avail++; 3727 bus_dmamap_unload(buf_tag, map); 3728 DBG_COUNTER_INC(encap_txq_avail_fail); 3729 DBG_COUNTER_INC(encap_txd_encap_fail); 3730 if ((txq->ift_task.gt_task.ta_flags & TASK_ENQUEUED) == 0) 3731 GROUPTASK_ENQUEUE(&txq->ift_task); 3732 return (ENOBUFS); 3733 } 3734 /* 3735 * On Intel cards we can greatly reduce the number of TX interrupts 3736 * we see by only setting report status on every Nth descriptor. 3737 * However, this also means that the driver will need to keep track 3738 * of the descriptors that RS was set on to check them for the DD bit. 3739 */ 3740 txq->ift_rs_pending += nsegs + 1; 3741 if (txq->ift_rs_pending > TXQ_MAX_RS_DEFERRED(txq) || 3742 iflib_no_tx_batch || (TXQ_AVAIL(txq) - nsegs) <= MAX_TX_DESC(ctx) + 2) { 3743 pi.ipi_flags |= IPI_TX_INTR; 3744 txq->ift_rs_pending = 0; 3745 } 3746 3747 pi.ipi_segs = segs; 3748 pi.ipi_nsegs = nsegs; 3749 3750 MPASS(pidx >= 0 && pidx < txq->ift_size); 3751 #ifdef PKT_DEBUG 3752 print_pkt(&pi); 3753 #endif 3754 if ((err = ctx->isc_txd_encap(ctx->ifc_softc, &pi)) == 0) { 3755 bus_dmamap_sync(buf_tag, map, BUS_DMASYNC_PREWRITE); 3756 DBG_COUNTER_INC(tx_encap); 3757 MPASS(pi.ipi_new_pidx < txq->ift_size); 3758 3759 ndesc = pi.ipi_new_pidx - pi.ipi_pidx; 3760 if (pi.ipi_new_pidx < pi.ipi_pidx) { 3761 ndesc += txq->ift_size; 3762 txq->ift_gen = 1; 3763 } 3764 /* 3765 * drivers can need as many as 3766 * two sentinels 3767 */ 3768 MPASS(ndesc <= pi.ipi_nsegs + 2); 3769 MPASS(pi.ipi_new_pidx != pidx); 3770 MPASS(ndesc > 0); 3771 txq->ift_in_use += ndesc; 3772 txq->ift_db_pending += ndesc; 3773 3774 /* 3775 * We update the last software descriptor again here because there may 3776 * be a sentinel and/or there may be more mbufs than segments 3777 */ 3778 txq->ift_pidx = pi.ipi_new_pidx; 3779 txq->ift_npending += pi.ipi_ndescs; 3780 } else { 3781 *m_headp = m_head = iflib_remove_mbuf(txq); 3782 if (err == EFBIG) { 3783 txq->ift_txd_encap_efbig++; 3784 if (remap < 2) { 3785 remap = 1; 3786 goto defrag; 3787 } 3788 } 3789 goto defrag_failed; 3790 } 3791 /* 3792 * err can't possibly be non-zero here, so we don't neet to test it 3793 * to see if we need to DBG_COUNTER_INC(encap_txd_encap_fail). 3794 */ 3795 return (err); 3796 3797 defrag_failed: 3798 txq->ift_mbuf_defrag_failed++; 3799 txq->ift_map_failed++; 3800 m_freem(*m_headp); 3801 DBG_COUNTER_INC(tx_frees); 3802 *m_headp = NULL; 3803 DBG_COUNTER_INC(encap_txd_encap_fail); 3804 return (ENOMEM); 3805 } 3806 3807 static void 3808 iflib_tx_desc_free(iflib_txq_t txq, int n) 3809 { 3810 uint32_t qsize, cidx, mask, gen; 3811 struct mbuf *m, **ifsd_m; 3812 bool do_prefetch; 3813 3814 cidx = txq->ift_cidx; 3815 gen = txq->ift_gen; 3816 qsize = txq->ift_size; 3817 mask = qsize - 1; 3818 ifsd_m = txq->ift_sds.ifsd_m; 3819 do_prefetch = (txq->ift_ctx->ifc_flags & IFC_PREFETCH); 3820 3821 while (n-- > 0) { 3822 if (do_prefetch) { 3823 prefetch(ifsd_m[(cidx + 3) & mask]); 3824 prefetch(ifsd_m[(cidx + 4) & mask]); 3825 } 3826 if ((m = ifsd_m[cidx]) != NULL) { 3827 prefetch(&ifsd_m[(cidx + CACHE_PTR_INCREMENT) & mask]); 3828 if (m->m_pkthdr.csum_flags & CSUM_TSO) { 3829 bus_dmamap_sync(txq->ift_tso_buf_tag, 3830 txq->ift_sds.ifsd_tso_map[cidx], 3831 BUS_DMASYNC_POSTWRITE); 3832 bus_dmamap_unload(txq->ift_tso_buf_tag, 3833 txq->ift_sds.ifsd_tso_map[cidx]); 3834 } else { 3835 bus_dmamap_sync(txq->ift_buf_tag, 3836 txq->ift_sds.ifsd_map[cidx], 3837 BUS_DMASYNC_POSTWRITE); 3838 bus_dmamap_unload(txq->ift_buf_tag, 3839 txq->ift_sds.ifsd_map[cidx]); 3840 } 3841 /* XXX we don't support any drivers that batch packets yet */ 3842 MPASS(m->m_nextpkt == NULL); 3843 m_freem(m); 3844 ifsd_m[cidx] = NULL; 3845 #if MEMORY_LOGGING 3846 txq->ift_dequeued++; 3847 #endif 3848 DBG_COUNTER_INC(tx_frees); 3849 } 3850 if (__predict_false(++cidx == qsize)) { 3851 cidx = 0; 3852 gen = 0; 3853 } 3854 } 3855 txq->ift_cidx = cidx; 3856 txq->ift_gen = gen; 3857 } 3858 3859 static __inline int 3860 iflib_completed_tx_reclaim(iflib_txq_t txq, int thresh) 3861 { 3862 int reclaim; 3863 if_ctx_t ctx = txq->ift_ctx; 3864 3865 KASSERT(thresh >= 0, ("invalid threshold to reclaim")); 3866 MPASS(thresh /*+ MAX_TX_DESC(txq->ift_ctx) */ < txq->ift_size); 3867 3868 /* 3869 * Need a rate-limiting check so that this isn't called every time 3870 */ 3871 iflib_tx_credits_update(ctx, txq); 3872 reclaim = DESC_RECLAIMABLE(txq); 3873 3874 if (reclaim <= thresh /* + MAX_TX_DESC(txq->ift_ctx) */) { 3875 #ifdef INVARIANTS 3876 if (iflib_verbose_debug) { 3877 printf("%s processed=%ju cleaned=%ju tx_nsegments=%d reclaim=%d thresh=%d\n", __func__, 3878 txq->ift_processed, txq->ift_cleaned, txq->ift_ctx->ifc_softc_ctx.isc_tx_nsegments, 3879 reclaim, thresh); 3880 } 3881 #endif 3882 return (0); 3883 } 3884 iflib_tx_desc_free(txq, reclaim); 3885 txq->ift_cleaned += reclaim; 3886 txq->ift_in_use -= reclaim; 3887 3888 return (reclaim); 3889 } 3890 3891 static struct mbuf ** 3892 _ring_peek_one(struct ifmp_ring *r, int cidx, int offset, int remaining) 3893 { 3894 int next, size; 3895 struct mbuf **items; 3896 3897 size = r->size; 3898 next = (cidx + CACHE_PTR_INCREMENT) & (size - 1); 3899 items = __DEVOLATILE(struct mbuf **, &r->items[0]); 3900 3901 prefetch(items[(cidx + offset) & (size - 1)]); 3902 if (remaining > 1) { 3903 prefetch2cachelines(&items[next]); 3904 prefetch2cachelines(items[(cidx + offset + 1) & (size - 1)]); 3905 prefetch2cachelines(items[(cidx + offset + 2) & (size - 1)]); 3906 prefetch2cachelines(items[(cidx + offset + 3) & (size - 1)]); 3907 } 3908 return (__DEVOLATILE(struct mbuf **, &r->items[(cidx + offset) & (size - 1)])); 3909 } 3910 3911 static void 3912 iflib_txq_check_drain(iflib_txq_t txq, int budget) 3913 { 3914 3915 ifmp_ring_check_drainage(txq->ift_br, budget); 3916 } 3917 3918 static uint32_t 3919 iflib_txq_can_drain(struct ifmp_ring *r) 3920 { 3921 iflib_txq_t txq = r->cookie; 3922 if_ctx_t ctx = txq->ift_ctx; 3923 3924 if (TXQ_AVAIL(txq) > MAX_TX_DESC(ctx) + 2) 3925 return (1); 3926 bus_dmamap_sync(txq->ift_ifdi->idi_tag, txq->ift_ifdi->idi_map, 3927 BUS_DMASYNC_POSTREAD); 3928 return (ctx->isc_txd_credits_update(ctx->ifc_softc, txq->ift_id, 3929 false)); 3930 } 3931 3932 static uint32_t 3933 iflib_txq_drain(struct ifmp_ring *r, uint32_t cidx, uint32_t pidx) 3934 { 3935 iflib_txq_t txq = r->cookie; 3936 if_ctx_t ctx = txq->ift_ctx; 3937 if_t ifp = ctx->ifc_ifp; 3938 struct mbuf *m, **mp; 3939 int avail, bytes_sent, skipped, count, err, i; 3940 int mcast_sent, pkt_sent, reclaimed; 3941 bool do_prefetch, rang, ring; 3942 3943 if (__predict_false(!(if_getdrvflags(ifp) & IFF_DRV_RUNNING) || 3944 !LINK_ACTIVE(ctx))) { 3945 DBG_COUNTER_INC(txq_drain_notready); 3946 return (0); 3947 } 3948 reclaimed = iflib_completed_tx_reclaim(txq, RECLAIM_THRESH(ctx)); 3949 rang = iflib_txd_db_check(txq, reclaimed && txq->ift_db_pending); 3950 avail = IDXDIFF(pidx, cidx, r->size); 3951 3952 if (__predict_false(ctx->ifc_flags & IFC_QFLUSH)) { 3953 /* 3954 * The driver is unloading so we need to free all pending packets. 3955 */ 3956 DBG_COUNTER_INC(txq_drain_flushing); 3957 for (i = 0; i < avail; i++) { 3958 if (__predict_true(r->items[(cidx + i) & (r->size - 1)] != (void *)txq)) 3959 m_freem(r->items[(cidx + i) & (r->size - 1)]); 3960 r->items[(cidx + i) & (r->size - 1)] = NULL; 3961 } 3962 return (avail); 3963 } 3964 3965 if (__predict_false(if_getdrvflags(ctx->ifc_ifp) & IFF_DRV_OACTIVE)) { 3966 txq->ift_qstatus = IFLIB_QUEUE_IDLE; 3967 CALLOUT_LOCK(txq); 3968 callout_stop(&txq->ift_timer); 3969 CALLOUT_UNLOCK(txq); 3970 DBG_COUNTER_INC(txq_drain_oactive); 3971 return (0); 3972 } 3973 3974 /* 3975 * If we've reclaimed any packets this queue cannot be hung. 3976 */ 3977 if (reclaimed) 3978 txq->ift_qstatus = IFLIB_QUEUE_IDLE; 3979 skipped = mcast_sent = bytes_sent = pkt_sent = 0; 3980 count = MIN(avail, TX_BATCH_SIZE); 3981 #ifdef INVARIANTS 3982 if (iflib_verbose_debug) 3983 printf("%s avail=%d ifc_flags=%x txq_avail=%d ", __func__, 3984 avail, ctx->ifc_flags, TXQ_AVAIL(txq)); 3985 #endif 3986 do_prefetch = (ctx->ifc_flags & IFC_PREFETCH); 3987 err = 0; 3988 for (i = 0; i < count && TXQ_AVAIL(txq) >= MAX_TX_DESC(ctx) + 2; i++) { 3989 int rem = do_prefetch ? count - i : 0; 3990 3991 mp = _ring_peek_one(r, cidx, i, rem); 3992 MPASS(mp != NULL && *mp != NULL); 3993 3994 /* 3995 * Completion interrupts will use the address of the txq 3996 * as a sentinel to enqueue _something_ in order to acquire 3997 * the lock on the mp_ring (there's no direct lock call). 3998 * We obviously whave to check for these sentinel cases 3999 * and skip them. 4000 */ 4001 if (__predict_false(*mp == (struct mbuf *)txq)) { 4002 skipped++; 4003 continue; 4004 } 4005 err = iflib_encap(txq, mp); 4006 if (__predict_false(err)) { 4007 /* no room - bail out */ 4008 if (err == ENOBUFS) 4009 break; 4010 skipped++; 4011 /* we can't send this packet - skip it */ 4012 continue; 4013 } 4014 pkt_sent++; 4015 m = *mp; 4016 DBG_COUNTER_INC(tx_sent); 4017 bytes_sent += m->m_pkthdr.len; 4018 mcast_sent += !!(m->m_flags & M_MCAST); 4019 4020 if (__predict_false(!(if_getdrvflags(ifp) & IFF_DRV_RUNNING))) 4021 break; 4022 ETHER_BPF_MTAP(ifp, m); 4023 rang = iflib_txd_db_check(txq, false); 4024 } 4025 4026 /* deliberate use of bitwise or to avoid gratuitous short-circuit */ 4027 ring = rang ? false : (iflib_min_tx_latency | err); 4028 iflib_txd_db_check(txq, ring); 4029 if_inc_counter(ifp, IFCOUNTER_OBYTES, bytes_sent); 4030 if_inc_counter(ifp, IFCOUNTER_OPACKETS, pkt_sent); 4031 if (mcast_sent) 4032 if_inc_counter(ifp, IFCOUNTER_OMCASTS, mcast_sent); 4033 #ifdef INVARIANTS 4034 if (iflib_verbose_debug) 4035 printf("consumed=%d\n", skipped + pkt_sent); 4036 #endif 4037 return (skipped + pkt_sent); 4038 } 4039 4040 static uint32_t 4041 iflib_txq_drain_always(struct ifmp_ring *r) 4042 { 4043 return (1); 4044 } 4045 4046 static uint32_t 4047 iflib_txq_drain_free(struct ifmp_ring *r, uint32_t cidx, uint32_t pidx) 4048 { 4049 int i, avail; 4050 struct mbuf **mp; 4051 iflib_txq_t txq; 4052 4053 txq = r->cookie; 4054 4055 txq->ift_qstatus = IFLIB_QUEUE_IDLE; 4056 CALLOUT_LOCK(txq); 4057 callout_stop(&txq->ift_timer); 4058 CALLOUT_UNLOCK(txq); 4059 4060 avail = IDXDIFF(pidx, cidx, r->size); 4061 for (i = 0; i < avail; i++) { 4062 mp = _ring_peek_one(r, cidx, i, avail - i); 4063 if (__predict_false(*mp == (struct mbuf *)txq)) 4064 continue; 4065 m_freem(*mp); 4066 DBG_COUNTER_INC(tx_frees); 4067 } 4068 MPASS(ifmp_ring_is_stalled(r) == 0); 4069 return (avail); 4070 } 4071 4072 static void 4073 iflib_ifmp_purge(iflib_txq_t txq) 4074 { 4075 struct ifmp_ring *r; 4076 4077 r = txq->ift_br; 4078 r->drain = iflib_txq_drain_free; 4079 r->can_drain = iflib_txq_drain_always; 4080 4081 ifmp_ring_check_drainage(r, r->size); 4082 4083 r->drain = iflib_txq_drain; 4084 r->can_drain = iflib_txq_can_drain; 4085 } 4086 4087 static void 4088 _task_fn_tx(void *context) 4089 { 4090 iflib_txq_t txq = context; 4091 if_ctx_t ctx = txq->ift_ctx; 4092 if_t ifp = ctx->ifc_ifp; 4093 int abdicate = ctx->ifc_sysctl_tx_abdicate; 4094 4095 #ifdef IFLIB_DIAGNOSTICS 4096 txq->ift_cpu_exec_count[curcpu]++; 4097 #endif 4098 if (!(if_getdrvflags(ifp) & IFF_DRV_RUNNING)) 4099 return; 4100 #ifdef DEV_NETMAP 4101 if ((if_getcapenable(ifp) & IFCAP_NETMAP) && 4102 netmap_tx_irq(ifp, txq->ift_id)) 4103 goto skip_ifmp; 4104 #endif 4105 #ifdef ALTQ 4106 if (if_altq_is_enabled(ifp)) 4107 iflib_altq_if_start(ifp); 4108 #endif 4109 if (txq->ift_db_pending) 4110 ifmp_ring_enqueue(txq->ift_br, (void **)&txq, 1, TX_BATCH_SIZE, abdicate); 4111 else if (!abdicate) 4112 ifmp_ring_check_drainage(txq->ift_br, TX_BATCH_SIZE); 4113 /* 4114 * When abdicating, we always need to check drainage, not just when we don't enqueue 4115 */ 4116 if (abdicate) 4117 ifmp_ring_check_drainage(txq->ift_br, TX_BATCH_SIZE); 4118 #ifdef DEV_NETMAP 4119 skip_ifmp: 4120 #endif 4121 if (ctx->ifc_flags & IFC_LEGACY) 4122 IFDI_INTR_ENABLE(ctx); 4123 else 4124 IFDI_TX_QUEUE_INTR_ENABLE(ctx, txq->ift_id); 4125 } 4126 4127 static void 4128 _task_fn_rx(void *context) 4129 { 4130 iflib_rxq_t rxq = context; 4131 if_ctx_t ctx = rxq->ifr_ctx; 4132 uint8_t more; 4133 uint16_t budget; 4134 #ifdef DEV_NETMAP 4135 u_int work = 0; 4136 int nmirq; 4137 #endif 4138 4139 #ifdef IFLIB_DIAGNOSTICS 4140 rxq->ifr_cpu_exec_count[curcpu]++; 4141 #endif 4142 DBG_COUNTER_INC(task_fn_rxs); 4143 if (__predict_false(!(if_getdrvflags(ctx->ifc_ifp) & IFF_DRV_RUNNING))) 4144 return; 4145 #ifdef DEV_NETMAP 4146 nmirq = netmap_rx_irq(ctx->ifc_ifp, rxq->ifr_id, &work); 4147 if (nmirq != NM_IRQ_PASS) { 4148 more = (nmirq == NM_IRQ_RESCHED) ? IFLIB_RXEOF_MORE : 0; 4149 goto skip_rxeof; 4150 } 4151 #endif 4152 budget = ctx->ifc_sysctl_rx_budget; 4153 if (budget == 0) 4154 budget = 16; /* XXX */ 4155 more = iflib_rxeof(rxq, budget); 4156 #ifdef DEV_NETMAP 4157 skip_rxeof: 4158 #endif 4159 if ((more & IFLIB_RXEOF_MORE) == 0) { 4160 if (ctx->ifc_flags & IFC_LEGACY) 4161 IFDI_INTR_ENABLE(ctx); 4162 else 4163 IFDI_RX_QUEUE_INTR_ENABLE(ctx, rxq->ifr_id); 4164 DBG_COUNTER_INC(rx_intr_enables); 4165 } 4166 if (__predict_false(!(if_getdrvflags(ctx->ifc_ifp) & IFF_DRV_RUNNING))) 4167 return; 4168 4169 if (more & IFLIB_RXEOF_MORE) 4170 GROUPTASK_ENQUEUE(&rxq->ifr_task); 4171 else if (more & IFLIB_RXEOF_EMPTY) 4172 callout_reset_curcpu(&rxq->ifr_watchdog, 1, &_task_fn_rx_watchdog, rxq); 4173 } 4174 4175 static void 4176 _task_fn_admin(void *context) 4177 { 4178 if_ctx_t ctx = context; 4179 if_softc_ctx_t sctx = &ctx->ifc_softc_ctx; 4180 iflib_txq_t txq; 4181 int i; 4182 bool oactive, running, do_reset, do_watchdog, in_detach; 4183 4184 STATE_LOCK(ctx); 4185 running = (if_getdrvflags(ctx->ifc_ifp) & IFF_DRV_RUNNING); 4186 oactive = (if_getdrvflags(ctx->ifc_ifp) & IFF_DRV_OACTIVE); 4187 do_reset = (ctx->ifc_flags & IFC_DO_RESET); 4188 do_watchdog = (ctx->ifc_flags & IFC_DO_WATCHDOG); 4189 in_detach = (ctx->ifc_flags & IFC_IN_DETACH); 4190 ctx->ifc_flags &= ~(IFC_DO_RESET | IFC_DO_WATCHDOG); 4191 STATE_UNLOCK(ctx); 4192 4193 if ((!running && !oactive) && !(ctx->ifc_sctx->isc_flags & IFLIB_ADMIN_ALWAYS_RUN)) 4194 return; 4195 if (in_detach) 4196 return; 4197 4198 CTX_LOCK(ctx); 4199 for (txq = ctx->ifc_txqs, i = 0; i < sctx->isc_ntxqsets; i++, txq++) { 4200 CALLOUT_LOCK(txq); 4201 callout_stop(&txq->ift_timer); 4202 CALLOUT_UNLOCK(txq); 4203 } 4204 if (ctx->ifc_sctx->isc_flags & IFLIB_HAS_ADMINCQ) 4205 IFDI_ADMIN_COMPLETION_HANDLE(ctx); 4206 if (do_watchdog) { 4207 ctx->ifc_watchdog_events++; 4208 IFDI_WATCHDOG_RESET(ctx); 4209 } 4210 IFDI_UPDATE_ADMIN_STATUS(ctx); 4211 for (txq = ctx->ifc_txqs, i = 0; i < sctx->isc_ntxqsets; i++, txq++) { 4212 callout_reset_on(&txq->ift_timer, iflib_timer_default, iflib_timer, txq, 4213 txq->ift_timer.c_cpu); 4214 } 4215 IFDI_LINK_INTR_ENABLE(ctx); 4216 if (do_reset) 4217 iflib_if_init_locked(ctx); 4218 CTX_UNLOCK(ctx); 4219 4220 if (LINK_ACTIVE(ctx) == 0) 4221 return; 4222 for (txq = ctx->ifc_txqs, i = 0; i < sctx->isc_ntxqsets; i++, txq++) 4223 iflib_txq_check_drain(txq, IFLIB_RESTART_BUDGET); 4224 } 4225 4226 static void 4227 _task_fn_iov(void *context) 4228 { 4229 if_ctx_t ctx = context; 4230 4231 if (!(if_getdrvflags(ctx->ifc_ifp) & IFF_DRV_RUNNING) && 4232 !(ctx->ifc_sctx->isc_flags & IFLIB_ADMIN_ALWAYS_RUN)) 4233 return; 4234 4235 CTX_LOCK(ctx); 4236 IFDI_VFLR_HANDLE(ctx); 4237 CTX_UNLOCK(ctx); 4238 } 4239 4240 static int 4241 iflib_sysctl_int_delay(SYSCTL_HANDLER_ARGS) 4242 { 4243 int err; 4244 if_int_delay_info_t info; 4245 if_ctx_t ctx; 4246 4247 info = (if_int_delay_info_t)arg1; 4248 ctx = info->iidi_ctx; 4249 info->iidi_req = req; 4250 info->iidi_oidp = oidp; 4251 CTX_LOCK(ctx); 4252 err = IFDI_SYSCTL_INT_DELAY(ctx, info); 4253 CTX_UNLOCK(ctx); 4254 return (err); 4255 } 4256 4257 /********************************************************************* 4258 * 4259 * IFNET FUNCTIONS 4260 * 4261 **********************************************************************/ 4262 4263 static void 4264 iflib_if_init_locked(if_ctx_t ctx) 4265 { 4266 iflib_stop(ctx); 4267 iflib_init_locked(ctx); 4268 } 4269 4270 static void 4271 iflib_if_init(void *arg) 4272 { 4273 if_ctx_t ctx = arg; 4274 4275 CTX_LOCK(ctx); 4276 iflib_if_init_locked(ctx); 4277 CTX_UNLOCK(ctx); 4278 } 4279 4280 static int 4281 iflib_if_transmit(if_t ifp, struct mbuf *m) 4282 { 4283 if_ctx_t ctx = if_getsoftc(ifp); 4284 iflib_txq_t txq; 4285 int err, qidx; 4286 int abdicate; 4287 4288 if (__predict_false((if_getdrvflags(ifp) & IFF_DRV_RUNNING) == 0 || !LINK_ACTIVE(ctx))) { 4289 DBG_COUNTER_INC(tx_frees); 4290 m_freem(m); 4291 return (ENETDOWN); 4292 } 4293 4294 MPASS(m->m_nextpkt == NULL); 4295 /* ALTQ-enabled interfaces always use queue 0. */ 4296 qidx = 0; 4297 /* Use driver-supplied queue selection method if it exists */ 4298 if (ctx->isc_txq_select_v2) { 4299 struct if_pkt_info pi; 4300 uint64_t early_pullups = 0; 4301 pkt_info_zero(&pi); 4302 4303 err = iflib_parse_header_partial(&pi, &m, &early_pullups); 4304 if (__predict_false(err != 0)) { 4305 /* Assign pullups for bad pkts to default queue */ 4306 ctx->ifc_txqs[0].ift_pullups += early_pullups; 4307 DBG_COUNTER_INC(encap_txd_encap_fail); 4308 return (err); 4309 } 4310 /* Let driver make queueing decision */ 4311 qidx = ctx->isc_txq_select_v2(ctx->ifc_softc, m, &pi); 4312 ctx->ifc_txqs[qidx].ift_pullups += early_pullups; 4313 } 4314 /* Backwards compatibility w/ simpler queue select */ 4315 else if (ctx->isc_txq_select) 4316 qidx = ctx->isc_txq_select(ctx->ifc_softc, m); 4317 /* If not, use iflib's standard method */ 4318 else if ((NTXQSETS(ctx) > 1) && M_HASHTYPE_GET(m) && !if_altq_is_enabled(ifp)) 4319 qidx = QIDX(ctx, m); 4320 4321 /* Set TX queue */ 4322 txq = &ctx->ifc_txqs[qidx]; 4323 4324 #ifdef DRIVER_BACKPRESSURE 4325 if (txq->ift_closed) { 4326 while (m != NULL) { 4327 next = m->m_nextpkt; 4328 m->m_nextpkt = NULL; 4329 m_freem(m); 4330 DBG_COUNTER_INC(tx_frees); 4331 m = next; 4332 } 4333 return (ENOBUFS); 4334 } 4335 #endif 4336 #ifdef notyet 4337 qidx = count = 0; 4338 mp = marr; 4339 next = m; 4340 do { 4341 count++; 4342 next = next->m_nextpkt; 4343 } while (next != NULL); 4344 4345 if (count > nitems(marr)) 4346 if ((mp = malloc(count * sizeof(struct mbuf *), M_IFLIB, M_NOWAIT)) == NULL) { 4347 /* XXX check nextpkt */ 4348 m_freem(m); 4349 /* XXX simplify for now */ 4350 DBG_COUNTER_INC(tx_frees); 4351 return (ENOBUFS); 4352 } 4353 for (next = m, i = 0; next != NULL; i++) { 4354 mp[i] = next; 4355 next = next->m_nextpkt; 4356 mp[i]->m_nextpkt = NULL; 4357 } 4358 #endif 4359 DBG_COUNTER_INC(tx_seen); 4360 abdicate = ctx->ifc_sysctl_tx_abdicate; 4361 4362 err = ifmp_ring_enqueue(txq->ift_br, (void **)&m, 1, TX_BATCH_SIZE, abdicate); 4363 4364 if (abdicate) 4365 GROUPTASK_ENQUEUE(&txq->ift_task); 4366 if (err) { 4367 if (!abdicate) 4368 GROUPTASK_ENQUEUE(&txq->ift_task); 4369 /* support forthcoming later */ 4370 #ifdef DRIVER_BACKPRESSURE 4371 txq->ift_closed = TRUE; 4372 #endif 4373 ifmp_ring_check_drainage(txq->ift_br, TX_BATCH_SIZE); 4374 m_freem(m); 4375 DBG_COUNTER_INC(tx_frees); 4376 } 4377 4378 return (err); 4379 } 4380 4381 #ifdef ALTQ 4382 /* 4383 * The overall approach to integrating iflib with ALTQ is to continue to use 4384 * the iflib mp_ring machinery between the ALTQ queue(s) and the hardware 4385 * ring. Technically, when using ALTQ, queueing to an intermediate mp_ring 4386 * is redundant/unnecessary, but doing so minimizes the amount of 4387 * ALTQ-specific code required in iflib. It is assumed that the overhead of 4388 * redundantly queueing to an intermediate mp_ring is swamped by the 4389 * performance limitations inherent in using ALTQ. 4390 * 4391 * When ALTQ support is compiled in, all iflib drivers will use a transmit 4392 * routine, iflib_altq_if_transmit(), that checks if ALTQ is enabled for the 4393 * given interface. If ALTQ is enabled for an interface, then all 4394 * transmitted packets for that interface will be submitted to the ALTQ 4395 * subsystem via IFQ_ENQUEUE(). We don't use the legacy if_transmit() 4396 * implementation because it uses IFQ_HANDOFF(), which will duplicatively 4397 * update stats that the iflib machinery handles, and which is sensitve to 4398 * the disused IFF_DRV_OACTIVE flag. Additionally, iflib_altq_if_start() 4399 * will be installed as the start routine for use by ALTQ facilities that 4400 * need to trigger queue drains on a scheduled basis. 4401 * 4402 */ 4403 static void 4404 iflib_altq_if_start(if_t ifp) 4405 { 4406 struct ifaltq *ifq = &ifp->if_snd; /* XXX - DRVAPI */ 4407 struct mbuf *m; 4408 4409 IFQ_LOCK(ifq); 4410 IFQ_DEQUEUE_NOLOCK(ifq, m); 4411 while (m != NULL) { 4412 iflib_if_transmit(ifp, m); 4413 IFQ_DEQUEUE_NOLOCK(ifq, m); 4414 } 4415 IFQ_UNLOCK(ifq); 4416 } 4417 4418 static int 4419 iflib_altq_if_transmit(if_t ifp, struct mbuf *m) 4420 { 4421 int err; 4422 4423 if (if_altq_is_enabled(ifp)) { 4424 IFQ_ENQUEUE(&ifp->if_snd, m, err); /* XXX - DRVAPI */ 4425 if (err == 0) 4426 iflib_altq_if_start(ifp); 4427 } else 4428 err = iflib_if_transmit(ifp, m); 4429 4430 return (err); 4431 } 4432 #endif /* ALTQ */ 4433 4434 static void 4435 iflib_if_qflush(if_t ifp) 4436 { 4437 if_ctx_t ctx = if_getsoftc(ifp); 4438 iflib_txq_t txq = ctx->ifc_txqs; 4439 int i; 4440 4441 STATE_LOCK(ctx); 4442 ctx->ifc_flags |= IFC_QFLUSH; 4443 STATE_UNLOCK(ctx); 4444 for (i = 0; i < NTXQSETS(ctx); i++, txq++) 4445 while (!(ifmp_ring_is_idle(txq->ift_br) || ifmp_ring_is_stalled(txq->ift_br))) 4446 iflib_txq_check_drain(txq, 0); 4447 STATE_LOCK(ctx); 4448 ctx->ifc_flags &= ~IFC_QFLUSH; 4449 STATE_UNLOCK(ctx); 4450 4451 /* 4452 * When ALTQ is enabled, this will also take care of purging the 4453 * ALTQ queue(s). 4454 */ 4455 if_qflush(ifp); 4456 } 4457 4458 #define IFCAP_FLAGS (IFCAP_HWCSUM_IPV6 | IFCAP_HWCSUM | IFCAP_LRO | \ 4459 IFCAP_TSO | IFCAP_VLAN_HWTAGGING | IFCAP_HWSTATS | \ 4460 IFCAP_VLAN_MTU | IFCAP_VLAN_HWFILTER | \ 4461 IFCAP_VLAN_HWTSO | IFCAP_VLAN_HWCSUM | IFCAP_MEXTPG) 4462 4463 static int 4464 iflib_if_ioctl(if_t ifp, u_long command, caddr_t data) 4465 { 4466 if_ctx_t ctx = if_getsoftc(ifp); 4467 struct ifreq *ifr = (struct ifreq *)data; 4468 #if defined(INET) || defined(INET6) 4469 struct ifaddr *ifa = (struct ifaddr *)data; 4470 #endif 4471 bool avoid_reset = false; 4472 int err = 0, reinit = 0, bits; 4473 4474 switch (command) { 4475 case SIOCSIFADDR: 4476 #ifdef INET 4477 if (ifa->ifa_addr->sa_family == AF_INET) 4478 avoid_reset = true; 4479 #endif 4480 #ifdef INET6 4481 if (ifa->ifa_addr->sa_family == AF_INET6) 4482 avoid_reset = true; 4483 #endif 4484 /* 4485 * Calling init results in link renegotiation, 4486 * so we avoid doing it when possible. 4487 */ 4488 if (avoid_reset) { 4489 if_setflagbits(ifp, IFF_UP, 0); 4490 if (!(if_getdrvflags(ifp) & IFF_DRV_RUNNING)) 4491 reinit = 1; 4492 #ifdef INET 4493 if (!(if_getflags(ifp) & IFF_NOARP)) 4494 arp_ifinit(ifp, ifa); 4495 #endif 4496 } else 4497 err = ether_ioctl(ifp, command, data); 4498 break; 4499 case SIOCSIFMTU: 4500 CTX_LOCK(ctx); 4501 if (ifr->ifr_mtu == if_getmtu(ifp)) { 4502 CTX_UNLOCK(ctx); 4503 break; 4504 } 4505 bits = if_getdrvflags(ifp); 4506 /* stop the driver and free any clusters before proceeding */ 4507 iflib_stop(ctx); 4508 4509 if ((err = IFDI_MTU_SET(ctx, ifr->ifr_mtu)) == 0) { 4510 STATE_LOCK(ctx); 4511 if (ifr->ifr_mtu > ctx->ifc_max_fl_buf_size) 4512 ctx->ifc_flags |= IFC_MULTISEG; 4513 else 4514 ctx->ifc_flags &= ~IFC_MULTISEG; 4515 STATE_UNLOCK(ctx); 4516 err = if_setmtu(ifp, ifr->ifr_mtu); 4517 } 4518 iflib_init_locked(ctx); 4519 STATE_LOCK(ctx); 4520 if_setdrvflags(ifp, bits); 4521 STATE_UNLOCK(ctx); 4522 CTX_UNLOCK(ctx); 4523 break; 4524 case SIOCSIFFLAGS: 4525 CTX_LOCK(ctx); 4526 if (if_getflags(ifp) & IFF_UP) { 4527 if (if_getdrvflags(ifp) & IFF_DRV_RUNNING) { 4528 if ((if_getflags(ifp) ^ ctx->ifc_if_flags) & 4529 (IFF_PROMISC | IFF_ALLMULTI)) { 4530 CTX_UNLOCK(ctx); 4531 err = IFDI_PROMISC_SET(ctx, if_getflags(ifp)); 4532 CTX_LOCK(ctx); 4533 } 4534 } else 4535 reinit = 1; 4536 } else if (if_getdrvflags(ifp) & IFF_DRV_RUNNING) { 4537 iflib_stop(ctx); 4538 } 4539 ctx->ifc_if_flags = if_getflags(ifp); 4540 CTX_UNLOCK(ctx); 4541 break; 4542 case SIOCADDMULTI: 4543 case SIOCDELMULTI: 4544 if (if_getdrvflags(ifp) & IFF_DRV_RUNNING) { 4545 CTX_LOCK(ctx); 4546 IFDI_INTR_DISABLE(ctx); 4547 IFDI_MULTI_SET(ctx); 4548 IFDI_INTR_ENABLE(ctx); 4549 CTX_UNLOCK(ctx); 4550 } 4551 break; 4552 case SIOCSIFMEDIA: 4553 CTX_LOCK(ctx); 4554 IFDI_MEDIA_SET(ctx); 4555 CTX_UNLOCK(ctx); 4556 /* FALLTHROUGH */ 4557 case SIOCGIFMEDIA: 4558 case SIOCGIFXMEDIA: 4559 err = ifmedia_ioctl(ifp, ifr, ctx->ifc_mediap, command); 4560 break; 4561 case SIOCGI2C: 4562 { 4563 struct ifi2creq i2c; 4564 4565 err = copyin(ifr_data_get_ptr(ifr), &i2c, sizeof(i2c)); 4566 if (err != 0) 4567 break; 4568 if (i2c.dev_addr != 0xA0 && i2c.dev_addr != 0xA2) { 4569 err = EINVAL; 4570 break; 4571 } 4572 if (i2c.len > sizeof(i2c.data)) { 4573 err = EINVAL; 4574 break; 4575 } 4576 4577 if ((err = IFDI_I2C_REQ(ctx, &i2c)) == 0) 4578 err = copyout(&i2c, ifr_data_get_ptr(ifr), 4579 sizeof(i2c)); 4580 break; 4581 } 4582 case SIOCSIFCAP: 4583 { 4584 int mask, setmask, oldmask; 4585 4586 oldmask = if_getcapenable(ifp); 4587 mask = ifr->ifr_reqcap ^ oldmask; 4588 mask &= ctx->ifc_softc_ctx.isc_capabilities | IFCAP_MEXTPG; 4589 setmask = 0; 4590 #ifdef TCP_OFFLOAD 4591 setmask |= mask & (IFCAP_TOE4 | IFCAP_TOE6); 4592 #endif 4593 setmask |= (mask & IFCAP_FLAGS); 4594 setmask |= (mask & IFCAP_WOL); 4595 4596 /* 4597 * If any RX csum has changed, change all the ones that 4598 * are supported by the driver. 4599 */ 4600 if (setmask & (IFCAP_RXCSUM | IFCAP_RXCSUM_IPV6)) { 4601 setmask |= ctx->ifc_softc_ctx.isc_capabilities & 4602 (IFCAP_RXCSUM | IFCAP_RXCSUM_IPV6); 4603 } 4604 4605 /* 4606 * want to ensure that traffic has stopped before we change any of the flags 4607 */ 4608 if (setmask) { 4609 CTX_LOCK(ctx); 4610 bits = if_getdrvflags(ifp); 4611 if (bits & IFF_DRV_RUNNING && setmask & ~IFCAP_WOL) 4612 iflib_stop(ctx); 4613 STATE_LOCK(ctx); 4614 if_togglecapenable(ifp, setmask); 4615 ctx->ifc_softc_ctx.isc_capenable ^= setmask; 4616 STATE_UNLOCK(ctx); 4617 if (bits & IFF_DRV_RUNNING && setmask & ~IFCAP_WOL) 4618 iflib_init_locked(ctx); 4619 STATE_LOCK(ctx); 4620 if_setdrvflags(ifp, bits); 4621 STATE_UNLOCK(ctx); 4622 CTX_UNLOCK(ctx); 4623 } 4624 if_vlancap(ifp); 4625 break; 4626 } 4627 case SIOCGPRIVATE_0: 4628 case SIOCSDRVSPEC: 4629 case SIOCGDRVSPEC: 4630 CTX_LOCK(ctx); 4631 err = IFDI_PRIV_IOCTL(ctx, command, data); 4632 CTX_UNLOCK(ctx); 4633 break; 4634 default: 4635 err = ether_ioctl(ifp, command, data); 4636 break; 4637 } 4638 if (reinit) 4639 iflib_if_init(ctx); 4640 return (err); 4641 } 4642 4643 static uint64_t 4644 iflib_if_get_counter(if_t ifp, ift_counter cnt) 4645 { 4646 if_ctx_t ctx = if_getsoftc(ifp); 4647 4648 return (IFDI_GET_COUNTER(ctx, cnt)); 4649 } 4650 4651 /********************************************************************* 4652 * 4653 * OTHER FUNCTIONS EXPORTED TO THE STACK 4654 * 4655 **********************************************************************/ 4656 4657 static void 4658 iflib_vlan_register(void *arg, if_t ifp, uint16_t vtag) 4659 { 4660 if_ctx_t ctx = if_getsoftc(ifp); 4661 4662 if ((void *)ctx != arg) 4663 return; 4664 4665 if ((vtag == 0) || (vtag > 4095)) 4666 return; 4667 4668 if (iflib_in_detach(ctx)) 4669 return; 4670 4671 CTX_LOCK(ctx); 4672 /* Driver may need all untagged packets to be flushed */ 4673 if (IFDI_NEEDS_RESTART(ctx, IFLIB_RESTART_VLAN_CONFIG)) 4674 iflib_stop(ctx); 4675 IFDI_VLAN_REGISTER(ctx, vtag); 4676 /* Re-init to load the changes, if required */ 4677 if (IFDI_NEEDS_RESTART(ctx, IFLIB_RESTART_VLAN_CONFIG)) 4678 iflib_init_locked(ctx); 4679 CTX_UNLOCK(ctx); 4680 } 4681 4682 static void 4683 iflib_vlan_unregister(void *arg, if_t ifp, uint16_t vtag) 4684 { 4685 if_ctx_t ctx = if_getsoftc(ifp); 4686 4687 if ((void *)ctx != arg) 4688 return; 4689 4690 if ((vtag == 0) || (vtag > 4095)) 4691 return; 4692 4693 CTX_LOCK(ctx); 4694 /* Driver may need all tagged packets to be flushed */ 4695 if (IFDI_NEEDS_RESTART(ctx, IFLIB_RESTART_VLAN_CONFIG)) 4696 iflib_stop(ctx); 4697 IFDI_VLAN_UNREGISTER(ctx, vtag); 4698 /* Re-init to load the changes, if required */ 4699 if (IFDI_NEEDS_RESTART(ctx, IFLIB_RESTART_VLAN_CONFIG)) 4700 iflib_init_locked(ctx); 4701 CTX_UNLOCK(ctx); 4702 } 4703 4704 static void 4705 iflib_led_func(void *arg, int onoff) 4706 { 4707 if_ctx_t ctx = arg; 4708 4709 CTX_LOCK(ctx); 4710 IFDI_LED_FUNC(ctx, onoff); 4711 CTX_UNLOCK(ctx); 4712 } 4713 4714 /********************************************************************* 4715 * 4716 * BUS FUNCTION DEFINITIONS 4717 * 4718 **********************************************************************/ 4719 4720 int 4721 iflib_device_probe(device_t dev) 4722 { 4723 const pci_vendor_info_t *ent; 4724 if_shared_ctx_t sctx; 4725 uint16_t pci_device_id, pci_rev_id, pci_subdevice_id, pci_subvendor_id; 4726 uint16_t pci_vendor_id; 4727 4728 if ((sctx = DEVICE_REGISTER(dev)) == NULL || sctx->isc_magic != IFLIB_MAGIC) 4729 return (ENOTSUP); 4730 4731 pci_vendor_id = pci_get_vendor(dev); 4732 pci_device_id = pci_get_device(dev); 4733 pci_subvendor_id = pci_get_subvendor(dev); 4734 pci_subdevice_id = pci_get_subdevice(dev); 4735 pci_rev_id = pci_get_revid(dev); 4736 if (sctx->isc_parse_devinfo != NULL) 4737 sctx->isc_parse_devinfo(&pci_device_id, &pci_subvendor_id, &pci_subdevice_id, &pci_rev_id); 4738 4739 ent = sctx->isc_vendor_info; 4740 while (ent->pvi_vendor_id != 0) { 4741 if (pci_vendor_id != ent->pvi_vendor_id) { 4742 ent++; 4743 continue; 4744 } 4745 if ((pci_device_id == ent->pvi_device_id) && 4746 ((pci_subvendor_id == ent->pvi_subvendor_id) || 4747 (ent->pvi_subvendor_id == 0)) && 4748 ((pci_subdevice_id == ent->pvi_subdevice_id) || 4749 (ent->pvi_subdevice_id == 0)) && 4750 ((pci_rev_id == ent->pvi_rev_id) || 4751 (ent->pvi_rev_id == 0))) { 4752 device_set_desc_copy(dev, ent->pvi_name); 4753 /* this needs to be changed to zero if the bus probing code 4754 * ever stops re-probing on best match because the sctx 4755 * may have its values over written by register calls 4756 * in subsequent probes 4757 */ 4758 return (BUS_PROBE_DEFAULT); 4759 } 4760 ent++; 4761 } 4762 return (ENXIO); 4763 } 4764 4765 int 4766 iflib_device_probe_vendor(device_t dev) 4767 { 4768 int probe; 4769 4770 probe = iflib_device_probe(dev); 4771 if (probe == BUS_PROBE_DEFAULT) 4772 return (BUS_PROBE_VENDOR); 4773 else 4774 return (probe); 4775 } 4776 4777 static void 4778 iflib_reset_qvalues(if_ctx_t ctx) 4779 { 4780 if_softc_ctx_t scctx = &ctx->ifc_softc_ctx; 4781 if_shared_ctx_t sctx = ctx->ifc_sctx; 4782 device_t dev = ctx->ifc_dev; 4783 int i; 4784 4785 if (ctx->ifc_sysctl_ntxqs != 0) 4786 scctx->isc_ntxqsets = ctx->ifc_sysctl_ntxqs; 4787 if (ctx->ifc_sysctl_nrxqs != 0) 4788 scctx->isc_nrxqsets = ctx->ifc_sysctl_nrxqs; 4789 4790 for (i = 0; i < sctx->isc_ntxqs; i++) { 4791 if (ctx->ifc_sysctl_ntxds[i] != 0) 4792 scctx->isc_ntxd[i] = ctx->ifc_sysctl_ntxds[i]; 4793 else 4794 scctx->isc_ntxd[i] = sctx->isc_ntxd_default[i]; 4795 } 4796 4797 for (i = 0; i < sctx->isc_nrxqs; i++) { 4798 if (ctx->ifc_sysctl_nrxds[i] != 0) 4799 scctx->isc_nrxd[i] = ctx->ifc_sysctl_nrxds[i]; 4800 else 4801 scctx->isc_nrxd[i] = sctx->isc_nrxd_default[i]; 4802 } 4803 4804 for (i = 0; i < sctx->isc_nrxqs; i++) { 4805 if (scctx->isc_nrxd[i] < sctx->isc_nrxd_min[i]) { 4806 device_printf(dev, "nrxd%d: %d less than nrxd_min %d - resetting to min\n", 4807 i, scctx->isc_nrxd[i], sctx->isc_nrxd_min[i]); 4808 scctx->isc_nrxd[i] = sctx->isc_nrxd_min[i]; 4809 } 4810 if (scctx->isc_nrxd[i] > sctx->isc_nrxd_max[i]) { 4811 device_printf(dev, "nrxd%d: %d greater than nrxd_max %d - resetting to max\n", 4812 i, scctx->isc_nrxd[i], sctx->isc_nrxd_max[i]); 4813 scctx->isc_nrxd[i] = sctx->isc_nrxd_max[i]; 4814 } 4815 if (!powerof2(scctx->isc_nrxd[i])) { 4816 device_printf(dev, "nrxd%d: %d is not a power of 2 - using default value of %d\n", 4817 i, scctx->isc_nrxd[i], sctx->isc_nrxd_default[i]); 4818 scctx->isc_nrxd[i] = sctx->isc_nrxd_default[i]; 4819 } 4820 } 4821 4822 for (i = 0; i < sctx->isc_ntxqs; i++) { 4823 if (scctx->isc_ntxd[i] < sctx->isc_ntxd_min[i]) { 4824 device_printf(dev, "ntxd%d: %d less than ntxd_min %d - resetting to min\n", 4825 i, scctx->isc_ntxd[i], sctx->isc_ntxd_min[i]); 4826 scctx->isc_ntxd[i] = sctx->isc_ntxd_min[i]; 4827 } 4828 if (scctx->isc_ntxd[i] > sctx->isc_ntxd_max[i]) { 4829 device_printf(dev, "ntxd%d: %d greater than ntxd_max %d - resetting to max\n", 4830 i, scctx->isc_ntxd[i], sctx->isc_ntxd_max[i]); 4831 scctx->isc_ntxd[i] = sctx->isc_ntxd_max[i]; 4832 } 4833 if (!powerof2(scctx->isc_ntxd[i])) { 4834 device_printf(dev, "ntxd%d: %d is not a power of 2 - using default value of %d\n", 4835 i, scctx->isc_ntxd[i], sctx->isc_ntxd_default[i]); 4836 scctx->isc_ntxd[i] = sctx->isc_ntxd_default[i]; 4837 } 4838 } 4839 } 4840 4841 static void 4842 iflib_add_pfil(if_ctx_t ctx) 4843 { 4844 struct pfil_head *pfil; 4845 struct pfil_head_args pa; 4846 iflib_rxq_t rxq; 4847 int i; 4848 4849 pa.pa_version = PFIL_VERSION; 4850 pa.pa_flags = PFIL_IN; 4851 pa.pa_type = PFIL_TYPE_ETHERNET; 4852 pa.pa_headname = if_name(ctx->ifc_ifp); 4853 pfil = pfil_head_register(&pa); 4854 4855 for (i = 0, rxq = ctx->ifc_rxqs; i < NRXQSETS(ctx); i++, rxq++) { 4856 rxq->pfil = pfil; 4857 } 4858 } 4859 4860 static void 4861 iflib_rem_pfil(if_ctx_t ctx) 4862 { 4863 struct pfil_head *pfil; 4864 iflib_rxq_t rxq; 4865 int i; 4866 4867 rxq = ctx->ifc_rxqs; 4868 pfil = rxq->pfil; 4869 for (i = 0; i < NRXQSETS(ctx); i++, rxq++) { 4870 rxq->pfil = NULL; 4871 } 4872 pfil_head_unregister(pfil); 4873 } 4874 4875 4876 /* 4877 * Advance forward by n members of the cpuset ctx->ifc_cpus starting from 4878 * cpuid and wrapping as necessary. 4879 */ 4880 static unsigned int 4881 cpuid_advance(if_ctx_t ctx, unsigned int cpuid, unsigned int n) 4882 { 4883 unsigned int first_valid; 4884 unsigned int last_valid; 4885 4886 /* cpuid should always be in the valid set */ 4887 MPASS(CPU_ISSET(cpuid, &ctx->ifc_cpus)); 4888 4889 /* valid set should never be empty */ 4890 MPASS(!CPU_EMPTY(&ctx->ifc_cpus)); 4891 4892 first_valid = CPU_FFS(&ctx->ifc_cpus) - 1; 4893 last_valid = CPU_FLS(&ctx->ifc_cpus) - 1; 4894 n = n % CPU_COUNT(&ctx->ifc_cpus); 4895 while (n > 0) { 4896 do { 4897 cpuid++; 4898 if (cpuid > last_valid) 4899 cpuid = first_valid; 4900 } while (!CPU_ISSET(cpuid, &ctx->ifc_cpus)); 4901 n--; 4902 } 4903 4904 return (cpuid); 4905 } 4906 4907 #if defined(SMP) && defined(SCHED_ULE) 4908 extern struct cpu_group *cpu_top; /* CPU topology */ 4909 4910 static int 4911 find_child_with_core(int cpu, struct cpu_group *grp) 4912 { 4913 int i; 4914 4915 if (grp->cg_children == 0) 4916 return (-1); 4917 4918 MPASS(grp->cg_child); 4919 for (i = 0; i < grp->cg_children; i++) { 4920 if (CPU_ISSET(cpu, &grp->cg_child[i].cg_mask)) 4921 return (i); 4922 } 4923 4924 return (-1); 4925 } 4926 4927 4928 /* 4929 * Find an L2 neighbor of the given CPU or return -1 if none found. This 4930 * does not distinguish among multiple L2 neighbors if the given CPU has 4931 * more than one (it will always return the same result in that case). 4932 */ 4933 static int 4934 find_l2_neighbor(int cpu) 4935 { 4936 struct cpu_group *grp; 4937 int i; 4938 4939 grp = cpu_top; 4940 if (grp == NULL) 4941 return (-1); 4942 4943 /* 4944 * Find the smallest CPU group that contains the given core. 4945 */ 4946 i = 0; 4947 while ((i = find_child_with_core(cpu, grp)) != -1) { 4948 /* 4949 * If the smallest group containing the given CPU has less 4950 * than two members, we conclude the given CPU has no 4951 * L2 neighbor. 4952 */ 4953 if (grp->cg_child[i].cg_count <= 1) 4954 return (-1); 4955 grp = &grp->cg_child[i]; 4956 } 4957 4958 /* Must share L2. */ 4959 if (grp->cg_level > CG_SHARE_L2 || grp->cg_level == CG_SHARE_NONE) 4960 return (-1); 4961 4962 /* 4963 * Select the first member of the set that isn't the reference 4964 * CPU, which at this point is guaranteed to exist. 4965 */ 4966 for (i = 0; i < CPU_SETSIZE; i++) { 4967 if (CPU_ISSET(i, &grp->cg_mask) && i != cpu) 4968 return (i); 4969 } 4970 4971 /* Should never be reached */ 4972 return (-1); 4973 } 4974 4975 #else 4976 static int 4977 find_l2_neighbor(int cpu) 4978 { 4979 4980 return (-1); 4981 } 4982 #endif 4983 4984 /* 4985 * CPU mapping behaviors 4986 * --------------------- 4987 * 'separate txrx' refers to the separate_txrx sysctl 4988 * 'use logical' refers to the use_logical_cores sysctl 4989 * 'INTR CPUS' indicates whether bus_get_cpus(INTR_CPUS) succeeded 4990 * 4991 * separate use INTR 4992 * txrx logical CPUS result 4993 * ---------- --------- ------ ------------------------------------------------ 4994 * - - X RX and TX queues mapped to consecutive physical 4995 * cores with RX/TX pairs on same core and excess 4996 * of either following 4997 * - X X RX and TX queues mapped to consecutive cores 4998 * of any type with RX/TX pairs on same core and 4999 * excess of either following 5000 * X - X RX and TX queues mapped to consecutive physical 5001 * cores; all RX then all TX 5002 * X X X RX queues mapped to consecutive physical cores 5003 * first, then TX queues mapped to L2 neighbor of 5004 * the corresponding RX queue if one exists, 5005 * otherwise to consecutive physical cores 5006 * - n/a - RX and TX queues mapped to consecutive cores of 5007 * any type with RX/TX pairs on same core and excess 5008 * of either following 5009 * X n/a - RX and TX queues mapped to consecutive cores of 5010 * any type; all RX then all TX 5011 */ 5012 static unsigned int 5013 get_cpuid_for_queue(if_ctx_t ctx, unsigned int base_cpuid, unsigned int qid, 5014 bool is_tx) 5015 { 5016 if_softc_ctx_t scctx = &ctx->ifc_softc_ctx; 5017 unsigned int core_index; 5018 5019 if (ctx->ifc_sysctl_separate_txrx) { 5020 /* 5021 * When using separate CPUs for TX and RX, the assignment 5022 * will always be of a consecutive CPU out of the set of 5023 * context CPUs, except for the specific case where the 5024 * context CPUs are phsyical cores, the use of logical cores 5025 * has been enabled, the assignment is for TX, the TX qid 5026 * corresponds to an RX qid, and the CPU assigned to the 5027 * corresponding RX queue has an L2 neighbor. 5028 */ 5029 if (ctx->ifc_sysctl_use_logical_cores && 5030 ctx->ifc_cpus_are_physical_cores && 5031 is_tx && qid < scctx->isc_nrxqsets) { 5032 int l2_neighbor; 5033 unsigned int rx_cpuid; 5034 5035 rx_cpuid = cpuid_advance(ctx, base_cpuid, qid); 5036 l2_neighbor = find_l2_neighbor(rx_cpuid); 5037 if (l2_neighbor != -1) { 5038 return (l2_neighbor); 5039 } 5040 /* 5041 * ... else fall through to the normal 5042 * consecutive-after-RX assignment scheme. 5043 * 5044 * Note that we are assuming that all RX queue CPUs 5045 * have an L2 neighbor, or all do not. If a mixed 5046 * scenario is possible, we will have to keep track 5047 * separately of how many queues prior to this one 5048 * were not able to be assigned to an L2 neighbor. 5049 */ 5050 } 5051 if (is_tx) 5052 core_index = scctx->isc_nrxqsets + qid; 5053 else 5054 core_index = qid; 5055 } else { 5056 core_index = qid; 5057 } 5058 5059 return (cpuid_advance(ctx, base_cpuid, core_index)); 5060 } 5061 5062 static uint16_t 5063 get_ctx_core_offset(if_ctx_t ctx) 5064 { 5065 if_softc_ctx_t scctx = &ctx->ifc_softc_ctx; 5066 struct cpu_offset *op; 5067 cpuset_t assigned_cpus; 5068 unsigned int cores_consumed; 5069 unsigned int base_cpuid = ctx->ifc_sysctl_core_offset; 5070 unsigned int first_valid; 5071 unsigned int last_valid; 5072 unsigned int i; 5073 5074 first_valid = CPU_FFS(&ctx->ifc_cpus) - 1; 5075 last_valid = CPU_FLS(&ctx->ifc_cpus) - 1; 5076 5077 if (base_cpuid != CORE_OFFSET_UNSPECIFIED) { 5078 /* 5079 * Align the user-chosen base CPU ID to the next valid CPU 5080 * for this device. If the chosen base CPU ID is smaller 5081 * than the first valid CPU or larger than the last valid 5082 * CPU, we assume the user does not know what the valid 5083 * range is for this device and is thinking in terms of a 5084 * zero-based reference frame, and so we shift the given 5085 * value into the valid range (and wrap accordingly) so the 5086 * intent is translated to the proper frame of reference. 5087 * If the base CPU ID is within the valid first/last, but 5088 * does not correspond to a valid CPU, it is advanced to the 5089 * next valid CPU (wrapping if necessary). 5090 */ 5091 if (base_cpuid < first_valid || base_cpuid > last_valid) { 5092 /* shift from zero-based to first_valid-based */ 5093 base_cpuid += first_valid; 5094 /* wrap to range [first_valid, last_valid] */ 5095 base_cpuid = (base_cpuid - first_valid) % 5096 (last_valid - first_valid + 1); 5097 } 5098 if (!CPU_ISSET(base_cpuid, &ctx->ifc_cpus)) { 5099 /* 5100 * base_cpuid is in [first_valid, last_valid], but 5101 * not a member of the valid set. In this case, 5102 * there will always be a member of the valid set 5103 * with a CPU ID that is greater than base_cpuid, 5104 * and we simply advance to it. 5105 */ 5106 while (!CPU_ISSET(base_cpuid, &ctx->ifc_cpus)) 5107 base_cpuid++; 5108 } 5109 return (base_cpuid); 5110 } 5111 5112 /* 5113 * Determine how many cores will be consumed by performing the CPU 5114 * assignments and counting how many of the assigned CPUs correspond 5115 * to CPUs in the set of context CPUs. This is done using the CPU 5116 * ID first_valid as the base CPU ID, as the base CPU must be within 5117 * the set of context CPUs. 5118 * 5119 * Note not all assigned CPUs will be in the set of context CPUs 5120 * when separate CPUs are being allocated to TX and RX queues, 5121 * assignment to logical cores has been enabled, the set of context 5122 * CPUs contains only physical CPUs, and TX queues are mapped to L2 5123 * neighbors of CPUs that RX queues have been mapped to - in this 5124 * case we do only want to count how many CPUs in the set of context 5125 * CPUs have been consumed, as that determines the next CPU in that 5126 * set to start allocating at for the next device for which 5127 * core_offset is not set. 5128 */ 5129 CPU_ZERO(&assigned_cpus); 5130 for (i = 0; i < scctx->isc_ntxqsets; i++) 5131 CPU_SET(get_cpuid_for_queue(ctx, first_valid, i, true), 5132 &assigned_cpus); 5133 for (i = 0; i < scctx->isc_nrxqsets; i++) 5134 CPU_SET(get_cpuid_for_queue(ctx, first_valid, i, false), 5135 &assigned_cpus); 5136 CPU_AND(&assigned_cpus, &assigned_cpus, &ctx->ifc_cpus); 5137 cores_consumed = CPU_COUNT(&assigned_cpus); 5138 5139 mtx_lock(&cpu_offset_mtx); 5140 SLIST_FOREACH(op, &cpu_offsets, entries) { 5141 if (CPU_CMP(&ctx->ifc_cpus, &op->set) == 0) { 5142 base_cpuid = op->next_cpuid; 5143 op->next_cpuid = cpuid_advance(ctx, op->next_cpuid, 5144 cores_consumed); 5145 MPASS(op->refcount < UINT_MAX); 5146 op->refcount++; 5147 break; 5148 } 5149 } 5150 if (base_cpuid == CORE_OFFSET_UNSPECIFIED) { 5151 base_cpuid = first_valid; 5152 op = malloc(sizeof(struct cpu_offset), M_IFLIB, 5153 M_NOWAIT | M_ZERO); 5154 if (op == NULL) { 5155 device_printf(ctx->ifc_dev, 5156 "allocation for cpu offset failed.\n"); 5157 } else { 5158 op->next_cpuid = cpuid_advance(ctx, base_cpuid, 5159 cores_consumed); 5160 op->refcount = 1; 5161 CPU_COPY(&ctx->ifc_cpus, &op->set); 5162 SLIST_INSERT_HEAD(&cpu_offsets, op, entries); 5163 } 5164 } 5165 mtx_unlock(&cpu_offset_mtx); 5166 5167 return (base_cpuid); 5168 } 5169 5170 static void 5171 unref_ctx_core_offset(if_ctx_t ctx) 5172 { 5173 struct cpu_offset *op, *top; 5174 5175 mtx_lock(&cpu_offset_mtx); 5176 SLIST_FOREACH_SAFE(op, &cpu_offsets, entries, top) { 5177 if (CPU_CMP(&ctx->ifc_cpus, &op->set) == 0) { 5178 MPASS(op->refcount > 0); 5179 op->refcount--; 5180 if (op->refcount == 0) { 5181 SLIST_REMOVE(&cpu_offsets, op, cpu_offset, entries); 5182 free(op, M_IFLIB); 5183 } 5184 break; 5185 } 5186 } 5187 mtx_unlock(&cpu_offset_mtx); 5188 } 5189 5190 int 5191 iflib_device_register(device_t dev, void *sc, if_shared_ctx_t sctx, if_ctx_t *ctxp) 5192 { 5193 if_ctx_t ctx; 5194 if_t ifp; 5195 if_softc_ctx_t scctx; 5196 kobjop_desc_t kobj_desc; 5197 kobj_method_t *kobj_method; 5198 int err, msix, rid; 5199 int num_txd, num_rxd; 5200 5201 ctx = malloc(sizeof(*ctx), M_IFLIB, M_WAITOK | M_ZERO); 5202 5203 if (sc == NULL) { 5204 sc = malloc(sctx->isc_driver->size, M_IFLIB, M_WAITOK | M_ZERO); 5205 device_set_softc(dev, ctx); 5206 ctx->ifc_flags |= IFC_SC_ALLOCATED; 5207 } 5208 5209 ctx->ifc_sctx = sctx; 5210 ctx->ifc_dev = dev; 5211 ctx->ifc_softc = sc; 5212 5213 if ((err = iflib_register(ctx)) != 0) { 5214 device_printf(dev, "iflib_register failed %d\n", err); 5215 goto fail_ctx_free; 5216 } 5217 iflib_add_device_sysctl_pre(ctx); 5218 5219 scctx = &ctx->ifc_softc_ctx; 5220 ifp = ctx->ifc_ifp; 5221 5222 iflib_reset_qvalues(ctx); 5223 IFNET_WLOCK(); 5224 CTX_LOCK(ctx); 5225 if ((err = IFDI_ATTACH_PRE(ctx)) != 0) { 5226 device_printf(dev, "IFDI_ATTACH_PRE failed %d\n", err); 5227 goto fail_unlock; 5228 } 5229 _iflib_pre_assert(scctx); 5230 ctx->ifc_txrx = *scctx->isc_txrx; 5231 5232 MPASS(scctx->isc_dma_width <= flsll(BUS_SPACE_MAXADDR)); 5233 5234 if (sctx->isc_flags & IFLIB_DRIVER_MEDIA) 5235 ctx->ifc_mediap = scctx->isc_media; 5236 5237 #ifdef INVARIANTS 5238 if (scctx->isc_capabilities & IFCAP_TXCSUM) 5239 MPASS(scctx->isc_tx_csum_flags); 5240 #endif 5241 5242 if_setcapabilities(ifp, 5243 scctx->isc_capabilities | IFCAP_HWSTATS | IFCAP_MEXTPG); 5244 if_setcapenable(ifp, 5245 scctx->isc_capenable | IFCAP_HWSTATS | IFCAP_MEXTPG); 5246 5247 if (scctx->isc_ntxqsets == 0 || (scctx->isc_ntxqsets_max && scctx->isc_ntxqsets_max < scctx->isc_ntxqsets)) 5248 scctx->isc_ntxqsets = scctx->isc_ntxqsets_max; 5249 if (scctx->isc_nrxqsets == 0 || (scctx->isc_nrxqsets_max && scctx->isc_nrxqsets_max < scctx->isc_nrxqsets)) 5250 scctx->isc_nrxqsets = scctx->isc_nrxqsets_max; 5251 5252 num_txd = iflib_num_tx_descs(ctx); 5253 num_rxd = iflib_num_rx_descs(ctx); 5254 5255 /* XXX change for per-queue sizes */ 5256 device_printf(dev, "Using %d TX descriptors and %d RX descriptors\n", 5257 num_txd, num_rxd); 5258 5259 if (scctx->isc_tx_nsegments > num_txd / MAX_SINGLE_PACKET_FRACTION) 5260 scctx->isc_tx_nsegments = max(1, num_txd / 5261 MAX_SINGLE_PACKET_FRACTION); 5262 if (scctx->isc_tx_tso_segments_max > num_txd / 5263 MAX_SINGLE_PACKET_FRACTION) 5264 scctx->isc_tx_tso_segments_max = max(1, 5265 num_txd / MAX_SINGLE_PACKET_FRACTION); 5266 5267 /* TSO parameters - dig these out of the data sheet - simply correspond to tag setup */ 5268 if (if_getcapabilities(ifp) & IFCAP_TSO) { 5269 /* 5270 * The stack can't handle a TSO size larger than IP_MAXPACKET, 5271 * but some MACs do. 5272 */ 5273 if_sethwtsomax(ifp, min(scctx->isc_tx_tso_size_max, 5274 IP_MAXPACKET)); 5275 /* 5276 * Take maximum number of m_pullup(9)'s in iflib_parse_header() 5277 * into account. In the worst case, each of these calls will 5278 * add another mbuf and, thus, the requirement for another DMA 5279 * segment. So for best performance, it doesn't make sense to 5280 * advertize a maximum of TSO segments that typically will 5281 * require defragmentation in iflib_encap(). 5282 */ 5283 if_sethwtsomaxsegcount(ifp, scctx->isc_tx_tso_segments_max - 3); 5284 if_sethwtsomaxsegsize(ifp, scctx->isc_tx_tso_segsize_max); 5285 } 5286 if (scctx->isc_rss_table_size == 0) 5287 scctx->isc_rss_table_size = 64; 5288 scctx->isc_rss_table_mask = scctx->isc_rss_table_size - 1; 5289 5290 GROUPTASK_INIT(&ctx->ifc_admin_task, 0, _task_fn_admin, ctx); 5291 /* XXX format name */ 5292 taskqgroup_attach(qgroup_if_config_tqg, &ctx->ifc_admin_task, ctx, 5293 NULL, NULL, "admin"); 5294 5295 /* Set up cpu set. If it fails, use the set of all CPUs. */ 5296 if (bus_get_cpus(dev, INTR_CPUS, sizeof(ctx->ifc_cpus), &ctx->ifc_cpus) != 0) { 5297 device_printf(dev, "Unable to fetch CPU list\n"); 5298 CPU_COPY(&all_cpus, &ctx->ifc_cpus); 5299 ctx->ifc_cpus_are_physical_cores = false; 5300 } else 5301 ctx->ifc_cpus_are_physical_cores = true; 5302 MPASS(CPU_COUNT(&ctx->ifc_cpus) > 0); 5303 5304 /* 5305 * Now set up MSI or MSI-X, should return us the number of supported 5306 * vectors (will be 1 for a legacy interrupt and MSI). 5307 */ 5308 if (sctx->isc_flags & IFLIB_SKIP_MSIX) { 5309 msix = scctx->isc_vectors; 5310 } else if (scctx->isc_msix_bar != 0) 5311 /* 5312 * The simple fact that isc_msix_bar is not 0 does not mean we 5313 * we have a good value there that is known to work. 5314 */ 5315 msix = iflib_msix_init(ctx); 5316 else { 5317 scctx->isc_vectors = 1; 5318 scctx->isc_ntxqsets = 1; 5319 scctx->isc_nrxqsets = 1; 5320 scctx->isc_intr = IFLIB_INTR_LEGACY; 5321 msix = 0; 5322 } 5323 /* Get memory for the station queues */ 5324 if ((err = iflib_queues_alloc(ctx))) { 5325 device_printf(dev, "Unable to allocate queue memory\n"); 5326 goto fail_intr_free; 5327 } 5328 5329 if ((err = iflib_qset_structures_setup(ctx))) 5330 goto fail_queues; 5331 5332 /* 5333 * Now that we know how many queues there are, get the core offset. 5334 */ 5335 ctx->ifc_sysctl_core_offset = get_ctx_core_offset(ctx); 5336 5337 if (msix > 1) { 5338 /* 5339 * When using MSI-X, ensure that ifdi_{r,t}x_queue_intr_enable 5340 * aren't the default NULL implementation. 5341 */ 5342 kobj_desc = &ifdi_rx_queue_intr_enable_desc; 5343 kobj_method = kobj_lookup_method(((kobj_t)ctx)->ops->cls, NULL, 5344 kobj_desc); 5345 if (kobj_method == &kobj_desc->deflt) { 5346 device_printf(dev, 5347 "MSI-X requires ifdi_rx_queue_intr_enable method"); 5348 err = EOPNOTSUPP; 5349 goto fail_queues; 5350 } 5351 kobj_desc = &ifdi_tx_queue_intr_enable_desc; 5352 kobj_method = kobj_lookup_method(((kobj_t)ctx)->ops->cls, NULL, 5353 kobj_desc); 5354 if (kobj_method == &kobj_desc->deflt) { 5355 device_printf(dev, 5356 "MSI-X requires ifdi_tx_queue_intr_enable method"); 5357 err = EOPNOTSUPP; 5358 goto fail_queues; 5359 } 5360 5361 /* 5362 * Assign the MSI-X vectors. 5363 * Note that the default NULL ifdi_msix_intr_assign method will 5364 * fail here, too. 5365 */ 5366 err = IFDI_MSIX_INTR_ASSIGN(ctx, msix); 5367 if (err != 0) { 5368 device_printf(dev, "IFDI_MSIX_INTR_ASSIGN failed %d\n", 5369 err); 5370 goto fail_queues; 5371 } 5372 } else if (scctx->isc_intr != IFLIB_INTR_MSIX) { 5373 rid = 0; 5374 if (scctx->isc_intr == IFLIB_INTR_MSI) { 5375 MPASS(msix == 1); 5376 rid = 1; 5377 } 5378 if ((err = iflib_legacy_setup(ctx, ctx->isc_legacy_intr, ctx->ifc_softc, &rid, "irq0")) != 0) { 5379 device_printf(dev, "iflib_legacy_setup failed %d\n", err); 5380 goto fail_queues; 5381 } 5382 } else { 5383 device_printf(dev, 5384 "Cannot use iflib with only 1 MSI-X interrupt!\n"); 5385 err = ENODEV; 5386 goto fail_queues; 5387 } 5388 5389 /* 5390 * It prevents a double-locking panic with iflib_media_status when 5391 * the driver loads. 5392 */ 5393 CTX_UNLOCK(ctx); 5394 ether_ifattach(ctx->ifc_ifp, ctx->ifc_mac.octet); 5395 CTX_LOCK(ctx); 5396 5397 if ((err = IFDI_ATTACH_POST(ctx)) != 0) { 5398 device_printf(dev, "IFDI_ATTACH_POST failed %d\n", err); 5399 goto fail_detach; 5400 } 5401 5402 /* 5403 * Tell the upper layer(s) if IFCAP_VLAN_MTU is supported. 5404 * This must appear after the call to ether_ifattach() because 5405 * ether_ifattach() sets if_hdrlen to the default value. 5406 */ 5407 if (if_getcapabilities(ifp) & IFCAP_VLAN_MTU) 5408 if_setifheaderlen(ifp, sizeof(struct ether_vlan_header)); 5409 5410 if ((err = iflib_netmap_attach(ctx))) { 5411 device_printf(ctx->ifc_dev, "netmap attach failed: %d\n", err); 5412 goto fail_detach; 5413 } 5414 *ctxp = ctx; 5415 5416 DEBUGNET_SET(ctx->ifc_ifp, iflib); 5417 5418 if_setgetcounterfn(ctx->ifc_ifp, iflib_if_get_counter); 5419 iflib_add_device_sysctl_post(ctx); 5420 iflib_add_pfil(ctx); 5421 ctx->ifc_flags |= IFC_INIT_DONE; 5422 CTX_UNLOCK(ctx); 5423 IFNET_WUNLOCK(); 5424 5425 return (0); 5426 5427 fail_detach: 5428 ether_ifdetach(ctx->ifc_ifp); 5429 fail_queues: 5430 iflib_tqg_detach(ctx); 5431 iflib_tx_structures_free(ctx); 5432 iflib_rx_structures_free(ctx); 5433 IFDI_DETACH(ctx); 5434 IFDI_QUEUES_FREE(ctx); 5435 fail_intr_free: 5436 iflib_free_intr_mem(ctx); 5437 fail_unlock: 5438 CTX_UNLOCK(ctx); 5439 IFNET_WUNLOCK(); 5440 iflib_deregister(ctx); 5441 fail_ctx_free: 5442 device_set_softc(ctx->ifc_dev, NULL); 5443 if (ctx->ifc_flags & IFC_SC_ALLOCATED) 5444 free(ctx->ifc_softc, M_IFLIB); 5445 free(ctx, M_IFLIB); 5446 return (err); 5447 } 5448 5449 int 5450 iflib_device_attach(device_t dev) 5451 { 5452 if_ctx_t ctx; 5453 if_shared_ctx_t sctx; 5454 5455 if ((sctx = DEVICE_REGISTER(dev)) == NULL || sctx->isc_magic != IFLIB_MAGIC) 5456 return (ENOTSUP); 5457 5458 pci_enable_busmaster(dev); 5459 5460 return (iflib_device_register(dev, NULL, sctx, &ctx)); 5461 } 5462 5463 int 5464 iflib_device_deregister(if_ctx_t ctx) 5465 { 5466 if_t ifp = ctx->ifc_ifp; 5467 device_t dev = ctx->ifc_dev; 5468 5469 /* Make sure VLANS are not using driver */ 5470 if (if_vlantrunkinuse(ifp)) { 5471 device_printf(dev, "Vlan in use, detach first\n"); 5472 return (EBUSY); 5473 } 5474 #ifdef PCI_IOV 5475 if (!CTX_IS_VF(ctx) && pci_iov_detach(dev) != 0) { 5476 device_printf(dev, "SR-IOV in use; detach first.\n"); 5477 return (EBUSY); 5478 } 5479 #endif 5480 5481 STATE_LOCK(ctx); 5482 ctx->ifc_flags |= IFC_IN_DETACH; 5483 STATE_UNLOCK(ctx); 5484 5485 /* Unregister VLAN handlers before calling iflib_stop() */ 5486 iflib_unregister_vlan_handlers(ctx); 5487 5488 iflib_netmap_detach(ifp); 5489 ether_ifdetach(ifp); 5490 5491 CTX_LOCK(ctx); 5492 iflib_stop(ctx); 5493 CTX_UNLOCK(ctx); 5494 5495 iflib_rem_pfil(ctx); 5496 if (ctx->ifc_led_dev != NULL) 5497 led_destroy(ctx->ifc_led_dev); 5498 5499 iflib_tqg_detach(ctx); 5500 iflib_tx_structures_free(ctx); 5501 iflib_rx_structures_free(ctx); 5502 5503 CTX_LOCK(ctx); 5504 IFDI_DETACH(ctx); 5505 IFDI_QUEUES_FREE(ctx); 5506 CTX_UNLOCK(ctx); 5507 5508 /* ether_ifdetach calls if_qflush - lock must be destroy afterwards*/ 5509 iflib_free_intr_mem(ctx); 5510 5511 bus_generic_detach(dev); 5512 5513 iflib_deregister(ctx); 5514 5515 device_set_softc(ctx->ifc_dev, NULL); 5516 if (ctx->ifc_flags & IFC_SC_ALLOCATED) 5517 free(ctx->ifc_softc, M_IFLIB); 5518 unref_ctx_core_offset(ctx); 5519 free(ctx, M_IFLIB); 5520 return (0); 5521 } 5522 5523 static void 5524 iflib_tqg_detach(if_ctx_t ctx) 5525 { 5526 iflib_txq_t txq; 5527 iflib_rxq_t rxq; 5528 int i; 5529 struct taskqgroup *tqg; 5530 5531 /* XXX drain any dependent tasks */ 5532 tqg = qgroup_if_io_tqg; 5533 for (txq = ctx->ifc_txqs, i = 0; i < NTXQSETS(ctx); i++, txq++) { 5534 callout_drain(&txq->ift_timer); 5535 #ifdef DEV_NETMAP 5536 callout_drain(&txq->ift_netmap_timer); 5537 #endif /* DEV_NETMAP */ 5538 if (txq->ift_task.gt_uniq != NULL) 5539 taskqgroup_detach(tqg, &txq->ift_task); 5540 } 5541 for (i = 0, rxq = ctx->ifc_rxqs; i < NRXQSETS(ctx); i++, rxq++) { 5542 if (rxq->ifr_task.gt_uniq != NULL) 5543 taskqgroup_detach(tqg, &rxq->ifr_task); 5544 } 5545 tqg = qgroup_if_config_tqg; 5546 if (ctx->ifc_admin_task.gt_uniq != NULL) 5547 taskqgroup_detach(tqg, &ctx->ifc_admin_task); 5548 if (ctx->ifc_vflr_task.gt_uniq != NULL) 5549 taskqgroup_detach(tqg, &ctx->ifc_vflr_task); 5550 } 5551 5552 static void 5553 iflib_free_intr_mem(if_ctx_t ctx) 5554 { 5555 5556 if (ctx->ifc_softc_ctx.isc_intr != IFLIB_INTR_MSIX) { 5557 iflib_irq_free(ctx, &ctx->ifc_legacy_irq); 5558 } 5559 if (ctx->ifc_softc_ctx.isc_intr != IFLIB_INTR_LEGACY) { 5560 pci_release_msi(ctx->ifc_dev); 5561 } 5562 if (ctx->ifc_msix_mem != NULL) { 5563 bus_release_resource(ctx->ifc_dev, SYS_RES_MEMORY, 5564 rman_get_rid(ctx->ifc_msix_mem), ctx->ifc_msix_mem); 5565 ctx->ifc_msix_mem = NULL; 5566 } 5567 } 5568 5569 int 5570 iflib_device_detach(device_t dev) 5571 { 5572 if_ctx_t ctx = device_get_softc(dev); 5573 5574 return (iflib_device_deregister(ctx)); 5575 } 5576 5577 int 5578 iflib_device_suspend(device_t dev) 5579 { 5580 if_ctx_t ctx = device_get_softc(dev); 5581 5582 CTX_LOCK(ctx); 5583 IFDI_SUSPEND(ctx); 5584 CTX_UNLOCK(ctx); 5585 5586 return (bus_generic_suspend(dev)); 5587 } 5588 int 5589 iflib_device_shutdown(device_t dev) 5590 { 5591 if_ctx_t ctx = device_get_softc(dev); 5592 5593 CTX_LOCK(ctx); 5594 IFDI_SHUTDOWN(ctx); 5595 CTX_UNLOCK(ctx); 5596 5597 return (bus_generic_suspend(dev)); 5598 } 5599 5600 int 5601 iflib_device_resume(device_t dev) 5602 { 5603 if_ctx_t ctx = device_get_softc(dev); 5604 iflib_txq_t txq = ctx->ifc_txqs; 5605 5606 CTX_LOCK(ctx); 5607 IFDI_RESUME(ctx); 5608 iflib_if_init_locked(ctx); 5609 CTX_UNLOCK(ctx); 5610 for (int i = 0; i < NTXQSETS(ctx); i++, txq++) 5611 iflib_txq_check_drain(txq, IFLIB_RESTART_BUDGET); 5612 5613 return (bus_generic_resume(dev)); 5614 } 5615 5616 int 5617 iflib_device_iov_init(device_t dev, uint16_t num_vfs, const nvlist_t *params) 5618 { 5619 int error; 5620 if_ctx_t ctx = device_get_softc(dev); 5621 5622 CTX_LOCK(ctx); 5623 error = IFDI_IOV_INIT(ctx, num_vfs, params); 5624 CTX_UNLOCK(ctx); 5625 5626 return (error); 5627 } 5628 5629 void 5630 iflib_device_iov_uninit(device_t dev) 5631 { 5632 if_ctx_t ctx = device_get_softc(dev); 5633 5634 CTX_LOCK(ctx); 5635 IFDI_IOV_UNINIT(ctx); 5636 CTX_UNLOCK(ctx); 5637 } 5638 5639 int 5640 iflib_device_iov_add_vf(device_t dev, uint16_t vfnum, const nvlist_t *params) 5641 { 5642 int error; 5643 if_ctx_t ctx = device_get_softc(dev); 5644 5645 CTX_LOCK(ctx); 5646 error = IFDI_IOV_VF_ADD(ctx, vfnum, params); 5647 CTX_UNLOCK(ctx); 5648 5649 return (error); 5650 } 5651 5652 /********************************************************************* 5653 * 5654 * MODULE FUNCTION DEFINITIONS 5655 * 5656 **********************************************************************/ 5657 5658 /* 5659 * - Start a fast taskqueue thread for each core 5660 * - Start a taskqueue for control operations 5661 */ 5662 static int 5663 iflib_module_init(void) 5664 { 5665 iflib_timer_default = hz / 2; 5666 return (0); 5667 } 5668 5669 static int 5670 iflib_module_event_handler(module_t mod, int what, void *arg) 5671 { 5672 int err; 5673 5674 switch (what) { 5675 case MOD_LOAD: 5676 if ((err = iflib_module_init()) != 0) 5677 return (err); 5678 break; 5679 case MOD_UNLOAD: 5680 return (EBUSY); 5681 default: 5682 return (EOPNOTSUPP); 5683 } 5684 5685 return (0); 5686 } 5687 5688 /********************************************************************* 5689 * 5690 * PUBLIC FUNCTION DEFINITIONS 5691 * ordered as in iflib.h 5692 * 5693 **********************************************************************/ 5694 5695 static void 5696 _iflib_assert(if_shared_ctx_t sctx) 5697 { 5698 int i; 5699 5700 MPASS(sctx->isc_tx_maxsize); 5701 MPASS(sctx->isc_tx_maxsegsize); 5702 5703 MPASS(sctx->isc_rx_maxsize); 5704 MPASS(sctx->isc_rx_nsegments); 5705 MPASS(sctx->isc_rx_maxsegsize); 5706 5707 MPASS(sctx->isc_nrxqs >= 1 && sctx->isc_nrxqs <= 8); 5708 for (i = 0; i < sctx->isc_nrxqs; i++) { 5709 MPASS(sctx->isc_nrxd_min[i]); 5710 MPASS(powerof2(sctx->isc_nrxd_min[i])); 5711 MPASS(sctx->isc_nrxd_max[i]); 5712 MPASS(powerof2(sctx->isc_nrxd_max[i])); 5713 MPASS(sctx->isc_nrxd_default[i]); 5714 MPASS(powerof2(sctx->isc_nrxd_default[i])); 5715 } 5716 5717 MPASS(sctx->isc_ntxqs >= 1 && sctx->isc_ntxqs <= 8); 5718 for (i = 0; i < sctx->isc_ntxqs; i++) { 5719 MPASS(sctx->isc_ntxd_min[i]); 5720 MPASS(powerof2(sctx->isc_ntxd_min[i])); 5721 MPASS(sctx->isc_ntxd_max[i]); 5722 MPASS(powerof2(sctx->isc_ntxd_max[i])); 5723 MPASS(sctx->isc_ntxd_default[i]); 5724 MPASS(powerof2(sctx->isc_ntxd_default[i])); 5725 } 5726 } 5727 5728 static void 5729 _iflib_pre_assert(if_softc_ctx_t scctx) 5730 { 5731 5732 MPASS(scctx->isc_txrx->ift_txd_encap); 5733 MPASS(scctx->isc_txrx->ift_txd_flush); 5734 MPASS(scctx->isc_txrx->ift_txd_credits_update); 5735 MPASS(scctx->isc_txrx->ift_rxd_available); 5736 MPASS(scctx->isc_txrx->ift_rxd_pkt_get); 5737 MPASS(scctx->isc_txrx->ift_rxd_refill); 5738 MPASS(scctx->isc_txrx->ift_rxd_flush); 5739 } 5740 5741 static int 5742 iflib_register(if_ctx_t ctx) 5743 { 5744 if_shared_ctx_t sctx = ctx->ifc_sctx; 5745 driver_t *driver = sctx->isc_driver; 5746 device_t dev = ctx->ifc_dev; 5747 if_t ifp; 5748 5749 _iflib_assert(sctx); 5750 5751 CTX_LOCK_INIT(ctx); 5752 STATE_LOCK_INIT(ctx, device_get_nameunit(ctx->ifc_dev)); 5753 ifp = ctx->ifc_ifp = if_alloc_dev(IFT_ETHER, dev); 5754 5755 /* 5756 * Initialize our context's device specific methods 5757 */ 5758 kobj_init((kobj_t) ctx, (kobj_class_t) driver); 5759 kobj_class_compile((kobj_class_t) driver); 5760 5761 if_initname(ifp, device_get_name(dev), device_get_unit(dev)); 5762 if_setsoftc(ifp, ctx); 5763 if_setdev(ifp, dev); 5764 if_setinitfn(ifp, iflib_if_init); 5765 if_setioctlfn(ifp, iflib_if_ioctl); 5766 #ifdef ALTQ 5767 if_setstartfn(ifp, iflib_altq_if_start); 5768 if_settransmitfn(ifp, iflib_altq_if_transmit); 5769 if_setsendqready(ifp); 5770 #else 5771 if_settransmitfn(ifp, iflib_if_transmit); 5772 #endif 5773 if_setqflushfn(ifp, iflib_if_qflush); 5774 if_setflags(ifp, IFF_BROADCAST | IFF_SIMPLEX | IFF_MULTICAST); 5775 ctx->ifc_vlan_attach_event = 5776 EVENTHANDLER_REGISTER(vlan_config, iflib_vlan_register, ctx, 5777 EVENTHANDLER_PRI_FIRST); 5778 ctx->ifc_vlan_detach_event = 5779 EVENTHANDLER_REGISTER(vlan_unconfig, iflib_vlan_unregister, ctx, 5780 EVENTHANDLER_PRI_FIRST); 5781 5782 if ((sctx->isc_flags & IFLIB_DRIVER_MEDIA) == 0) { 5783 ctx->ifc_mediap = &ctx->ifc_media; 5784 ifmedia_init(ctx->ifc_mediap, IFM_IMASK, 5785 iflib_media_change, iflib_media_status); 5786 } 5787 return (0); 5788 } 5789 5790 static void 5791 iflib_unregister_vlan_handlers(if_ctx_t ctx) 5792 { 5793 /* Unregister VLAN events */ 5794 if (ctx->ifc_vlan_attach_event != NULL) { 5795 EVENTHANDLER_DEREGISTER(vlan_config, ctx->ifc_vlan_attach_event); 5796 ctx->ifc_vlan_attach_event = NULL; 5797 } 5798 if (ctx->ifc_vlan_detach_event != NULL) { 5799 EVENTHANDLER_DEREGISTER(vlan_unconfig, ctx->ifc_vlan_detach_event); 5800 ctx->ifc_vlan_detach_event = NULL; 5801 } 5802 5803 } 5804 5805 static void 5806 iflib_deregister(if_ctx_t ctx) 5807 { 5808 if_t ifp = ctx->ifc_ifp; 5809 5810 /* Remove all media */ 5811 ifmedia_removeall(&ctx->ifc_media); 5812 5813 /* Ensure that VLAN event handlers are unregistered */ 5814 iflib_unregister_vlan_handlers(ctx); 5815 5816 /* Release kobject reference */ 5817 kobj_delete((kobj_t) ctx, NULL); 5818 5819 /* Free the ifnet structure */ 5820 if_free(ifp); 5821 5822 STATE_LOCK_DESTROY(ctx); 5823 5824 /* ether_ifdetach calls if_qflush - lock must be destroy afterwards*/ 5825 CTX_LOCK_DESTROY(ctx); 5826 } 5827 5828 static int 5829 iflib_queues_alloc(if_ctx_t ctx) 5830 { 5831 if_shared_ctx_t sctx = ctx->ifc_sctx; 5832 if_softc_ctx_t scctx = &ctx->ifc_softc_ctx; 5833 device_t dev = ctx->ifc_dev; 5834 int nrxqsets = scctx->isc_nrxqsets; 5835 int ntxqsets = scctx->isc_ntxqsets; 5836 iflib_txq_t txq; 5837 iflib_rxq_t rxq; 5838 iflib_fl_t fl = NULL; 5839 int i, j, cpu, err, txconf, rxconf; 5840 iflib_dma_info_t ifdip; 5841 uint32_t *rxqsizes = scctx->isc_rxqsizes; 5842 uint32_t *txqsizes = scctx->isc_txqsizes; 5843 uint8_t nrxqs = sctx->isc_nrxqs; 5844 uint8_t ntxqs = sctx->isc_ntxqs; 5845 int nfree_lists = sctx->isc_nfl ? sctx->isc_nfl : 1; 5846 int fl_offset = (sctx->isc_flags & IFLIB_HAS_RXCQ ? 1 : 0); 5847 caddr_t *vaddrs; 5848 uint64_t *paddrs; 5849 5850 KASSERT(ntxqs > 0, ("number of queues per qset must be at least 1")); 5851 KASSERT(nrxqs > 0, ("number of queues per qset must be at least 1")); 5852 KASSERT(nrxqs >= fl_offset + nfree_lists, 5853 ("there must be at least a rxq for each free list")); 5854 5855 /* Allocate the TX ring struct memory */ 5856 if (!(ctx->ifc_txqs = 5857 (iflib_txq_t) malloc(sizeof(struct iflib_txq) * 5858 ntxqsets, M_IFLIB, M_NOWAIT | M_ZERO))) { 5859 device_printf(dev, "Unable to allocate TX ring memory\n"); 5860 err = ENOMEM; 5861 goto fail; 5862 } 5863 5864 /* Now allocate the RX */ 5865 if (!(ctx->ifc_rxqs = 5866 (iflib_rxq_t) malloc(sizeof(struct iflib_rxq) * 5867 nrxqsets, M_IFLIB, M_NOWAIT | M_ZERO))) { 5868 device_printf(dev, "Unable to allocate RX ring memory\n"); 5869 err = ENOMEM; 5870 goto rx_fail; 5871 } 5872 5873 txq = ctx->ifc_txqs; 5874 rxq = ctx->ifc_rxqs; 5875 5876 /* 5877 * XXX handle allocation failure 5878 */ 5879 for (txconf = i = 0, cpu = CPU_FIRST(); i < ntxqsets; i++, txconf++, txq++, cpu = CPU_NEXT(cpu)) { 5880 /* Set up some basics */ 5881 5882 if ((ifdip = malloc(sizeof(struct iflib_dma_info) * ntxqs, 5883 M_IFLIB, M_NOWAIT | M_ZERO)) == NULL) { 5884 device_printf(dev, 5885 "Unable to allocate TX DMA info memory\n"); 5886 err = ENOMEM; 5887 goto err_tx_desc; 5888 } 5889 txq->ift_ifdi = ifdip; 5890 for (j = 0; j < ntxqs; j++, ifdip++) { 5891 if (iflib_dma_alloc(ctx, txqsizes[j], ifdip, 0)) { 5892 device_printf(dev, 5893 "Unable to allocate TX descriptors\n"); 5894 err = ENOMEM; 5895 goto err_tx_desc; 5896 } 5897 txq->ift_txd_size[j] = scctx->isc_txd_size[j]; 5898 bzero((void *)ifdip->idi_vaddr, txqsizes[j]); 5899 } 5900 txq->ift_ctx = ctx; 5901 txq->ift_id = i; 5902 if (sctx->isc_flags & IFLIB_HAS_TXCQ) { 5903 txq->ift_br_offset = 1; 5904 } else { 5905 txq->ift_br_offset = 0; 5906 } 5907 5908 if (iflib_txsd_alloc(txq)) { 5909 device_printf(dev, "Critical Failure setting up TX buffers\n"); 5910 err = ENOMEM; 5911 goto err_tx_desc; 5912 } 5913 5914 /* Initialize the TX lock */ 5915 snprintf(txq->ift_mtx_name, MTX_NAME_LEN, "%s:TX(%d):callout", 5916 device_get_nameunit(dev), txq->ift_id); 5917 mtx_init(&txq->ift_mtx, txq->ift_mtx_name, NULL, MTX_DEF); 5918 callout_init_mtx(&txq->ift_timer, &txq->ift_mtx, 0); 5919 txq->ift_timer.c_cpu = cpu; 5920 #ifdef DEV_NETMAP 5921 callout_init_mtx(&txq->ift_netmap_timer, &txq->ift_mtx, 0); 5922 txq->ift_netmap_timer.c_cpu = cpu; 5923 #endif /* DEV_NETMAP */ 5924 5925 err = ifmp_ring_alloc(&txq->ift_br, 2048, txq, iflib_txq_drain, 5926 iflib_txq_can_drain, M_IFLIB, M_WAITOK); 5927 if (err) { 5928 /* XXX free any allocated rings */ 5929 device_printf(dev, "Unable to allocate buf_ring\n"); 5930 goto err_tx_desc; 5931 } 5932 } 5933 5934 for (rxconf = i = 0; i < nrxqsets; i++, rxconf++, rxq++) { 5935 /* Set up some basics */ 5936 callout_init(&rxq->ifr_watchdog, 1); 5937 5938 if ((ifdip = malloc(sizeof(struct iflib_dma_info) * nrxqs, 5939 M_IFLIB, M_NOWAIT | M_ZERO)) == NULL) { 5940 device_printf(dev, 5941 "Unable to allocate RX DMA info memory\n"); 5942 err = ENOMEM; 5943 goto err_tx_desc; 5944 } 5945 5946 rxq->ifr_ifdi = ifdip; 5947 /* XXX this needs to be changed if #rx queues != #tx queues */ 5948 rxq->ifr_ntxqirq = 1; 5949 rxq->ifr_txqid[0] = i; 5950 for (j = 0; j < nrxqs; j++, ifdip++) { 5951 if (iflib_dma_alloc(ctx, rxqsizes[j], ifdip, 0)) { 5952 device_printf(dev, 5953 "Unable to allocate RX descriptors\n"); 5954 err = ENOMEM; 5955 goto err_tx_desc; 5956 } 5957 bzero((void *)ifdip->idi_vaddr, rxqsizes[j]); 5958 } 5959 rxq->ifr_ctx = ctx; 5960 rxq->ifr_id = i; 5961 rxq->ifr_fl_offset = fl_offset; 5962 rxq->ifr_nfl = nfree_lists; 5963 if (!(fl = 5964 (iflib_fl_t) malloc(sizeof(struct iflib_fl) * nfree_lists, M_IFLIB, M_NOWAIT | M_ZERO))) { 5965 device_printf(dev, "Unable to allocate free list memory\n"); 5966 err = ENOMEM; 5967 goto err_tx_desc; 5968 } 5969 rxq->ifr_fl = fl; 5970 for (j = 0; j < nfree_lists; j++) { 5971 fl[j].ifl_rxq = rxq; 5972 fl[j].ifl_id = j; 5973 fl[j].ifl_ifdi = &rxq->ifr_ifdi[j + rxq->ifr_fl_offset]; 5974 fl[j].ifl_rxd_size = scctx->isc_rxd_size[j]; 5975 } 5976 /* Allocate receive buffers for the ring */ 5977 if (iflib_rxsd_alloc(rxq)) { 5978 device_printf(dev, 5979 "Critical Failure setting up receive buffers\n"); 5980 err = ENOMEM; 5981 goto err_rx_desc; 5982 } 5983 5984 for (j = 0, fl = rxq->ifr_fl; j < rxq->ifr_nfl; j++, fl++) 5985 fl->ifl_rx_bitmap = bit_alloc(fl->ifl_size, M_IFLIB, 5986 M_WAITOK); 5987 } 5988 5989 /* TXQs */ 5990 vaddrs = malloc(sizeof(caddr_t) * ntxqsets * ntxqs, M_IFLIB, M_WAITOK); 5991 paddrs = malloc(sizeof(uint64_t) * ntxqsets * ntxqs, M_IFLIB, M_WAITOK); 5992 for (i = 0; i < ntxqsets; i++) { 5993 iflib_dma_info_t di = ctx->ifc_txqs[i].ift_ifdi; 5994 5995 for (j = 0; j < ntxqs; j++, di++) { 5996 vaddrs[i * ntxqs + j] = di->idi_vaddr; 5997 paddrs[i * ntxqs + j] = di->idi_paddr; 5998 } 5999 } 6000 if ((err = IFDI_TX_QUEUES_ALLOC(ctx, vaddrs, paddrs, ntxqs, ntxqsets)) != 0) { 6001 device_printf(ctx->ifc_dev, 6002 "Unable to allocate device TX queue\n"); 6003 iflib_tx_structures_free(ctx); 6004 free(vaddrs, M_IFLIB); 6005 free(paddrs, M_IFLIB); 6006 goto err_rx_desc; 6007 } 6008 free(vaddrs, M_IFLIB); 6009 free(paddrs, M_IFLIB); 6010 6011 /* RXQs */ 6012 vaddrs = malloc(sizeof(caddr_t) * nrxqsets * nrxqs, M_IFLIB, M_WAITOK); 6013 paddrs = malloc(sizeof(uint64_t) * nrxqsets * nrxqs, M_IFLIB, M_WAITOK); 6014 for (i = 0; i < nrxqsets; i++) { 6015 iflib_dma_info_t di = ctx->ifc_rxqs[i].ifr_ifdi; 6016 6017 for (j = 0; j < nrxqs; j++, di++) { 6018 vaddrs[i * nrxqs + j] = di->idi_vaddr; 6019 paddrs[i * nrxqs + j] = di->idi_paddr; 6020 } 6021 } 6022 if ((err = IFDI_RX_QUEUES_ALLOC(ctx, vaddrs, paddrs, nrxqs, nrxqsets)) != 0) { 6023 device_printf(ctx->ifc_dev, 6024 "Unable to allocate device RX queue\n"); 6025 iflib_tx_structures_free(ctx); 6026 free(vaddrs, M_IFLIB); 6027 free(paddrs, M_IFLIB); 6028 goto err_rx_desc; 6029 } 6030 free(vaddrs, M_IFLIB); 6031 free(paddrs, M_IFLIB); 6032 6033 return (0); 6034 6035 /* XXX handle allocation failure changes */ 6036 err_rx_desc: 6037 err_tx_desc: 6038 rx_fail: 6039 if (ctx->ifc_rxqs != NULL) 6040 free(ctx->ifc_rxqs, M_IFLIB); 6041 ctx->ifc_rxqs = NULL; 6042 if (ctx->ifc_txqs != NULL) 6043 free(ctx->ifc_txqs, M_IFLIB); 6044 ctx->ifc_txqs = NULL; 6045 fail: 6046 return (err); 6047 } 6048 6049 static int 6050 iflib_tx_structures_setup(if_ctx_t ctx) 6051 { 6052 iflib_txq_t txq = ctx->ifc_txqs; 6053 int i; 6054 6055 for (i = 0; i < NTXQSETS(ctx); i++, txq++) 6056 iflib_txq_setup(txq); 6057 6058 return (0); 6059 } 6060 6061 static void 6062 iflib_tx_structures_free(if_ctx_t ctx) 6063 { 6064 iflib_txq_t txq = ctx->ifc_txqs; 6065 if_shared_ctx_t sctx = ctx->ifc_sctx; 6066 int i, j; 6067 6068 for (i = 0; i < NTXQSETS(ctx); i++, txq++) { 6069 for (j = 0; j < sctx->isc_ntxqs; j++) 6070 iflib_dma_free(&txq->ift_ifdi[j]); 6071 iflib_txq_destroy(txq); 6072 } 6073 free(ctx->ifc_txqs, M_IFLIB); 6074 ctx->ifc_txqs = NULL; 6075 } 6076 6077 /********************************************************************* 6078 * 6079 * Initialize all receive rings. 6080 * 6081 **********************************************************************/ 6082 static int 6083 iflib_rx_structures_setup(if_ctx_t ctx) 6084 { 6085 iflib_rxq_t rxq = ctx->ifc_rxqs; 6086 int q; 6087 #if defined(INET6) || defined(INET) 6088 int err, i; 6089 #endif 6090 6091 for (q = 0; q < ctx->ifc_softc_ctx.isc_nrxqsets; q++, rxq++) { 6092 #if defined(INET6) || defined(INET) 6093 err = tcp_lro_init_args(&rxq->ifr_lc, ctx->ifc_ifp, 6094 TCP_LRO_ENTRIES, min(1024, 6095 ctx->ifc_softc_ctx.isc_nrxd[rxq->ifr_fl_offset])); 6096 if (err != 0) { 6097 device_printf(ctx->ifc_dev, 6098 "LRO Initialization failed!\n"); 6099 goto fail; 6100 } 6101 #endif 6102 IFDI_RXQ_SETUP(ctx, rxq->ifr_id); 6103 } 6104 return (0); 6105 #if defined(INET6) || defined(INET) 6106 fail: 6107 /* 6108 * Free LRO resources allocated so far, we will only handle 6109 * the rings that completed, the failing case will have 6110 * cleaned up for itself. 'q' failed, so its the terminus. 6111 */ 6112 rxq = ctx->ifc_rxqs; 6113 for (i = 0; i < q; ++i, rxq++) { 6114 tcp_lro_free(&rxq->ifr_lc); 6115 } 6116 return (err); 6117 #endif 6118 } 6119 6120 /********************************************************************* 6121 * 6122 * Free all receive rings. 6123 * 6124 **********************************************************************/ 6125 static void 6126 iflib_rx_structures_free(if_ctx_t ctx) 6127 { 6128 iflib_rxq_t rxq = ctx->ifc_rxqs; 6129 if_shared_ctx_t sctx = ctx->ifc_sctx; 6130 int i, j; 6131 6132 for (i = 0; i < ctx->ifc_softc_ctx.isc_nrxqsets; i++, rxq++) { 6133 for (j = 0; j < sctx->isc_nrxqs; j++) 6134 iflib_dma_free(&rxq->ifr_ifdi[j]); 6135 iflib_rx_sds_free(rxq); 6136 #if defined(INET6) || defined(INET) 6137 tcp_lro_free(&rxq->ifr_lc); 6138 #endif 6139 } 6140 free(ctx->ifc_rxqs, M_IFLIB); 6141 ctx->ifc_rxqs = NULL; 6142 } 6143 6144 static int 6145 iflib_qset_structures_setup(if_ctx_t ctx) 6146 { 6147 int err; 6148 6149 /* 6150 * It is expected that the caller takes care of freeing queues if this 6151 * fails. 6152 */ 6153 if ((err = iflib_tx_structures_setup(ctx)) != 0) { 6154 device_printf(ctx->ifc_dev, "iflib_tx_structures_setup failed: %d\n", err); 6155 return (err); 6156 } 6157 6158 if ((err = iflib_rx_structures_setup(ctx)) != 0) 6159 device_printf(ctx->ifc_dev, "iflib_rx_structures_setup failed: %d\n", err); 6160 6161 return (err); 6162 } 6163 6164 int 6165 iflib_irq_alloc(if_ctx_t ctx, if_irq_t irq, int rid, 6166 driver_filter_t filter, void *filter_arg, driver_intr_t handler, void *arg, const char *name) 6167 { 6168 6169 return (_iflib_irq_alloc(ctx, irq, rid, filter, handler, arg, name)); 6170 } 6171 6172 /* Just to avoid copy/paste */ 6173 static inline int 6174 iflib_irq_set_affinity(if_ctx_t ctx, if_irq_t irq, iflib_intr_type_t type, 6175 int qid, struct grouptask *gtask, struct taskqgroup *tqg, void *uniq, 6176 const char *name) 6177 { 6178 device_t dev; 6179 unsigned int base_cpuid, cpuid; 6180 int err; 6181 6182 dev = ctx->ifc_dev; 6183 base_cpuid = ctx->ifc_sysctl_core_offset; 6184 cpuid = get_cpuid_for_queue(ctx, base_cpuid, qid, type == IFLIB_INTR_TX); 6185 err = taskqgroup_attach_cpu(tqg, gtask, uniq, cpuid, dev, 6186 irq ? irq->ii_res : NULL, name); 6187 if (err) { 6188 device_printf(dev, "taskqgroup_attach_cpu failed %d\n", err); 6189 return (err); 6190 } 6191 #ifdef notyet 6192 if (cpuid > ctx->ifc_cpuid_highest) 6193 ctx->ifc_cpuid_highest = cpuid; 6194 #endif 6195 return (0); 6196 } 6197 6198 /* 6199 * Allocate a hardware interrupt for subctx using the parent (ctx)'s hardware 6200 * resources. 6201 * 6202 * Similar to iflib_irq_alloc_generic(), but for interrupt type IFLIB_INTR_RXTX 6203 * only. 6204 * 6205 * XXX: Could be removed if subctx's dev has its intr resource allocation 6206 * methods replaced with custom ones? 6207 */ 6208 int 6209 iflib_irq_alloc_generic_subctx(if_ctx_t ctx, if_ctx_t subctx, if_irq_t irq, 6210 int rid, iflib_intr_type_t type, 6211 driver_filter_t *filter, void *filter_arg, 6212 int qid, const char *name) 6213 { 6214 device_t dev, subdev; 6215 struct grouptask *gtask; 6216 struct taskqgroup *tqg; 6217 iflib_filter_info_t info; 6218 gtask_fn_t *fn; 6219 int tqrid, err; 6220 driver_filter_t *intr_fast; 6221 void *q; 6222 6223 MPASS(ctx != NULL); 6224 MPASS(subctx != NULL); 6225 6226 tqrid = rid; 6227 dev = ctx->ifc_dev; 6228 subdev = subctx->ifc_dev; 6229 6230 switch (type) { 6231 case IFLIB_INTR_RXTX: 6232 q = &subctx->ifc_rxqs[qid]; 6233 info = &subctx->ifc_rxqs[qid].ifr_filter_info; 6234 gtask = &subctx->ifc_rxqs[qid].ifr_task; 6235 tqg = qgroup_if_io_tqg; 6236 fn = _task_fn_rx; 6237 intr_fast = iflib_fast_intr_rxtx; 6238 NET_GROUPTASK_INIT(gtask, 0, fn, q); 6239 break; 6240 default: 6241 device_printf(dev, "%s: unknown net intr type for subctx %s (%d)\n", 6242 __func__, device_get_nameunit(subdev), type); 6243 return (EINVAL); 6244 } 6245 6246 info->ifi_filter = filter; 6247 info->ifi_filter_arg = filter_arg; 6248 info->ifi_task = gtask; 6249 info->ifi_ctx = q; 6250 6251 NET_GROUPTASK_INIT(gtask, 0, fn, q); 6252 6253 /* Allocate interrupts from hardware using parent context */ 6254 err = _iflib_irq_alloc(ctx, irq, rid, intr_fast, NULL, info, name); 6255 if (err != 0) { 6256 device_printf(dev, "_iflib_irq_alloc failed for subctx %s: %d\n", 6257 device_get_nameunit(subdev), err); 6258 return (err); 6259 } 6260 6261 if (tqrid != -1) { 6262 err = iflib_irq_set_affinity(ctx, irq, type, qid, gtask, tqg, q, 6263 name); 6264 if (err) 6265 return (err); 6266 } else { 6267 taskqgroup_attach(tqg, gtask, q, dev, irq->ii_res, name); 6268 } 6269 6270 return (0); 6271 } 6272 6273 int 6274 iflib_irq_alloc_generic(if_ctx_t ctx, if_irq_t irq, int rid, 6275 iflib_intr_type_t type, driver_filter_t *filter, 6276 void *filter_arg, int qid, const char *name) 6277 { 6278 device_t dev; 6279 struct grouptask *gtask; 6280 struct taskqgroup *tqg; 6281 iflib_filter_info_t info; 6282 gtask_fn_t *fn; 6283 int tqrid, err; 6284 driver_filter_t *intr_fast; 6285 void *q; 6286 6287 info = &ctx->ifc_filter_info; 6288 tqrid = rid; 6289 6290 switch (type) { 6291 /* XXX merge tx/rx for netmap? */ 6292 case IFLIB_INTR_TX: 6293 q = &ctx->ifc_txqs[qid]; 6294 info = &ctx->ifc_txqs[qid].ift_filter_info; 6295 gtask = &ctx->ifc_txqs[qid].ift_task; 6296 tqg = qgroup_if_io_tqg; 6297 fn = _task_fn_tx; 6298 intr_fast = iflib_fast_intr; 6299 GROUPTASK_INIT(gtask, 0, fn, q); 6300 ctx->ifc_flags |= IFC_NETMAP_TX_IRQ; 6301 break; 6302 case IFLIB_INTR_RX: 6303 q = &ctx->ifc_rxqs[qid]; 6304 info = &ctx->ifc_rxqs[qid].ifr_filter_info; 6305 gtask = &ctx->ifc_rxqs[qid].ifr_task; 6306 tqg = qgroup_if_io_tqg; 6307 fn = _task_fn_rx; 6308 intr_fast = iflib_fast_intr; 6309 NET_GROUPTASK_INIT(gtask, 0, fn, q); 6310 break; 6311 case IFLIB_INTR_RXTX: 6312 q = &ctx->ifc_rxqs[qid]; 6313 info = &ctx->ifc_rxqs[qid].ifr_filter_info; 6314 gtask = &ctx->ifc_rxqs[qid].ifr_task; 6315 tqg = qgroup_if_io_tqg; 6316 fn = _task_fn_rx; 6317 intr_fast = iflib_fast_intr_rxtx; 6318 NET_GROUPTASK_INIT(gtask, 0, fn, q); 6319 break; 6320 case IFLIB_INTR_ADMIN: 6321 q = ctx; 6322 tqrid = -1; 6323 info = &ctx->ifc_filter_info; 6324 gtask = &ctx->ifc_admin_task; 6325 tqg = qgroup_if_config_tqg; 6326 fn = _task_fn_admin; 6327 intr_fast = iflib_fast_intr_ctx; 6328 break; 6329 default: 6330 device_printf(ctx->ifc_dev, "%s: unknown net intr type\n", 6331 __func__); 6332 return (EINVAL); 6333 } 6334 6335 info->ifi_filter = filter; 6336 info->ifi_filter_arg = filter_arg; 6337 info->ifi_task = gtask; 6338 info->ifi_ctx = q; 6339 6340 dev = ctx->ifc_dev; 6341 err = _iflib_irq_alloc(ctx, irq, rid, intr_fast, NULL, info, name); 6342 if (err != 0) { 6343 device_printf(dev, "_iflib_irq_alloc failed %d\n", err); 6344 return (err); 6345 } 6346 if (type == IFLIB_INTR_ADMIN) 6347 return (0); 6348 6349 if (tqrid != -1) { 6350 err = iflib_irq_set_affinity(ctx, irq, type, qid, gtask, tqg, q, 6351 name); 6352 if (err) 6353 return (err); 6354 } else { 6355 taskqgroup_attach(tqg, gtask, q, dev, irq->ii_res, name); 6356 } 6357 6358 return (0); 6359 } 6360 6361 void 6362 iflib_softirq_alloc_generic(if_ctx_t ctx, if_irq_t irq, iflib_intr_type_t type, 6363 void *arg, int qid, const char *name) 6364 { 6365 device_t dev; 6366 struct grouptask *gtask; 6367 struct taskqgroup *tqg; 6368 gtask_fn_t *fn; 6369 void *q; 6370 int err; 6371 6372 switch (type) { 6373 case IFLIB_INTR_TX: 6374 q = &ctx->ifc_txqs[qid]; 6375 gtask = &ctx->ifc_txqs[qid].ift_task; 6376 tqg = qgroup_if_io_tqg; 6377 fn = _task_fn_tx; 6378 GROUPTASK_INIT(gtask, 0, fn, q); 6379 break; 6380 case IFLIB_INTR_RX: 6381 q = &ctx->ifc_rxqs[qid]; 6382 gtask = &ctx->ifc_rxqs[qid].ifr_task; 6383 tqg = qgroup_if_io_tqg; 6384 fn = _task_fn_rx; 6385 NET_GROUPTASK_INIT(gtask, 0, fn, q); 6386 break; 6387 case IFLIB_INTR_IOV: 6388 q = ctx; 6389 gtask = &ctx->ifc_vflr_task; 6390 tqg = qgroup_if_config_tqg; 6391 fn = _task_fn_iov; 6392 GROUPTASK_INIT(gtask, 0, fn, q); 6393 break; 6394 default: 6395 panic("unknown net intr type"); 6396 } 6397 err = iflib_irq_set_affinity(ctx, irq, type, qid, gtask, tqg, q, name); 6398 if (err) { 6399 dev = ctx->ifc_dev; 6400 taskqgroup_attach(tqg, gtask, q, dev, irq ? irq->ii_res : NULL, 6401 name); 6402 } 6403 } 6404 6405 void 6406 iflib_irq_free(if_ctx_t ctx, if_irq_t irq) 6407 { 6408 6409 if (irq->ii_tag) 6410 bus_teardown_intr(ctx->ifc_dev, irq->ii_res, irq->ii_tag); 6411 6412 if (irq->ii_res) 6413 bus_release_resource(ctx->ifc_dev, SYS_RES_IRQ, 6414 rman_get_rid(irq->ii_res), irq->ii_res); 6415 } 6416 6417 static int 6418 iflib_legacy_setup(if_ctx_t ctx, driver_filter_t filter, void *filter_arg, int *rid, const char *name) 6419 { 6420 iflib_txq_t txq = ctx->ifc_txqs; 6421 iflib_rxq_t rxq = ctx->ifc_rxqs; 6422 if_irq_t irq = &ctx->ifc_legacy_irq; 6423 iflib_filter_info_t info; 6424 device_t dev; 6425 struct grouptask *gtask; 6426 struct resource *res; 6427 int err, tqrid; 6428 bool rx_only; 6429 6430 info = &rxq->ifr_filter_info; 6431 gtask = &rxq->ifr_task; 6432 tqrid = *rid; 6433 rx_only = (ctx->ifc_sctx->isc_flags & IFLIB_SINGLE_IRQ_RX_ONLY) != 0; 6434 6435 ctx->ifc_flags |= IFC_LEGACY; 6436 info->ifi_filter = filter; 6437 info->ifi_filter_arg = filter_arg; 6438 info->ifi_task = gtask; 6439 info->ifi_ctx = rxq; 6440 6441 dev = ctx->ifc_dev; 6442 /* We allocate a single interrupt resource */ 6443 err = _iflib_irq_alloc(ctx, irq, tqrid, rx_only ? iflib_fast_intr : 6444 iflib_fast_intr_rxtx, NULL, info, name); 6445 if (err != 0) 6446 return (err); 6447 NET_GROUPTASK_INIT(gtask, 0, _task_fn_rx, rxq); 6448 res = irq->ii_res; 6449 taskqgroup_attach(qgroup_if_io_tqg, gtask, rxq, dev, res, name); 6450 6451 GROUPTASK_INIT(&txq->ift_task, 0, _task_fn_tx, txq); 6452 taskqgroup_attach(qgroup_if_io_tqg, &txq->ift_task, txq, dev, res, 6453 "tx"); 6454 return (0); 6455 } 6456 6457 void 6458 iflib_led_create(if_ctx_t ctx) 6459 { 6460 6461 ctx->ifc_led_dev = led_create(iflib_led_func, ctx, 6462 device_get_nameunit(ctx->ifc_dev)); 6463 } 6464 6465 void 6466 iflib_tx_intr_deferred(if_ctx_t ctx, int txqid) 6467 { 6468 6469 GROUPTASK_ENQUEUE(&ctx->ifc_txqs[txqid].ift_task); 6470 } 6471 6472 void 6473 iflib_rx_intr_deferred(if_ctx_t ctx, int rxqid) 6474 { 6475 6476 GROUPTASK_ENQUEUE(&ctx->ifc_rxqs[rxqid].ifr_task); 6477 } 6478 6479 void 6480 iflib_admin_intr_deferred(if_ctx_t ctx) 6481 { 6482 6483 MPASS(ctx->ifc_admin_task.gt_taskqueue != NULL); 6484 GROUPTASK_ENQUEUE(&ctx->ifc_admin_task); 6485 } 6486 6487 void 6488 iflib_iov_intr_deferred(if_ctx_t ctx) 6489 { 6490 6491 GROUPTASK_ENQUEUE(&ctx->ifc_vflr_task); 6492 } 6493 6494 void 6495 iflib_io_tqg_attach(struct grouptask *gt, void *uniq, int cpu, const char *name) 6496 { 6497 6498 taskqgroup_attach_cpu(qgroup_if_io_tqg, gt, uniq, cpu, NULL, NULL, 6499 name); 6500 } 6501 6502 void 6503 iflib_config_gtask_init(void *ctx, struct grouptask *gtask, gtask_fn_t *fn, 6504 const char *name) 6505 { 6506 6507 GROUPTASK_INIT(gtask, 0, fn, ctx); 6508 taskqgroup_attach(qgroup_if_config_tqg, gtask, gtask, NULL, NULL, 6509 name); 6510 } 6511 6512 void 6513 iflib_config_gtask_deinit(struct grouptask *gtask) 6514 { 6515 6516 taskqgroup_detach(qgroup_if_config_tqg, gtask); 6517 } 6518 6519 void 6520 iflib_link_state_change(if_ctx_t ctx, int link_state, uint64_t baudrate) 6521 { 6522 if_t ifp = ctx->ifc_ifp; 6523 iflib_txq_t txq = ctx->ifc_txqs; 6524 6525 if_setbaudrate(ifp, baudrate); 6526 if (baudrate >= IF_Gbps(10)) { 6527 STATE_LOCK(ctx); 6528 ctx->ifc_flags |= IFC_PREFETCH; 6529 STATE_UNLOCK(ctx); 6530 } 6531 /* If link down, disable watchdog */ 6532 if ((ctx->ifc_link_state == LINK_STATE_UP) && (link_state == LINK_STATE_DOWN)) { 6533 for (int i = 0; i < ctx->ifc_softc_ctx.isc_ntxqsets; i++, txq++) 6534 txq->ift_qstatus = IFLIB_QUEUE_IDLE; 6535 } 6536 ctx->ifc_link_state = link_state; 6537 if_link_state_change(ifp, link_state); 6538 } 6539 6540 static int 6541 iflib_tx_credits_update(if_ctx_t ctx, iflib_txq_t txq) 6542 { 6543 int credits; 6544 #ifdef INVARIANTS 6545 int credits_pre = txq->ift_cidx_processed; 6546 #endif 6547 6548 bus_dmamap_sync(txq->ift_ifdi->idi_tag, txq->ift_ifdi->idi_map, 6549 BUS_DMASYNC_POSTREAD); 6550 if ((credits = ctx->isc_txd_credits_update(ctx->ifc_softc, txq->ift_id, true)) == 0) 6551 return (0); 6552 6553 txq->ift_processed += credits; 6554 txq->ift_cidx_processed += credits; 6555 6556 MPASS(credits_pre + credits == txq->ift_cidx_processed); 6557 if (txq->ift_cidx_processed >= txq->ift_size) 6558 txq->ift_cidx_processed -= txq->ift_size; 6559 return (credits); 6560 } 6561 6562 static int 6563 iflib_rxd_avail(if_ctx_t ctx, iflib_rxq_t rxq, qidx_t cidx, qidx_t budget) 6564 { 6565 iflib_fl_t fl; 6566 u_int i; 6567 6568 for (i = 0, fl = &rxq->ifr_fl[0]; i < rxq->ifr_nfl; i++, fl++) 6569 bus_dmamap_sync(fl->ifl_ifdi->idi_tag, fl->ifl_ifdi->idi_map, 6570 BUS_DMASYNC_POSTREAD | BUS_DMASYNC_POSTWRITE); 6571 return (ctx->isc_rxd_available(ctx->ifc_softc, rxq->ifr_id, cidx, 6572 budget)); 6573 } 6574 6575 void 6576 iflib_add_int_delay_sysctl(if_ctx_t ctx, const char *name, 6577 const char *description, if_int_delay_info_t info, 6578 int offset, int value) 6579 { 6580 info->iidi_ctx = ctx; 6581 info->iidi_offset = offset; 6582 info->iidi_value = value; 6583 SYSCTL_ADD_PROC(device_get_sysctl_ctx(ctx->ifc_dev), 6584 SYSCTL_CHILDREN(device_get_sysctl_tree(ctx->ifc_dev)), 6585 OID_AUTO, name, CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE, 6586 info, 0, iflib_sysctl_int_delay, "I", description); 6587 } 6588 6589 struct sx * 6590 iflib_ctx_lock_get(if_ctx_t ctx) 6591 { 6592 6593 return (&ctx->ifc_ctx_sx); 6594 } 6595 6596 static int 6597 iflib_msix_init(if_ctx_t ctx) 6598 { 6599 device_t dev = ctx->ifc_dev; 6600 if_shared_ctx_t sctx = ctx->ifc_sctx; 6601 if_softc_ctx_t scctx = &ctx->ifc_softc_ctx; 6602 int admincnt, bar, err, iflib_num_rx_queues, iflib_num_tx_queues; 6603 int msgs, queuemsgs, queues, rx_queues, tx_queues, vectors; 6604 6605 iflib_num_tx_queues = ctx->ifc_sysctl_ntxqs; 6606 iflib_num_rx_queues = ctx->ifc_sysctl_nrxqs; 6607 6608 if (bootverbose) 6609 device_printf(dev, "msix_init qsets capped at %d\n", 6610 imax(scctx->isc_ntxqsets, scctx->isc_nrxqsets)); 6611 6612 /* Override by tuneable */ 6613 if (scctx->isc_disable_msix) 6614 goto msi; 6615 6616 /* First try MSI-X */ 6617 if ((msgs = pci_msix_count(dev)) == 0) { 6618 if (bootverbose) 6619 device_printf(dev, "MSI-X not supported or disabled\n"); 6620 goto msi; 6621 } 6622 6623 bar = ctx->ifc_softc_ctx.isc_msix_bar; 6624 /* 6625 * bar == -1 => "trust me I know what I'm doing" 6626 * Some drivers are for hardware that is so shoddily 6627 * documented that no one knows which bars are which 6628 * so the developer has to map all bars. This hack 6629 * allows shoddy garbage to use MSI-X in this framework. 6630 */ 6631 if (bar != -1) { 6632 ctx->ifc_msix_mem = bus_alloc_resource_any(dev, 6633 SYS_RES_MEMORY, &bar, RF_ACTIVE); 6634 if (ctx->ifc_msix_mem == NULL) { 6635 device_printf(dev, "Unable to map MSI-X table\n"); 6636 goto msi; 6637 } 6638 } 6639 6640 admincnt = sctx->isc_admin_intrcnt; 6641 #if IFLIB_DEBUG 6642 /* use only 1 qset in debug mode */ 6643 queuemsgs = min(msgs - admincnt, 1); 6644 #else 6645 queuemsgs = msgs - admincnt; 6646 #endif 6647 #ifdef RSS 6648 queues = imin(queuemsgs, rss_getnumbuckets()); 6649 #else 6650 queues = queuemsgs; 6651 #endif 6652 queues = imin(CPU_COUNT(&ctx->ifc_cpus), queues); 6653 if (bootverbose) 6654 device_printf(dev, 6655 "intr CPUs: %d queue msgs: %d admincnt: %d\n", 6656 CPU_COUNT(&ctx->ifc_cpus), queuemsgs, admincnt); 6657 #ifdef RSS 6658 /* If we're doing RSS, clamp at the number of RSS buckets */ 6659 if (queues > rss_getnumbuckets()) 6660 queues = rss_getnumbuckets(); 6661 #endif 6662 if (iflib_num_rx_queues > 0 && iflib_num_rx_queues < queuemsgs - admincnt) 6663 rx_queues = iflib_num_rx_queues; 6664 else 6665 rx_queues = queues; 6666 6667 if (rx_queues > scctx->isc_nrxqsets) 6668 rx_queues = scctx->isc_nrxqsets; 6669 6670 /* 6671 * We want this to be all logical CPUs by default 6672 */ 6673 if (iflib_num_tx_queues > 0 && iflib_num_tx_queues < queues) 6674 tx_queues = iflib_num_tx_queues; 6675 else 6676 tx_queues = mp_ncpus; 6677 6678 if (tx_queues > scctx->isc_ntxqsets) 6679 tx_queues = scctx->isc_ntxqsets; 6680 6681 if (ctx->ifc_sysctl_qs_eq_override == 0) { 6682 #ifdef INVARIANTS 6683 if (tx_queues != rx_queues) 6684 device_printf(dev, 6685 "queue equality override not set, capping rx_queues at %d and tx_queues at %d\n", 6686 min(rx_queues, tx_queues), min(rx_queues, tx_queues)); 6687 #endif 6688 tx_queues = min(rx_queues, tx_queues); 6689 rx_queues = min(rx_queues, tx_queues); 6690 } 6691 6692 vectors = rx_queues + admincnt; 6693 if (msgs < vectors) { 6694 device_printf(dev, 6695 "insufficient number of MSI-X vectors " 6696 "(supported %d, need %d)\n", msgs, vectors); 6697 goto msi; 6698 } 6699 6700 device_printf(dev, "Using %d RX queues %d TX queues\n", rx_queues, 6701 tx_queues); 6702 msgs = vectors; 6703 if ((err = pci_alloc_msix(dev, &vectors)) == 0) { 6704 if (vectors != msgs) { 6705 device_printf(dev, 6706 "Unable to allocate sufficient MSI-X vectors " 6707 "(got %d, need %d)\n", vectors, msgs); 6708 pci_release_msi(dev); 6709 if (bar != -1) { 6710 bus_release_resource(dev, SYS_RES_MEMORY, bar, 6711 ctx->ifc_msix_mem); 6712 ctx->ifc_msix_mem = NULL; 6713 } 6714 goto msi; 6715 } 6716 device_printf(dev, "Using MSI-X interrupts with %d vectors\n", 6717 vectors); 6718 scctx->isc_vectors = vectors; 6719 scctx->isc_nrxqsets = rx_queues; 6720 scctx->isc_ntxqsets = tx_queues; 6721 scctx->isc_intr = IFLIB_INTR_MSIX; 6722 6723 return (vectors); 6724 } else { 6725 device_printf(dev, 6726 "failed to allocate %d MSI-X vectors, err: %d\n", vectors, 6727 err); 6728 if (bar != -1) { 6729 bus_release_resource(dev, SYS_RES_MEMORY, bar, 6730 ctx->ifc_msix_mem); 6731 ctx->ifc_msix_mem = NULL; 6732 } 6733 } 6734 6735 msi: 6736 vectors = pci_msi_count(dev); 6737 scctx->isc_nrxqsets = 1; 6738 scctx->isc_ntxqsets = 1; 6739 scctx->isc_vectors = vectors; 6740 if (vectors == 1 && pci_alloc_msi(dev, &vectors) == 0) { 6741 device_printf(dev, "Using an MSI interrupt\n"); 6742 scctx->isc_intr = IFLIB_INTR_MSI; 6743 } else { 6744 scctx->isc_vectors = 1; 6745 device_printf(dev, "Using a Legacy interrupt\n"); 6746 scctx->isc_intr = IFLIB_INTR_LEGACY; 6747 } 6748 6749 return (vectors); 6750 } 6751 6752 static const char *ring_states[] = { "IDLE", "BUSY", "STALLED", "ABDICATED" }; 6753 6754 static int 6755 mp_ring_state_handler(SYSCTL_HANDLER_ARGS) 6756 { 6757 int rc; 6758 uint16_t *state = ((uint16_t *)oidp->oid_arg1); 6759 struct sbuf *sb; 6760 const char *ring_state = "UNKNOWN"; 6761 6762 /* XXX needed ? */ 6763 rc = sysctl_wire_old_buffer(req, 0); 6764 MPASS(rc == 0); 6765 if (rc != 0) 6766 return (rc); 6767 sb = sbuf_new_for_sysctl(NULL, NULL, 80, req); 6768 MPASS(sb != NULL); 6769 if (sb == NULL) 6770 return (ENOMEM); 6771 if (state[3] <= 3) 6772 ring_state = ring_states[state[3]]; 6773 6774 sbuf_printf(sb, "pidx_head: %04hd pidx_tail: %04hd cidx: %04hd state: %s", 6775 state[0], state[1], state[2], ring_state); 6776 rc = sbuf_finish(sb); 6777 sbuf_delete(sb); 6778 return (rc); 6779 } 6780 6781 enum iflib_ndesc_handler { 6782 IFLIB_NTXD_HANDLER, 6783 IFLIB_NRXD_HANDLER, 6784 }; 6785 6786 static int 6787 mp_ndesc_handler(SYSCTL_HANDLER_ARGS) 6788 { 6789 if_ctx_t ctx = (void *)arg1; 6790 enum iflib_ndesc_handler type = arg2; 6791 char buf[256] = {0}; 6792 qidx_t *ndesc; 6793 char *p, *next; 6794 int nqs, rc, i; 6795 6796 nqs = 8; 6797 switch (type) { 6798 case IFLIB_NTXD_HANDLER: 6799 ndesc = ctx->ifc_sysctl_ntxds; 6800 if (ctx->ifc_sctx) 6801 nqs = ctx->ifc_sctx->isc_ntxqs; 6802 break; 6803 case IFLIB_NRXD_HANDLER: 6804 ndesc = ctx->ifc_sysctl_nrxds; 6805 if (ctx->ifc_sctx) 6806 nqs = ctx->ifc_sctx->isc_nrxqs; 6807 break; 6808 default: 6809 printf("%s: unhandled type\n", __func__); 6810 return (EINVAL); 6811 } 6812 if (nqs == 0) 6813 nqs = 8; 6814 6815 for (i = 0; i < 8; i++) { 6816 if (i >= nqs) 6817 break; 6818 if (i) 6819 strcat(buf, ","); 6820 sprintf(strchr(buf, 0), "%d", ndesc[i]); 6821 } 6822 6823 rc = sysctl_handle_string(oidp, buf, sizeof(buf), req); 6824 if (rc || req->newptr == NULL) 6825 return (rc); 6826 6827 for (i = 0, next = buf, p = strsep(&next, " ,"); i < 8 && p; 6828 i++, p = strsep(&next, " ,")) { 6829 ndesc[i] = strtoul(p, NULL, 10); 6830 } 6831 6832 return (rc); 6833 } 6834 6835 #define NAME_BUFLEN 32 6836 static void 6837 iflib_add_device_sysctl_pre(if_ctx_t ctx) 6838 { 6839 device_t dev = iflib_get_dev(ctx); 6840 struct sysctl_oid_list *child, *oid_list; 6841 struct sysctl_ctx_list *ctx_list; 6842 struct sysctl_oid *node; 6843 6844 ctx_list = device_get_sysctl_ctx(dev); 6845 child = SYSCTL_CHILDREN(device_get_sysctl_tree(dev)); 6846 ctx->ifc_sysctl_node = node = SYSCTL_ADD_NODE(ctx_list, child, 6847 OID_AUTO, "iflib", CTLFLAG_RD | CTLFLAG_MPSAFE, NULL, 6848 "IFLIB fields"); 6849 oid_list = SYSCTL_CHILDREN(node); 6850 6851 SYSCTL_ADD_CONST_STRING(ctx_list, oid_list, OID_AUTO, "driver_version", 6852 CTLFLAG_RD, ctx->ifc_sctx->isc_driver_version, "driver version"); 6853 6854 SYSCTL_ADD_U16(ctx_list, oid_list, OID_AUTO, "override_ntxqs", 6855 CTLFLAG_RWTUN, &ctx->ifc_sysctl_ntxqs, 0, 6856 "# of txqs to use, 0 => use default #"); 6857 SYSCTL_ADD_U16(ctx_list, oid_list, OID_AUTO, "override_nrxqs", 6858 CTLFLAG_RWTUN, &ctx->ifc_sysctl_nrxqs, 0, 6859 "# of rxqs to use, 0 => use default #"); 6860 SYSCTL_ADD_U16(ctx_list, oid_list, OID_AUTO, "override_qs_enable", 6861 CTLFLAG_RWTUN, &ctx->ifc_sysctl_qs_eq_override, 0, 6862 "permit #txq != #rxq"); 6863 SYSCTL_ADD_INT(ctx_list, oid_list, OID_AUTO, "disable_msix", 6864 CTLFLAG_RWTUN, &ctx->ifc_softc_ctx.isc_disable_msix, 0, 6865 "disable MSI-X (default 0)"); 6866 SYSCTL_ADD_U16(ctx_list, oid_list, OID_AUTO, "rx_budget", 6867 CTLFLAG_RWTUN, &ctx->ifc_sysctl_rx_budget, 0, "set the RX budget"); 6868 SYSCTL_ADD_U16(ctx_list, oid_list, OID_AUTO, "tx_abdicate", 6869 CTLFLAG_RWTUN, &ctx->ifc_sysctl_tx_abdicate, 0, 6870 "cause TX to abdicate instead of running to completion"); 6871 ctx->ifc_sysctl_core_offset = CORE_OFFSET_UNSPECIFIED; 6872 SYSCTL_ADD_U16(ctx_list, oid_list, OID_AUTO, "core_offset", 6873 CTLFLAG_RDTUN, &ctx->ifc_sysctl_core_offset, 0, 6874 "offset to start using cores at"); 6875 SYSCTL_ADD_U8(ctx_list, oid_list, OID_AUTO, "separate_txrx", 6876 CTLFLAG_RDTUN, &ctx->ifc_sysctl_separate_txrx, 0, 6877 "use separate cores for TX and RX"); 6878 SYSCTL_ADD_U8(ctx_list, oid_list, OID_AUTO, "use_logical_cores", 6879 CTLFLAG_RDTUN, &ctx->ifc_sysctl_use_logical_cores, 0, 6880 "try to make use of logical cores for TX and RX"); 6881 SYSCTL_ADD_U16(ctx_list, oid_list, OID_AUTO, "use_extra_msix_vectors", 6882 CTLFLAG_RDTUN, &ctx->ifc_sysctl_extra_msix_vectors, 0, 6883 "attempt to reserve the given number of extra MSI-X vectors during driver load for the creation of additional interfaces later"); 6884 SYSCTL_ADD_INT(ctx_list, oid_list, OID_AUTO, "allocated_msix_vectors", 6885 CTLFLAG_RDTUN, &ctx->ifc_softc_ctx.isc_vectors, 0, 6886 "total # of MSI-X vectors allocated by driver"); 6887 6888 /* XXX change for per-queue sizes */ 6889 SYSCTL_ADD_PROC(ctx_list, oid_list, OID_AUTO, "override_ntxds", 6890 CTLTYPE_STRING | CTLFLAG_RWTUN | CTLFLAG_NEEDGIANT, ctx, 6891 IFLIB_NTXD_HANDLER, mp_ndesc_handler, "A", 6892 "list of # of TX descriptors to use, 0 = use default #"); 6893 SYSCTL_ADD_PROC(ctx_list, oid_list, OID_AUTO, "override_nrxds", 6894 CTLTYPE_STRING | CTLFLAG_RWTUN | CTLFLAG_NEEDGIANT, ctx, 6895 IFLIB_NRXD_HANDLER, mp_ndesc_handler, "A", 6896 "list of # of RX descriptors to use, 0 = use default #"); 6897 } 6898 6899 static void 6900 iflib_add_device_sysctl_post(if_ctx_t ctx) 6901 { 6902 if_shared_ctx_t sctx = ctx->ifc_sctx; 6903 if_softc_ctx_t scctx = &ctx->ifc_softc_ctx; 6904 device_t dev = iflib_get_dev(ctx); 6905 struct sysctl_oid_list *child; 6906 struct sysctl_ctx_list *ctx_list; 6907 iflib_fl_t fl; 6908 iflib_txq_t txq; 6909 iflib_rxq_t rxq; 6910 int i, j; 6911 char namebuf[NAME_BUFLEN]; 6912 char *qfmt; 6913 struct sysctl_oid *queue_node, *fl_node, *node; 6914 struct sysctl_oid_list *queue_list, *fl_list; 6915 ctx_list = device_get_sysctl_ctx(dev); 6916 6917 node = ctx->ifc_sysctl_node; 6918 child = SYSCTL_CHILDREN(node); 6919 6920 if (scctx->isc_ntxqsets > 100) 6921 qfmt = "txq%03d"; 6922 else if (scctx->isc_ntxqsets > 10) 6923 qfmt = "txq%02d"; 6924 else 6925 qfmt = "txq%d"; 6926 for (i = 0, txq = ctx->ifc_txqs; i < scctx->isc_ntxqsets; i++, txq++) { 6927 snprintf(namebuf, NAME_BUFLEN, qfmt, i); 6928 queue_node = SYSCTL_ADD_NODE(ctx_list, child, OID_AUTO, namebuf, 6929 CTLFLAG_RD | CTLFLAG_MPSAFE, NULL, "Queue Name"); 6930 queue_list = SYSCTL_CHILDREN(queue_node); 6931 SYSCTL_ADD_INT(ctx_list, queue_list, OID_AUTO, "cpu", 6932 CTLFLAG_RD, &txq->ift_task.gt_cpu, 0, 6933 "cpu this queue is bound to"); 6934 #if MEMORY_LOGGING 6935 SYSCTL_ADD_UQUAD(ctx_list, queue_list, OID_AUTO, "txq_dequeued", 6936 CTLFLAG_RD, &txq->ift_dequeued, "total mbufs freed"); 6937 SYSCTL_ADD_UQUAD(ctx_list, queue_list, OID_AUTO, "txq_enqueued", 6938 CTLFLAG_RD, &txq->ift_enqueued, "total mbufs enqueued"); 6939 #endif 6940 SYSCTL_ADD_UQUAD(ctx_list, queue_list, OID_AUTO, "mbuf_defrag", 6941 CTLFLAG_RD, &txq->ift_mbuf_defrag, 6942 "# of times m_defrag was called"); 6943 SYSCTL_ADD_UQUAD(ctx_list, queue_list, OID_AUTO, "m_pullups", 6944 CTLFLAG_RD, &txq->ift_pullups, 6945 "# of times m_pullup was called"); 6946 SYSCTL_ADD_UQUAD(ctx_list, queue_list, OID_AUTO, 6947 "mbuf_defrag_failed", CTLFLAG_RD, 6948 &txq->ift_mbuf_defrag_failed, "# of times m_defrag failed"); 6949 SYSCTL_ADD_UQUAD(ctx_list, queue_list, OID_AUTO, 6950 "no_desc_avail", CTLFLAG_RD, &txq->ift_no_desc_avail, 6951 "# of times no descriptors were available"); 6952 SYSCTL_ADD_UQUAD(ctx_list, queue_list, OID_AUTO, 6953 "tx_map_failed", CTLFLAG_RD, &txq->ift_map_failed, 6954 "# of times DMA map failed"); 6955 SYSCTL_ADD_UQUAD(ctx_list, queue_list, OID_AUTO, 6956 "txd_encap_efbig", CTLFLAG_RD, &txq->ift_txd_encap_efbig, 6957 "# of times txd_encap returned EFBIG"); 6958 SYSCTL_ADD_UQUAD(ctx_list, queue_list, OID_AUTO, 6959 "no_tx_dma_setup", CTLFLAG_RD, &txq->ift_no_tx_dma_setup, 6960 "# of times map failed for other than EFBIG"); 6961 SYSCTL_ADD_U16(ctx_list, queue_list, OID_AUTO, "txq_pidx", 6962 CTLFLAG_RD, &txq->ift_pidx, 1, "Producer Index"); 6963 SYSCTL_ADD_U16(ctx_list, queue_list, OID_AUTO, "txq_cidx", 6964 CTLFLAG_RD, &txq->ift_cidx, 1, "Consumer Index"); 6965 SYSCTL_ADD_U16(ctx_list, queue_list, OID_AUTO, 6966 "txq_cidx_processed", CTLFLAG_RD, &txq->ift_cidx_processed, 6967 1, "Consumer Index seen by credit update"); 6968 SYSCTL_ADD_U16(ctx_list, queue_list, OID_AUTO, "txq_in_use", 6969 CTLFLAG_RD, &txq->ift_in_use, 1, "descriptors in use"); 6970 SYSCTL_ADD_UQUAD(ctx_list, queue_list, OID_AUTO, 6971 "txq_processed", CTLFLAG_RD, &txq->ift_processed, 6972 "descriptors procesed for clean"); 6973 SYSCTL_ADD_UQUAD(ctx_list, queue_list, OID_AUTO, "txq_cleaned", 6974 CTLFLAG_RD, &txq->ift_cleaned, "total cleaned"); 6975 SYSCTL_ADD_PROC(ctx_list, queue_list, OID_AUTO, "ring_state", 6976 CTLTYPE_STRING | CTLFLAG_RD | CTLFLAG_NEEDGIANT, 6977 __DEVOLATILE(uint64_t *, &txq->ift_br->state), 0, 6978 mp_ring_state_handler, "A", "soft ring state"); 6979 SYSCTL_ADD_COUNTER_U64(ctx_list, queue_list, OID_AUTO, 6980 "r_enqueues", CTLFLAG_RD, &txq->ift_br->enqueues, 6981 "# of enqueues to the mp_ring for this queue"); 6982 SYSCTL_ADD_COUNTER_U64(ctx_list, queue_list, OID_AUTO, 6983 "r_drops", CTLFLAG_RD, &txq->ift_br->drops, 6984 "# of drops in the mp_ring for this queue"); 6985 SYSCTL_ADD_COUNTER_U64(ctx_list, queue_list, OID_AUTO, 6986 "r_starts", CTLFLAG_RD, &txq->ift_br->starts, 6987 "# of normal consumer starts in mp_ring for this queue"); 6988 SYSCTL_ADD_COUNTER_U64(ctx_list, queue_list, OID_AUTO, 6989 "r_stalls", CTLFLAG_RD, &txq->ift_br->stalls, 6990 "# of consumer stalls in the mp_ring for this queue"); 6991 SYSCTL_ADD_COUNTER_U64(ctx_list, queue_list, OID_AUTO, 6992 "r_restarts", CTLFLAG_RD, &txq->ift_br->restarts, 6993 "# of consumer restarts in the mp_ring for this queue"); 6994 SYSCTL_ADD_COUNTER_U64(ctx_list, queue_list, OID_AUTO, 6995 "r_abdications", CTLFLAG_RD, &txq->ift_br->abdications, 6996 "# of consumer abdications in the mp_ring for this queue"); 6997 } 6998 6999 if (scctx->isc_nrxqsets > 100) 7000 qfmt = "rxq%03d"; 7001 else if (scctx->isc_nrxqsets > 10) 7002 qfmt = "rxq%02d"; 7003 else 7004 qfmt = "rxq%d"; 7005 for (i = 0, rxq = ctx->ifc_rxqs; i < scctx->isc_nrxqsets; i++, rxq++) { 7006 snprintf(namebuf, NAME_BUFLEN, qfmt, i); 7007 queue_node = SYSCTL_ADD_NODE(ctx_list, child, OID_AUTO, namebuf, 7008 CTLFLAG_RD | CTLFLAG_MPSAFE, NULL, "Queue Name"); 7009 queue_list = SYSCTL_CHILDREN(queue_node); 7010 SYSCTL_ADD_INT(ctx_list, queue_list, OID_AUTO, "cpu", 7011 CTLFLAG_RD, &rxq->ifr_task.gt_cpu, 0, 7012 "cpu this queue is bound to"); 7013 if (sctx->isc_flags & IFLIB_HAS_RXCQ) { 7014 SYSCTL_ADD_U16(ctx_list, queue_list, OID_AUTO, 7015 "rxq_cq_cidx", CTLFLAG_RD, &rxq->ifr_cq_cidx, 1, 7016 "Consumer Index"); 7017 } 7018 7019 for (j = 0, fl = rxq->ifr_fl; j < rxq->ifr_nfl; j++, fl++) { 7020 snprintf(namebuf, NAME_BUFLEN, "rxq_fl%d", j); 7021 fl_node = SYSCTL_ADD_NODE(ctx_list, queue_list, 7022 OID_AUTO, namebuf, CTLFLAG_RD | CTLFLAG_MPSAFE, 7023 NULL, "freelist Name"); 7024 fl_list = SYSCTL_CHILDREN(fl_node); 7025 SYSCTL_ADD_U16(ctx_list, fl_list, OID_AUTO, "pidx", 7026 CTLFLAG_RD, &fl->ifl_pidx, 1, "Producer Index"); 7027 SYSCTL_ADD_U16(ctx_list, fl_list, OID_AUTO, "cidx", 7028 CTLFLAG_RD, &fl->ifl_cidx, 1, "Consumer Index"); 7029 SYSCTL_ADD_U16(ctx_list, fl_list, OID_AUTO, "credits", 7030 CTLFLAG_RD, &fl->ifl_credits, 1, 7031 "credits available"); 7032 SYSCTL_ADD_U16(ctx_list, fl_list, OID_AUTO, "buf_size", 7033 CTLFLAG_RD, &fl->ifl_buf_size, 1, "buffer size"); 7034 #if MEMORY_LOGGING 7035 SYSCTL_ADD_UQUAD(ctx_list, fl_list, OID_AUTO, 7036 "fl_m_enqueued", CTLFLAG_RD, &fl->ifl_m_enqueued, 7037 "mbufs allocated"); 7038 SYSCTL_ADD_UQUAD(ctx_list, fl_list, OID_AUTO, 7039 "fl_m_dequeued", CTLFLAG_RD, &fl->ifl_m_dequeued, 7040 "mbufs freed"); 7041 SYSCTL_ADD_UQUAD(ctx_list, fl_list, OID_AUTO, 7042 "fl_cl_enqueued", CTLFLAG_RD, &fl->ifl_cl_enqueued, 7043 "clusters allocated"); 7044 SYSCTL_ADD_UQUAD(ctx_list, fl_list, OID_AUTO, 7045 "fl_cl_dequeued", CTLFLAG_RD, &fl->ifl_cl_dequeued, 7046 "clusters freed"); 7047 #endif 7048 } 7049 } 7050 7051 } 7052 7053 void 7054 iflib_request_reset(if_ctx_t ctx) 7055 { 7056 7057 STATE_LOCK(ctx); 7058 ctx->ifc_flags |= IFC_DO_RESET; 7059 STATE_UNLOCK(ctx); 7060 } 7061 7062 #ifndef __NO_STRICT_ALIGNMENT 7063 static struct mbuf * 7064 iflib_fixup_rx(struct mbuf *m) 7065 { 7066 struct mbuf *n; 7067 7068 if (m->m_len <= (MCLBYTES - ETHER_HDR_LEN)) { 7069 bcopy(m->m_data, m->m_data + ETHER_HDR_LEN, m->m_len); 7070 m->m_data += ETHER_HDR_LEN; 7071 n = m; 7072 } else { 7073 MGETHDR(n, M_NOWAIT, MT_DATA); 7074 if (n == NULL) { 7075 m_freem(m); 7076 return (NULL); 7077 } 7078 bcopy(m->m_data, n->m_data, ETHER_HDR_LEN); 7079 m->m_data += ETHER_HDR_LEN; 7080 m->m_len -= ETHER_HDR_LEN; 7081 n->m_len = ETHER_HDR_LEN; 7082 M_MOVE_PKTHDR(n, m); 7083 n->m_next = m; 7084 } 7085 return (n); 7086 } 7087 #endif 7088 7089 #ifdef DEBUGNET 7090 static void 7091 iflib_debugnet_init(if_t ifp, int *nrxr, int *ncl, int *clsize) 7092 { 7093 if_ctx_t ctx; 7094 7095 ctx = if_getsoftc(ifp); 7096 CTX_LOCK(ctx); 7097 *nrxr = NRXQSETS(ctx); 7098 *ncl = ctx->ifc_rxqs[0].ifr_fl->ifl_size; 7099 *clsize = ctx->ifc_rxqs[0].ifr_fl->ifl_buf_size; 7100 CTX_UNLOCK(ctx); 7101 } 7102 7103 static void 7104 iflib_debugnet_event(if_t ifp, enum debugnet_ev event) 7105 { 7106 if_ctx_t ctx; 7107 if_softc_ctx_t scctx; 7108 iflib_fl_t fl; 7109 iflib_rxq_t rxq; 7110 int i, j; 7111 7112 ctx = if_getsoftc(ifp); 7113 scctx = &ctx->ifc_softc_ctx; 7114 7115 switch (event) { 7116 case DEBUGNET_START: 7117 for (i = 0; i < scctx->isc_nrxqsets; i++) { 7118 rxq = &ctx->ifc_rxqs[i]; 7119 for (j = 0; j < rxq->ifr_nfl; j++) { 7120 fl = rxq->ifr_fl; 7121 fl->ifl_zone = m_getzone(fl->ifl_buf_size); 7122 } 7123 } 7124 iflib_no_tx_batch = 1; 7125 break; 7126 default: 7127 break; 7128 } 7129 } 7130 7131 static int 7132 iflib_debugnet_transmit(if_t ifp, struct mbuf *m) 7133 { 7134 if_ctx_t ctx; 7135 iflib_txq_t txq; 7136 int error; 7137 7138 ctx = if_getsoftc(ifp); 7139 if ((if_getdrvflags(ifp) & (IFF_DRV_RUNNING | IFF_DRV_OACTIVE)) != 7140 IFF_DRV_RUNNING) 7141 return (EBUSY); 7142 7143 txq = &ctx->ifc_txqs[0]; 7144 error = iflib_encap(txq, &m); 7145 if (error == 0) 7146 (void)iflib_txd_db_check(txq, true); 7147 return (error); 7148 } 7149 7150 static int 7151 iflib_debugnet_poll(if_t ifp, int count) 7152 { 7153 struct epoch_tracker et; 7154 if_ctx_t ctx; 7155 if_softc_ctx_t scctx; 7156 iflib_txq_t txq; 7157 int i; 7158 7159 ctx = if_getsoftc(ifp); 7160 scctx = &ctx->ifc_softc_ctx; 7161 7162 if ((if_getdrvflags(ifp) & (IFF_DRV_RUNNING | IFF_DRV_OACTIVE)) != 7163 IFF_DRV_RUNNING) 7164 return (EBUSY); 7165 7166 txq = &ctx->ifc_txqs[0]; 7167 (void)iflib_completed_tx_reclaim(txq, RECLAIM_THRESH(ctx)); 7168 7169 NET_EPOCH_ENTER(et); 7170 for (i = 0; i < scctx->isc_nrxqsets; i++) 7171 (void)iflib_rxeof(&ctx->ifc_rxqs[i], 16 /* XXX */); 7172 NET_EPOCH_EXIT(et); 7173 return (0); 7174 } 7175 #endif /* DEBUGNET */ 7176