1 /*- 2 * Copyright (c) 2014-2017, Matthew Macy <mmacy@nextbsd.org> 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 __FBSDID("$FreeBSD$"); 30 31 #include "opt_inet.h" 32 #include "opt_inet6.h" 33 #include "opt_acpi.h" 34 35 #include <sys/param.h> 36 #include <sys/types.h> 37 #include <sys/bus.h> 38 #include <sys/eventhandler.h> 39 #include <sys/sockio.h> 40 #include <sys/kernel.h> 41 #include <sys/lock.h> 42 #include <sys/mutex.h> 43 #include <sys/module.h> 44 #include <sys/kobj.h> 45 #include <sys/rman.h> 46 #include <sys/sbuf.h> 47 #include <sys/smp.h> 48 #include <sys/socket.h> 49 #include <sys/sysctl.h> 50 #include <sys/syslog.h> 51 #include <sys/taskqueue.h> 52 #include <sys/limits.h> 53 54 55 #include <net/if.h> 56 #include <net/if_var.h> 57 #include <net/if_types.h> 58 #include <net/if_media.h> 59 #include <net/bpf.h> 60 #include <net/ethernet.h> 61 #include <net/mp_ring.h> 62 #include <net/vnet.h> 63 64 #include <netinet/in.h> 65 #include <netinet/in_pcb.h> 66 #include <netinet/tcp_lro.h> 67 #include <netinet/in_systm.h> 68 #include <netinet/if_ether.h> 69 #include <netinet/ip.h> 70 #include <netinet/ip6.h> 71 #include <netinet/tcp.h> 72 #include <netinet/ip_var.h> 73 #include <netinet6/ip6_var.h> 74 75 #include <machine/bus.h> 76 #include <machine/in_cksum.h> 77 78 #include <vm/vm.h> 79 #include <vm/pmap.h> 80 81 #include <dev/led/led.h> 82 #include <dev/pci/pcireg.h> 83 #include <dev/pci/pcivar.h> 84 #include <dev/pci/pci_private.h> 85 86 #include <net/iflib.h> 87 88 #include "ifdi_if.h" 89 90 #if defined(__i386__) || defined(__amd64__) 91 #include <sys/memdesc.h> 92 #include <machine/bus.h> 93 #include <machine/md_var.h> 94 #include <machine/specialreg.h> 95 #include <x86/include/busdma_impl.h> 96 #include <x86/iommu/busdma_dmar.h> 97 #endif 98 99 #include <sys/bitstring.h> 100 /* 101 * enable accounting of every mbuf as it comes in to and goes out of 102 * iflib's software descriptor references 103 */ 104 #define MEMORY_LOGGING 0 105 /* 106 * Enable mbuf vectors for compressing long mbuf chains 107 */ 108 109 /* 110 * NB: 111 * - Prefetching in tx cleaning should perhaps be a tunable. The distance ahead 112 * we prefetch needs to be determined by the time spent in m_free vis a vis 113 * the cost of a prefetch. This will of course vary based on the workload: 114 * - NFLX's m_free path is dominated by vm-based M_EXT manipulation which 115 * is quite expensive, thus suggesting very little prefetch. 116 * - small packet forwarding which is just returning a single mbuf to 117 * UMA will typically be very fast vis a vis the cost of a memory 118 * access. 119 */ 120 121 122 /* 123 * File organization: 124 * - private structures 125 * - iflib private utility functions 126 * - ifnet functions 127 * - vlan registry and other exported functions 128 * - iflib public core functions 129 * 130 * 131 */ 132 static MALLOC_DEFINE(M_IFLIB, "iflib", "ifnet library"); 133 134 struct iflib_txq; 135 typedef struct iflib_txq *iflib_txq_t; 136 struct iflib_rxq; 137 typedef struct iflib_rxq *iflib_rxq_t; 138 struct iflib_fl; 139 typedef struct iflib_fl *iflib_fl_t; 140 141 struct iflib_ctx; 142 143 static void iru_init(if_rxd_update_t iru, iflib_rxq_t rxq, uint8_t flid); 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 mtx ifc_mtx; 166 167 uint16_t ifc_nhwtxqs; 168 uint16_t ifc_nhwrxqs; 169 170 iflib_txq_t ifc_txqs; 171 iflib_rxq_t ifc_rxqs; 172 uint32_t ifc_if_flags; 173 uint32_t ifc_flags; 174 uint32_t ifc_max_fl_buf_size; 175 int ifc_in_detach; 176 177 int ifc_link_state; 178 int ifc_link_irq; 179 int ifc_watchdog_events; 180 struct cdev *ifc_led_dev; 181 struct resource *ifc_msix_mem; 182 183 struct if_irq ifc_legacy_irq; 184 struct grouptask ifc_admin_task; 185 struct grouptask ifc_vflr_task; 186 struct iflib_filter_info ifc_filter_info; 187 struct ifmedia ifc_media; 188 189 struct sysctl_oid *ifc_sysctl_node; 190 uint16_t ifc_sysctl_ntxqs; 191 uint16_t ifc_sysctl_nrxqs; 192 uint16_t ifc_sysctl_qs_eq_override; 193 uint16_t ifc_sysctl_rx_budget; 194 195 qidx_t ifc_sysctl_ntxds[8]; 196 qidx_t ifc_sysctl_nrxds[8]; 197 struct if_txrx ifc_txrx; 198 #define isc_txd_encap ifc_txrx.ift_txd_encap 199 #define isc_txd_flush ifc_txrx.ift_txd_flush 200 #define isc_txd_credits_update ifc_txrx.ift_txd_credits_update 201 #define isc_rxd_available ifc_txrx.ift_rxd_available 202 #define isc_rxd_pkt_get ifc_txrx.ift_rxd_pkt_get 203 #define isc_rxd_refill ifc_txrx.ift_rxd_refill 204 #define isc_rxd_flush ifc_txrx.ift_rxd_flush 205 #define isc_rxd_refill ifc_txrx.ift_rxd_refill 206 #define isc_rxd_refill ifc_txrx.ift_rxd_refill 207 #define isc_legacy_intr ifc_txrx.ift_legacy_intr 208 eventhandler_tag ifc_vlan_attach_event; 209 eventhandler_tag ifc_vlan_detach_event; 210 uint8_t ifc_mac[ETHER_ADDR_LEN]; 211 char ifc_mtx_name[16]; 212 }; 213 214 215 void * 216 iflib_get_softc(if_ctx_t ctx) 217 { 218 219 return (ctx->ifc_softc); 220 } 221 222 device_t 223 iflib_get_dev(if_ctx_t ctx) 224 { 225 226 return (ctx->ifc_dev); 227 } 228 229 if_t 230 iflib_get_ifp(if_ctx_t ctx) 231 { 232 233 return (ctx->ifc_ifp); 234 } 235 236 struct ifmedia * 237 iflib_get_media(if_ctx_t ctx) 238 { 239 240 return (&ctx->ifc_media); 241 } 242 243 void 244 iflib_set_mac(if_ctx_t ctx, uint8_t mac[ETHER_ADDR_LEN]) 245 { 246 247 bcopy(mac, ctx->ifc_mac, ETHER_ADDR_LEN); 248 } 249 250 if_softc_ctx_t 251 iflib_get_softc_ctx(if_ctx_t ctx) 252 { 253 254 return (&ctx->ifc_softc_ctx); 255 } 256 257 if_shared_ctx_t 258 iflib_get_sctx(if_ctx_t ctx) 259 { 260 261 return (ctx->ifc_sctx); 262 } 263 264 #define IP_ALIGNED(m) ((((uintptr_t)(m)->m_data) & 0x3) == 0x2) 265 #define CACHE_PTR_INCREMENT (CACHE_LINE_SIZE/sizeof(void*)) 266 #define CACHE_PTR_NEXT(ptr) ((void *)(((uintptr_t)(ptr)+CACHE_LINE_SIZE-1) & (CACHE_LINE_SIZE-1))) 267 268 #define LINK_ACTIVE(ctx) ((ctx)->ifc_link_state == LINK_STATE_UP) 269 #define CTX_IS_VF(ctx) ((ctx)->ifc_sctx->isc_flags & IFLIB_IS_VF) 270 271 #define RX_SW_DESC_MAP_CREATED (1 << 0) 272 #define TX_SW_DESC_MAP_CREATED (1 << 1) 273 #define RX_SW_DESC_INUSE (1 << 3) 274 #define TX_SW_DESC_MAPPED (1 << 4) 275 276 #define M_TOOBIG M_PROTO1 277 278 typedef struct iflib_sw_rx_desc_array { 279 bus_dmamap_t *ifsd_map; /* bus_dma maps for packet */ 280 struct mbuf **ifsd_m; /* pkthdr mbufs */ 281 caddr_t *ifsd_cl; /* direct cluster pointer for rx */ 282 uint8_t *ifsd_flags; 283 } iflib_rxsd_array_t; 284 285 typedef struct iflib_sw_tx_desc_array { 286 bus_dmamap_t *ifsd_map; /* bus_dma maps for packet */ 287 struct mbuf **ifsd_m; /* pkthdr mbufs */ 288 uint8_t *ifsd_flags; 289 } if_txsd_vec_t; 290 291 292 /* magic number that should be high enough for any hardware */ 293 #define IFLIB_MAX_TX_SEGS 128 294 /* bnxt supports 64 with hardware LRO enabled */ 295 #define IFLIB_MAX_RX_SEGS 64 296 #define IFLIB_RX_COPY_THRESH 128 297 #define IFLIB_MAX_RX_REFRESH 32 298 /* The minimum descriptors per second before we start coalescing */ 299 #define IFLIB_MIN_DESC_SEC 16384 300 #define IFLIB_DEFAULT_TX_UPDATE_FREQ 16 301 #define IFLIB_QUEUE_IDLE 0 302 #define IFLIB_QUEUE_HUNG 1 303 #define IFLIB_QUEUE_WORKING 2 304 /* maximum number of txqs that can share an rx interrupt */ 305 #define IFLIB_MAX_TX_SHARED_INTR 4 306 307 /* this should really scale with ring size - this is a fairly arbitrary value */ 308 #define TX_BATCH_SIZE 32 309 310 #define IFLIB_RESTART_BUDGET 8 311 312 #define IFC_LEGACY 0x001 313 #define IFC_QFLUSH 0x002 314 #define IFC_MULTISEG 0x004 315 #define IFC_DMAR 0x008 316 #define IFC_SC_ALLOCATED 0x010 317 #define IFC_INIT_DONE 0x020 318 #define IFC_PREFETCH 0x040 319 #define IFC_DO_RESET 0x080 320 #define IFC_CHECK_HUNG 0x100 321 322 #define CSUM_OFFLOAD (CSUM_IP_TSO|CSUM_IP6_TSO|CSUM_IP| \ 323 CSUM_IP_UDP|CSUM_IP_TCP|CSUM_IP_SCTP| \ 324 CSUM_IP6_UDP|CSUM_IP6_TCP|CSUM_IP6_SCTP) 325 struct iflib_txq { 326 qidx_t ift_in_use; 327 qidx_t ift_cidx; 328 qidx_t ift_cidx_processed; 329 qidx_t ift_pidx; 330 uint8_t ift_gen; 331 uint8_t ift_br_offset; 332 uint16_t ift_npending; 333 uint16_t ift_db_pending; 334 uint16_t ift_rs_pending; 335 /* implicit pad */ 336 uint8_t ift_txd_size[8]; 337 uint64_t ift_processed; 338 uint64_t ift_cleaned; 339 uint64_t ift_cleaned_prev; 340 #if MEMORY_LOGGING 341 uint64_t ift_enqueued; 342 uint64_t ift_dequeued; 343 #endif 344 uint64_t ift_no_tx_dma_setup; 345 uint64_t ift_no_desc_avail; 346 uint64_t ift_mbuf_defrag_failed; 347 uint64_t ift_mbuf_defrag; 348 uint64_t ift_map_failed; 349 uint64_t ift_txd_encap_efbig; 350 uint64_t ift_pullups; 351 352 struct mtx ift_mtx; 353 struct mtx ift_db_mtx; 354 355 /* constant values */ 356 if_ctx_t ift_ctx; 357 struct ifmp_ring *ift_br; 358 struct grouptask ift_task; 359 qidx_t ift_size; 360 uint16_t ift_id; 361 struct callout ift_timer; 362 363 if_txsd_vec_t ift_sds; 364 uint8_t ift_qstatus; 365 uint8_t ift_closed; 366 uint8_t ift_update_freq; 367 struct iflib_filter_info ift_filter_info; 368 bus_dma_tag_t ift_desc_tag; 369 bus_dma_tag_t ift_tso_desc_tag; 370 iflib_dma_info_t ift_ifdi; 371 #define MTX_NAME_LEN 16 372 char ift_mtx_name[MTX_NAME_LEN]; 373 char ift_db_mtx_name[MTX_NAME_LEN]; 374 bus_dma_segment_t ift_segs[IFLIB_MAX_TX_SEGS] __aligned(CACHE_LINE_SIZE); 375 #ifdef IFLIB_DIAGNOSTICS 376 uint64_t ift_cpu_exec_count[256]; 377 #endif 378 } __aligned(CACHE_LINE_SIZE); 379 380 struct iflib_fl { 381 qidx_t ifl_cidx; 382 qidx_t ifl_pidx; 383 qidx_t ifl_credits; 384 uint8_t ifl_gen; 385 uint8_t ifl_rxd_size; 386 #if MEMORY_LOGGING 387 uint64_t ifl_m_enqueued; 388 uint64_t ifl_m_dequeued; 389 uint64_t ifl_cl_enqueued; 390 uint64_t ifl_cl_dequeued; 391 #endif 392 /* implicit pad */ 393 394 bitstr_t *ifl_rx_bitmap; 395 qidx_t ifl_fragidx; 396 /* constant */ 397 qidx_t ifl_size; 398 uint16_t ifl_buf_size; 399 uint16_t ifl_cltype; 400 uma_zone_t ifl_zone; 401 iflib_rxsd_array_t ifl_sds; 402 iflib_rxq_t ifl_rxq; 403 uint8_t ifl_id; 404 bus_dma_tag_t ifl_desc_tag; 405 iflib_dma_info_t ifl_ifdi; 406 uint64_t ifl_bus_addrs[IFLIB_MAX_RX_REFRESH] __aligned(CACHE_LINE_SIZE); 407 caddr_t ifl_vm_addrs[IFLIB_MAX_RX_REFRESH]; 408 qidx_t ifl_rxd_idxs[IFLIB_MAX_RX_REFRESH]; 409 } __aligned(CACHE_LINE_SIZE); 410 411 static inline qidx_t 412 get_inuse(int size, qidx_t cidx, qidx_t pidx, uint8_t gen) 413 { 414 qidx_t used; 415 416 if (pidx > cidx) 417 used = pidx - cidx; 418 else if (pidx < cidx) 419 used = size - cidx + pidx; 420 else if (gen == 0 && pidx == cidx) 421 used = 0; 422 else if (gen == 1 && pidx == cidx) 423 used = size; 424 else 425 panic("bad state"); 426 427 return (used); 428 } 429 430 #define TXQ_AVAIL(txq) (txq->ift_size - get_inuse(txq->ift_size, txq->ift_cidx, txq->ift_pidx, txq->ift_gen)) 431 432 #define IDXDIFF(head, tail, wrap) \ 433 ((head) >= (tail) ? (head) - (tail) : (wrap) - (tail) + (head)) 434 435 struct iflib_rxq { 436 /* If there is a separate completion queue - 437 * these are the cq cidx and pidx. Otherwise 438 * these are unused. 439 */ 440 qidx_t ifr_size; 441 qidx_t ifr_cq_cidx; 442 qidx_t ifr_cq_pidx; 443 uint8_t ifr_cq_gen; 444 uint8_t ifr_fl_offset; 445 446 if_ctx_t ifr_ctx; 447 iflib_fl_t ifr_fl; 448 uint64_t ifr_rx_irq; 449 uint16_t ifr_id; 450 uint8_t ifr_lro_enabled; 451 uint8_t ifr_nfl; 452 uint8_t ifr_ntxqirq; 453 uint8_t ifr_txqid[IFLIB_MAX_TX_SHARED_INTR]; 454 struct lro_ctrl ifr_lc; 455 struct grouptask ifr_task; 456 struct iflib_filter_info ifr_filter_info; 457 iflib_dma_info_t ifr_ifdi; 458 459 /* dynamically allocate if any drivers need a value substantially larger than this */ 460 struct if_rxd_frag ifr_frags[IFLIB_MAX_RX_SEGS] __aligned(CACHE_LINE_SIZE); 461 #ifdef IFLIB_DIAGNOSTICS 462 uint64_t ifr_cpu_exec_count[256]; 463 #endif 464 } __aligned(CACHE_LINE_SIZE); 465 466 typedef struct if_rxsd { 467 caddr_t *ifsd_cl; 468 struct mbuf **ifsd_m; 469 iflib_fl_t ifsd_fl; 470 qidx_t ifsd_cidx; 471 } *if_rxsd_t; 472 473 /* multiple of word size */ 474 #ifdef __LP64__ 475 #define PKT_INFO_SIZE 6 476 #define RXD_INFO_SIZE 5 477 #define PKT_TYPE uint64_t 478 #else 479 #define PKT_INFO_SIZE 11 480 #define RXD_INFO_SIZE 8 481 #define PKT_TYPE uint32_t 482 #endif 483 #define PKT_LOOP_BOUND ((PKT_INFO_SIZE/3)*3) 484 #define RXD_LOOP_BOUND ((RXD_INFO_SIZE/4)*4) 485 486 typedef struct if_pkt_info_pad { 487 PKT_TYPE pkt_val[PKT_INFO_SIZE]; 488 } *if_pkt_info_pad_t; 489 typedef struct if_rxd_info_pad { 490 PKT_TYPE rxd_val[RXD_INFO_SIZE]; 491 } *if_rxd_info_pad_t; 492 493 CTASSERT(sizeof(struct if_pkt_info_pad) == sizeof(struct if_pkt_info)); 494 CTASSERT(sizeof(struct if_rxd_info_pad) == sizeof(struct if_rxd_info)); 495 496 497 static inline void 498 pkt_info_zero(if_pkt_info_t pi) 499 { 500 if_pkt_info_pad_t pi_pad; 501 502 pi_pad = (if_pkt_info_pad_t)pi; 503 pi_pad->pkt_val[0] = 0; pi_pad->pkt_val[1] = 0; pi_pad->pkt_val[2] = 0; 504 pi_pad->pkt_val[3] = 0; pi_pad->pkt_val[4] = 0; pi_pad->pkt_val[5] = 0; 505 #ifndef __LP64__ 506 pi_pad->pkt_val[6] = 0; pi_pad->pkt_val[7] = 0; pi_pad->pkt_val[8] = 0; 507 pi_pad->pkt_val[9] = 0; pi_pad->pkt_val[10] = 0; 508 #endif 509 } 510 511 static inline void 512 rxd_info_zero(if_rxd_info_t ri) 513 { 514 if_rxd_info_pad_t ri_pad; 515 int i; 516 517 ri_pad = (if_rxd_info_pad_t)ri; 518 for (i = 0; i < RXD_LOOP_BOUND; i += 4) { 519 ri_pad->rxd_val[i] = 0; 520 ri_pad->rxd_val[i+1] = 0; 521 ri_pad->rxd_val[i+2] = 0; 522 ri_pad->rxd_val[i+3] = 0; 523 } 524 #ifdef __LP64__ 525 ri_pad->rxd_val[RXD_INFO_SIZE-1] = 0; 526 #endif 527 } 528 529 /* 530 * Only allow a single packet to take up most 1/nth of the tx ring 531 */ 532 #define MAX_SINGLE_PACKET_FRACTION 12 533 #define IF_BAD_DMA (bus_addr_t)-1 534 535 #define CTX_ACTIVE(ctx) ((if_getdrvflags((ctx)->ifc_ifp) & IFF_DRV_RUNNING)) 536 537 #define CTX_LOCK_INIT(_sc, _name) mtx_init(&(_sc)->ifc_mtx, _name, "iflib ctx lock", MTX_DEF) 538 539 #define CTX_LOCK(ctx) mtx_lock(&(ctx)->ifc_mtx) 540 #define CTX_UNLOCK(ctx) mtx_unlock(&(ctx)->ifc_mtx) 541 #define CTX_LOCK_DESTROY(ctx) mtx_destroy(&(ctx)->ifc_mtx) 542 543 544 #define CALLOUT_LOCK(txq) mtx_lock(&txq->ift_mtx) 545 #define CALLOUT_UNLOCK(txq) mtx_unlock(&txq->ift_mtx) 546 547 548 /* Our boot-time initialization hook */ 549 static int iflib_module_event_handler(module_t, int, void *); 550 551 static moduledata_t iflib_moduledata = { 552 "iflib", 553 iflib_module_event_handler, 554 NULL 555 }; 556 557 DECLARE_MODULE(iflib, iflib_moduledata, SI_SUB_INIT_IF, SI_ORDER_ANY); 558 MODULE_VERSION(iflib, 1); 559 560 MODULE_DEPEND(iflib, pci, 1, 1, 1); 561 MODULE_DEPEND(iflib, ether, 1, 1, 1); 562 563 TASKQGROUP_DEFINE(if_io_tqg, mp_ncpus, 1); 564 TASKQGROUP_DEFINE(if_config_tqg, 1, 1); 565 566 #ifndef IFLIB_DEBUG_COUNTERS 567 #ifdef INVARIANTS 568 #define IFLIB_DEBUG_COUNTERS 1 569 #else 570 #define IFLIB_DEBUG_COUNTERS 0 571 #endif /* !INVARIANTS */ 572 #endif 573 574 static SYSCTL_NODE(_net, OID_AUTO, iflib, CTLFLAG_RD, 0, 575 "iflib driver parameters"); 576 577 /* 578 * XXX need to ensure that this can't accidentally cause the head to be moved backwards 579 */ 580 static int iflib_min_tx_latency = 0; 581 SYSCTL_INT(_net_iflib, OID_AUTO, min_tx_latency, CTLFLAG_RW, 582 &iflib_min_tx_latency, 0, "minimize transmit latency at the possible expense of throughput"); 583 static int iflib_no_tx_batch = 0; 584 SYSCTL_INT(_net_iflib, OID_AUTO, no_tx_batch, CTLFLAG_RW, 585 &iflib_no_tx_batch, 0, "minimize transmit latency at the possible expense of throughput"); 586 587 588 #if IFLIB_DEBUG_COUNTERS 589 590 static int iflib_tx_seen; 591 static int iflib_tx_sent; 592 static int iflib_tx_encap; 593 static int iflib_rx_allocs; 594 static int iflib_fl_refills; 595 static int iflib_fl_refills_large; 596 static int iflib_tx_frees; 597 598 SYSCTL_INT(_net_iflib, OID_AUTO, tx_seen, CTLFLAG_RD, 599 &iflib_tx_seen, 0, "# tx mbufs seen"); 600 SYSCTL_INT(_net_iflib, OID_AUTO, tx_sent, CTLFLAG_RD, 601 &iflib_tx_sent, 0, "# tx mbufs sent"); 602 SYSCTL_INT(_net_iflib, OID_AUTO, tx_encap, CTLFLAG_RD, 603 &iflib_tx_encap, 0, "# tx mbufs encapped"); 604 SYSCTL_INT(_net_iflib, OID_AUTO, tx_frees, CTLFLAG_RD, 605 &iflib_tx_frees, 0, "# tx frees"); 606 SYSCTL_INT(_net_iflib, OID_AUTO, rx_allocs, CTLFLAG_RD, 607 &iflib_rx_allocs, 0, "# rx allocations"); 608 SYSCTL_INT(_net_iflib, OID_AUTO, fl_refills, CTLFLAG_RD, 609 &iflib_fl_refills, 0, "# refills"); 610 SYSCTL_INT(_net_iflib, OID_AUTO, fl_refills_large, CTLFLAG_RD, 611 &iflib_fl_refills_large, 0, "# large refills"); 612 613 614 static int iflib_txq_drain_flushing; 615 static int iflib_txq_drain_oactive; 616 static int iflib_txq_drain_notready; 617 static int iflib_txq_drain_encapfail; 618 619 SYSCTL_INT(_net_iflib, OID_AUTO, txq_drain_flushing, CTLFLAG_RD, 620 &iflib_txq_drain_flushing, 0, "# drain flushes"); 621 SYSCTL_INT(_net_iflib, OID_AUTO, txq_drain_oactive, CTLFLAG_RD, 622 &iflib_txq_drain_oactive, 0, "# drain oactives"); 623 SYSCTL_INT(_net_iflib, OID_AUTO, txq_drain_notready, CTLFLAG_RD, 624 &iflib_txq_drain_notready, 0, "# drain notready"); 625 SYSCTL_INT(_net_iflib, OID_AUTO, txq_drain_encapfail, CTLFLAG_RD, 626 &iflib_txq_drain_encapfail, 0, "# drain encap fails"); 627 628 629 static int iflib_encap_load_mbuf_fail; 630 static int iflib_encap_pad_mbuf_fail; 631 static int iflib_encap_txq_avail_fail; 632 static int iflib_encap_txd_encap_fail; 633 634 SYSCTL_INT(_net_iflib, OID_AUTO, encap_load_mbuf_fail, CTLFLAG_RD, 635 &iflib_encap_load_mbuf_fail, 0, "# busdma load failures"); 636 SYSCTL_INT(_net_iflib, OID_AUTO, encap_pad_mbuf_fail, CTLFLAG_RD, 637 &iflib_encap_pad_mbuf_fail, 0, "# runt frame pad failures"); 638 SYSCTL_INT(_net_iflib, OID_AUTO, encap_txq_avail_fail, CTLFLAG_RD, 639 &iflib_encap_txq_avail_fail, 0, "# txq avail failures"); 640 SYSCTL_INT(_net_iflib, OID_AUTO, encap_txd_encap_fail, CTLFLAG_RD, 641 &iflib_encap_txd_encap_fail, 0, "# driver encap failures"); 642 643 static int iflib_task_fn_rxs; 644 static int iflib_rx_intr_enables; 645 static int iflib_fast_intrs; 646 static int iflib_intr_link; 647 static int iflib_intr_msix; 648 static int iflib_rx_unavail; 649 static int iflib_rx_ctx_inactive; 650 static int iflib_rx_zero_len; 651 static int iflib_rx_if_input; 652 static int iflib_rx_mbuf_null; 653 static int iflib_rxd_flush; 654 655 static int iflib_verbose_debug; 656 657 SYSCTL_INT(_net_iflib, OID_AUTO, intr_link, CTLFLAG_RD, 658 &iflib_intr_link, 0, "# intr link calls"); 659 SYSCTL_INT(_net_iflib, OID_AUTO, intr_msix, CTLFLAG_RD, 660 &iflib_intr_msix, 0, "# intr msix calls"); 661 SYSCTL_INT(_net_iflib, OID_AUTO, task_fn_rx, CTLFLAG_RD, 662 &iflib_task_fn_rxs, 0, "# task_fn_rx calls"); 663 SYSCTL_INT(_net_iflib, OID_AUTO, rx_intr_enables, CTLFLAG_RD, 664 &iflib_rx_intr_enables, 0, "# rx intr enables"); 665 SYSCTL_INT(_net_iflib, OID_AUTO, fast_intrs, CTLFLAG_RD, 666 &iflib_fast_intrs, 0, "# fast_intr calls"); 667 SYSCTL_INT(_net_iflib, OID_AUTO, rx_unavail, CTLFLAG_RD, 668 &iflib_rx_unavail, 0, "# times rxeof called with no available data"); 669 SYSCTL_INT(_net_iflib, OID_AUTO, rx_ctx_inactive, CTLFLAG_RD, 670 &iflib_rx_ctx_inactive, 0, "# times rxeof called with inactive context"); 671 SYSCTL_INT(_net_iflib, OID_AUTO, rx_zero_len, CTLFLAG_RD, 672 &iflib_rx_zero_len, 0, "# times rxeof saw zero len mbuf"); 673 SYSCTL_INT(_net_iflib, OID_AUTO, rx_if_input, CTLFLAG_RD, 674 &iflib_rx_if_input, 0, "# times rxeof called if_input"); 675 SYSCTL_INT(_net_iflib, OID_AUTO, rx_mbuf_null, CTLFLAG_RD, 676 &iflib_rx_mbuf_null, 0, "# times rxeof got null mbuf"); 677 SYSCTL_INT(_net_iflib, OID_AUTO, rxd_flush, CTLFLAG_RD, 678 &iflib_rxd_flush, 0, "# times rxd_flush called"); 679 SYSCTL_INT(_net_iflib, OID_AUTO, verbose_debug, CTLFLAG_RW, 680 &iflib_verbose_debug, 0, "enable verbose debugging"); 681 682 #define DBG_COUNTER_INC(name) atomic_add_int(&(iflib_ ## name), 1) 683 static void 684 iflib_debug_reset(void) 685 { 686 iflib_tx_seen = iflib_tx_sent = iflib_tx_encap = iflib_rx_allocs = 687 iflib_fl_refills = iflib_fl_refills_large = iflib_tx_frees = 688 iflib_txq_drain_flushing = iflib_txq_drain_oactive = 689 iflib_txq_drain_notready = iflib_txq_drain_encapfail = 690 iflib_encap_load_mbuf_fail = iflib_encap_pad_mbuf_fail = 691 iflib_encap_txq_avail_fail = iflib_encap_txd_encap_fail = 692 iflib_task_fn_rxs = iflib_rx_intr_enables = iflib_fast_intrs = 693 iflib_intr_link = iflib_intr_msix = iflib_rx_unavail = 694 iflib_rx_ctx_inactive = iflib_rx_zero_len = iflib_rx_if_input = 695 iflib_rx_mbuf_null = iflib_rxd_flush = 0; 696 } 697 698 #else 699 #define DBG_COUNTER_INC(name) 700 static void iflib_debug_reset(void) {} 701 #endif 702 703 704 705 #define IFLIB_DEBUG 0 706 707 static void iflib_tx_structures_free(if_ctx_t ctx); 708 static void iflib_rx_structures_free(if_ctx_t ctx); 709 static int iflib_queues_alloc(if_ctx_t ctx); 710 static int iflib_tx_credits_update(if_ctx_t ctx, iflib_txq_t txq); 711 static int iflib_rxd_avail(if_ctx_t ctx, iflib_rxq_t rxq, qidx_t cidx, qidx_t budget); 712 static int iflib_qset_structures_setup(if_ctx_t ctx); 713 static int iflib_msix_init(if_ctx_t ctx); 714 static int iflib_legacy_setup(if_ctx_t ctx, driver_filter_t filter, void *filterarg, int *rid, char *str); 715 static void iflib_txq_check_drain(iflib_txq_t txq, int budget); 716 static uint32_t iflib_txq_can_drain(struct ifmp_ring *); 717 static int iflib_register(if_ctx_t); 718 static void iflib_init_locked(if_ctx_t ctx); 719 static void iflib_add_device_sysctl_pre(if_ctx_t ctx); 720 static void iflib_add_device_sysctl_post(if_ctx_t ctx); 721 static void iflib_ifmp_purge(iflib_txq_t txq); 722 static void _iflib_pre_assert(if_softc_ctx_t scctx); 723 static void iflib_stop(if_ctx_t ctx); 724 static void iflib_if_init_locked(if_ctx_t ctx); 725 #ifndef __NO_STRICT_ALIGNMENT 726 static struct mbuf * iflib_fixup_rx(struct mbuf *m); 727 #endif 728 729 #ifdef DEV_NETMAP 730 #include <sys/selinfo.h> 731 #include <net/netmap.h> 732 #include <dev/netmap/netmap_kern.h> 733 734 MODULE_DEPEND(iflib, netmap, 1, 1, 1); 735 736 static int netmap_fl_refill(iflib_rxq_t rxq, struct netmap_kring *kring, uint32_t nm_i, bool init); 737 738 /* 739 * device-specific sysctl variables: 740 * 741 * iflib_crcstrip: 0: keep CRC in rx frames (default), 1: strip it. 742 * During regular operations the CRC is stripped, but on some 743 * hardware reception of frames not multiple of 64 is slower, 744 * so using crcstrip=0 helps in benchmarks. 745 * 746 * iflib_rx_miss, iflib_rx_miss_bufs: 747 * count packets that might be missed due to lost interrupts. 748 */ 749 SYSCTL_DECL(_dev_netmap); 750 /* 751 * The xl driver by default strips CRCs and we do not override it. 752 */ 753 754 int iflib_crcstrip = 1; 755 SYSCTL_INT(_dev_netmap, OID_AUTO, iflib_crcstrip, 756 CTLFLAG_RW, &iflib_crcstrip, 1, "strip CRC on rx frames"); 757 758 int iflib_rx_miss, iflib_rx_miss_bufs; 759 SYSCTL_INT(_dev_netmap, OID_AUTO, iflib_rx_miss, 760 CTLFLAG_RW, &iflib_rx_miss, 0, "potentially missed rx intr"); 761 SYSCTL_INT(_dev_netmap, OID_AUTO, iflib_rx_miss_bufs, 762 CTLFLAG_RW, &iflib_rx_miss_bufs, 0, "potentially missed rx intr bufs"); 763 764 /* 765 * Register/unregister. We are already under netmap lock. 766 * Only called on the first register or the last unregister. 767 */ 768 static int 769 iflib_netmap_register(struct netmap_adapter *na, int onoff) 770 { 771 struct ifnet *ifp = na->ifp; 772 if_ctx_t ctx = ifp->if_softc; 773 int status; 774 775 CTX_LOCK(ctx); 776 IFDI_INTR_DISABLE(ctx); 777 778 /* Tell the stack that the interface is no longer active */ 779 ifp->if_drv_flags &= ~(IFF_DRV_RUNNING | IFF_DRV_OACTIVE); 780 781 if (!CTX_IS_VF(ctx)) 782 IFDI_CRCSTRIP_SET(ctx, onoff, iflib_crcstrip); 783 784 /* enable or disable flags and callbacks in na and ifp */ 785 if (onoff) { 786 nm_set_native_flags(na); 787 } else { 788 nm_clear_native_flags(na); 789 } 790 iflib_stop(ctx); 791 iflib_init_locked(ctx); 792 IFDI_CRCSTRIP_SET(ctx, onoff, iflib_crcstrip); // XXX why twice ? 793 status = ifp->if_drv_flags & IFF_DRV_RUNNING ? 0 : 1; 794 if (status) 795 nm_clear_native_flags(na); 796 CTX_UNLOCK(ctx); 797 return (status); 798 } 799 800 static int 801 netmap_fl_refill(iflib_rxq_t rxq, struct netmap_kring *kring, uint32_t nm_i, bool init) 802 { 803 struct netmap_adapter *na = kring->na; 804 u_int const lim = kring->nkr_num_slots - 1; 805 u_int head = kring->rhead; 806 struct netmap_ring *ring = kring->ring; 807 bus_dmamap_t *map; 808 struct if_rxd_update iru; 809 if_ctx_t ctx = rxq->ifr_ctx; 810 iflib_fl_t fl = &rxq->ifr_fl[0]; 811 uint32_t refill_pidx, nic_i; 812 813 if (nm_i == head && __predict_true(!init)) 814 return 0; 815 iru_init(&iru, rxq, 0 /* flid */); 816 map = fl->ifl_sds.ifsd_map; 817 refill_pidx = netmap_idx_k2n(kring, nm_i); 818 /* 819 * IMPORTANT: we must leave one free slot in the ring, 820 * so move head back by one unit 821 */ 822 head = nm_prev(head, lim); 823 while (nm_i != head) { 824 for (int tmp_pidx = 0; tmp_pidx < IFLIB_MAX_RX_REFRESH && nm_i != head; tmp_pidx++) { 825 struct netmap_slot *slot = &ring->slot[nm_i]; 826 void *addr = PNMB(na, slot, &fl->ifl_bus_addrs[tmp_pidx]); 827 uint32_t nic_i_dma = refill_pidx; 828 nic_i = netmap_idx_k2n(kring, nm_i); 829 830 MPASS(tmp_pidx < IFLIB_MAX_RX_REFRESH); 831 832 if (addr == NETMAP_BUF_BASE(na)) /* bad buf */ 833 return netmap_ring_reinit(kring); 834 835 fl->ifl_vm_addrs[tmp_pidx] = addr; 836 if (__predict_false(init) && map) { 837 netmap_load_map(na, fl->ifl_ifdi->idi_tag, map[nic_i], addr); 838 } else if (map && (slot->flags & NS_BUF_CHANGED)) { 839 /* buffer has changed, reload map */ 840 netmap_reload_map(na, fl->ifl_ifdi->idi_tag, map[nic_i], addr); 841 } 842 slot->flags &= ~NS_BUF_CHANGED; 843 844 nm_i = nm_next(nm_i, lim); 845 fl->ifl_rxd_idxs[tmp_pidx] = nic_i = nm_next(nic_i, lim); 846 if (nm_i != head && tmp_pidx < IFLIB_MAX_RX_REFRESH-1) 847 continue; 848 849 iru.iru_pidx = refill_pidx; 850 iru.iru_count = tmp_pidx+1; 851 ctx->isc_rxd_refill(ctx->ifc_softc, &iru); 852 853 refill_pidx = nic_i; 854 if (map == NULL) 855 continue; 856 857 for (int n = 0; n < iru.iru_count; n++) { 858 bus_dmamap_sync(fl->ifl_ifdi->idi_tag, map[nic_i_dma], 859 BUS_DMASYNC_PREREAD); 860 /* XXX - change this to not use the netmap func*/ 861 nic_i_dma = nm_next(nic_i_dma, lim); 862 } 863 } 864 } 865 kring->nr_hwcur = head; 866 867 if (map) 868 bus_dmamap_sync(fl->ifl_ifdi->idi_tag, fl->ifl_ifdi->idi_map, 869 BUS_DMASYNC_PREREAD | BUS_DMASYNC_PREWRITE); 870 ctx->isc_rxd_flush(ctx->ifc_softc, rxq->ifr_id, fl->ifl_id, nic_i); 871 return (0); 872 } 873 874 /* 875 * Reconcile kernel and user view of the transmit ring. 876 * 877 * All information is in the kring. 878 * Userspace wants to send packets up to the one before kring->rhead, 879 * kernel knows kring->nr_hwcur is the first unsent packet. 880 * 881 * Here we push packets out (as many as possible), and possibly 882 * reclaim buffers from previously completed transmission. 883 * 884 * The caller (netmap) guarantees that there is only one instance 885 * running at any time. Any interference with other driver 886 * methods should be handled by the individual drivers. 887 */ 888 static int 889 iflib_netmap_txsync(struct netmap_kring *kring, int flags) 890 { 891 struct netmap_adapter *na = kring->na; 892 struct ifnet *ifp = na->ifp; 893 struct netmap_ring *ring = kring->ring; 894 u_int nm_i; /* index into the netmap ring */ 895 u_int nic_i; /* index into the NIC ring */ 896 u_int n; 897 u_int const lim = kring->nkr_num_slots - 1; 898 u_int const head = kring->rhead; 899 struct if_pkt_info pi; 900 901 /* 902 * interrupts on every tx packet are expensive so request 903 * them every half ring, or where NS_REPORT is set 904 */ 905 u_int report_frequency = kring->nkr_num_slots >> 1; 906 /* device-specific */ 907 if_ctx_t ctx = ifp->if_softc; 908 iflib_txq_t txq = &ctx->ifc_txqs[kring->ring_id]; 909 910 if (txq->ift_sds.ifsd_map) 911 bus_dmamap_sync(txq->ift_desc_tag, txq->ift_ifdi->idi_map, 912 BUS_DMASYNC_POSTREAD | BUS_DMASYNC_POSTWRITE); 913 914 915 /* 916 * First part: process new packets to send. 917 * nm_i is the current index in the netmap ring, 918 * nic_i is the corresponding index in the NIC ring. 919 * 920 * If we have packets to send (nm_i != head) 921 * iterate over the netmap ring, fetch length and update 922 * the corresponding slot in the NIC ring. Some drivers also 923 * need to update the buffer's physical address in the NIC slot 924 * even NS_BUF_CHANGED is not set (PNMB computes the addresses). 925 * 926 * The netmap_reload_map() calls is especially expensive, 927 * even when (as in this case) the tag is 0, so do only 928 * when the buffer has actually changed. 929 * 930 * If possible do not set the report/intr bit on all slots, 931 * but only a few times per ring or when NS_REPORT is set. 932 * 933 * Finally, on 10G and faster drivers, it might be useful 934 * to prefetch the next slot and txr entry. 935 */ 936 937 nm_i = netmap_idx_n2k(kring, kring->nr_hwcur); 938 pkt_info_zero(&pi); 939 pi.ipi_segs = txq->ift_segs; 940 pi.ipi_qsidx = kring->ring_id; 941 if (nm_i != head) { /* we have new packets to send */ 942 nic_i = netmap_idx_k2n(kring, nm_i); 943 944 __builtin_prefetch(&ring->slot[nm_i]); 945 __builtin_prefetch(&txq->ift_sds.ifsd_m[nic_i]); 946 if (txq->ift_sds.ifsd_map) 947 __builtin_prefetch(&txq->ift_sds.ifsd_map[nic_i]); 948 949 for (n = 0; nm_i != head; n++) { 950 struct netmap_slot *slot = &ring->slot[nm_i]; 951 u_int len = slot->len; 952 uint64_t paddr; 953 void *addr = PNMB(na, slot, &paddr); 954 int flags = (slot->flags & NS_REPORT || 955 nic_i == 0 || nic_i == report_frequency) ? 956 IPI_TX_INTR : 0; 957 958 /* device-specific */ 959 pi.ipi_len = len; 960 pi.ipi_segs[0].ds_addr = paddr; 961 pi.ipi_segs[0].ds_len = len; 962 pi.ipi_nsegs = 1; 963 pi.ipi_ndescs = 0; 964 pi.ipi_pidx = nic_i; 965 pi.ipi_flags = flags; 966 967 /* Fill the slot in the NIC ring. */ 968 ctx->isc_txd_encap(ctx->ifc_softc, &pi); 969 970 /* prefetch for next round */ 971 __builtin_prefetch(&ring->slot[nm_i + 1]); 972 __builtin_prefetch(&txq->ift_sds.ifsd_m[nic_i + 1]); 973 if (txq->ift_sds.ifsd_map) { 974 __builtin_prefetch(&txq->ift_sds.ifsd_map[nic_i + 1]); 975 976 NM_CHECK_ADDR_LEN(na, addr, len); 977 978 if (slot->flags & NS_BUF_CHANGED) { 979 /* buffer has changed, reload map */ 980 netmap_reload_map(na, txq->ift_desc_tag, txq->ift_sds.ifsd_map[nic_i], addr); 981 } 982 /* make sure changes to the buffer are synced */ 983 bus_dmamap_sync(txq->ift_ifdi->idi_tag, txq->ift_sds.ifsd_map[nic_i], 984 BUS_DMASYNC_PREWRITE); 985 } 986 slot->flags &= ~(NS_REPORT | NS_BUF_CHANGED); 987 nm_i = nm_next(nm_i, lim); 988 nic_i = nm_next(nic_i, lim); 989 } 990 kring->nr_hwcur = head; 991 992 /* synchronize the NIC ring */ 993 if (txq->ift_sds.ifsd_map) 994 bus_dmamap_sync(txq->ift_desc_tag, txq->ift_ifdi->idi_map, 995 BUS_DMASYNC_PREREAD | BUS_DMASYNC_PREWRITE); 996 997 /* (re)start the tx unit up to slot nic_i (excluded) */ 998 ctx->isc_txd_flush(ctx->ifc_softc, txq->ift_id, nic_i); 999 } 1000 1001 /* 1002 * Second part: reclaim buffers for completed transmissions. 1003 */ 1004 if (iflib_tx_credits_update(ctx, txq)) { 1005 /* some tx completed, increment avail */ 1006 nic_i = txq->ift_cidx_processed; 1007 kring->nr_hwtail = nm_prev(netmap_idx_n2k(kring, nic_i), lim); 1008 } 1009 return (0); 1010 } 1011 1012 /* 1013 * Reconcile kernel and user view of the receive ring. 1014 * Same as for the txsync, this routine must be efficient. 1015 * The caller guarantees a single invocations, but races against 1016 * the rest of the driver should be handled here. 1017 * 1018 * On call, kring->rhead is the first packet that userspace wants 1019 * to keep, and kring->rcur is the wakeup point. 1020 * The kernel has previously reported packets up to kring->rtail. 1021 * 1022 * If (flags & NAF_FORCE_READ) also check for incoming packets irrespective 1023 * of whether or not we received an interrupt. 1024 */ 1025 static int 1026 iflib_netmap_rxsync(struct netmap_kring *kring, int flags) 1027 { 1028 struct netmap_adapter *na = kring->na; 1029 struct netmap_ring *ring = kring->ring; 1030 uint32_t nm_i; /* index into the netmap ring */ 1031 uint32_t nic_i; /* index into the NIC ring */ 1032 u_int i, n; 1033 u_int const lim = kring->nkr_num_slots - 1; 1034 u_int const head = netmap_idx_n2k(kring, kring->rhead); 1035 int force_update = (flags & NAF_FORCE_READ) || kring->nr_kflags & NKR_PENDINTR; 1036 struct if_rxd_info ri; 1037 1038 struct ifnet *ifp = na->ifp; 1039 if_ctx_t ctx = ifp->if_softc; 1040 iflib_rxq_t rxq = &ctx->ifc_rxqs[kring->ring_id]; 1041 iflib_fl_t fl = rxq->ifr_fl; 1042 if (head > lim) 1043 return netmap_ring_reinit(kring); 1044 1045 /* XXX check sync modes */ 1046 for (i = 0, fl = rxq->ifr_fl; i < rxq->ifr_nfl; i++, fl++) { 1047 if (fl->ifl_sds.ifsd_map == NULL) 1048 continue; 1049 bus_dmamap_sync(rxq->ifr_fl[i].ifl_desc_tag, fl->ifl_ifdi->idi_map, 1050 BUS_DMASYNC_POSTREAD | BUS_DMASYNC_POSTWRITE); 1051 } 1052 /* 1053 * First part: import newly received packets. 1054 * 1055 * nm_i is the index of the next free slot in the netmap ring, 1056 * nic_i is the index of the next received packet in the NIC ring, 1057 * and they may differ in case if_init() has been called while 1058 * in netmap mode. For the receive ring we have 1059 * 1060 * nic_i = rxr->next_check; 1061 * nm_i = kring->nr_hwtail (previous) 1062 * and 1063 * nm_i == (nic_i + kring->nkr_hwofs) % ring_size 1064 * 1065 * rxr->next_check is set to 0 on a ring reinit 1066 */ 1067 if (netmap_no_pendintr || force_update) { 1068 int crclen = iflib_crcstrip ? 0 : 4; 1069 int error, avail; 1070 uint16_t slot_flags = kring->nkr_slot_flags; 1071 1072 for (i = 0; i < rxq->ifr_nfl; i++) { 1073 fl = &rxq->ifr_fl[i]; 1074 nic_i = fl->ifl_cidx; 1075 nm_i = netmap_idx_n2k(kring, nic_i); 1076 avail = iflib_rxd_avail(ctx, rxq, nic_i, USHRT_MAX); 1077 for (n = 0; avail > 0; n++, avail--) { 1078 rxd_info_zero(&ri); 1079 ri.iri_frags = rxq->ifr_frags; 1080 ri.iri_qsidx = kring->ring_id; 1081 ri.iri_ifp = ctx->ifc_ifp; 1082 ri.iri_cidx = nic_i; 1083 1084 error = ctx->isc_rxd_pkt_get(ctx->ifc_softc, &ri); 1085 ring->slot[nm_i].len = error ? 0 : ri.iri_len - crclen; 1086 ring->slot[nm_i].flags = slot_flags; 1087 if (fl->ifl_sds.ifsd_map) 1088 bus_dmamap_sync(fl->ifl_ifdi->idi_tag, 1089 fl->ifl_sds.ifsd_map[nic_i], BUS_DMASYNC_POSTREAD); 1090 nm_i = nm_next(nm_i, lim); 1091 nic_i = nm_next(nic_i, lim); 1092 } 1093 if (n) { /* update the state variables */ 1094 if (netmap_no_pendintr && !force_update) { 1095 /* diagnostics */ 1096 iflib_rx_miss ++; 1097 iflib_rx_miss_bufs += n; 1098 } 1099 fl->ifl_cidx = nic_i; 1100 kring->nr_hwtail = netmap_idx_k2n(kring, nm_i); 1101 } 1102 kring->nr_kflags &= ~NKR_PENDINTR; 1103 } 1104 } 1105 /* 1106 * Second part: skip past packets that userspace has released. 1107 * (kring->nr_hwcur to head excluded), 1108 * and make the buffers available for reception. 1109 * As usual nm_i is the index in the netmap ring, 1110 * nic_i is the index in the NIC ring, and 1111 * nm_i == (nic_i + kring->nkr_hwofs) % ring_size 1112 */ 1113 /* XXX not sure how this will work with multiple free lists */ 1114 nm_i = netmap_idx_n2k(kring, kring->nr_hwcur); 1115 1116 return (netmap_fl_refill(rxq, kring, nm_i, false)); 1117 } 1118 1119 static void 1120 iflib_netmap_intr(struct netmap_adapter *na, int onoff) 1121 { 1122 struct ifnet *ifp = na->ifp; 1123 if_ctx_t ctx = ifp->if_softc; 1124 1125 CTX_LOCK(ctx); 1126 if (onoff) { 1127 IFDI_INTR_ENABLE(ctx); 1128 } else { 1129 IFDI_INTR_DISABLE(ctx); 1130 } 1131 CTX_UNLOCK(ctx); 1132 } 1133 1134 1135 static int 1136 iflib_netmap_attach(if_ctx_t ctx) 1137 { 1138 struct netmap_adapter na; 1139 if_softc_ctx_t scctx = &ctx->ifc_softc_ctx; 1140 1141 bzero(&na, sizeof(na)); 1142 1143 na.ifp = ctx->ifc_ifp; 1144 na.na_flags = NAF_BDG_MAYSLEEP; 1145 MPASS(ctx->ifc_softc_ctx.isc_ntxqsets); 1146 MPASS(ctx->ifc_softc_ctx.isc_nrxqsets); 1147 1148 na.num_tx_desc = scctx->isc_ntxd[0]; 1149 na.num_rx_desc = scctx->isc_nrxd[0]; 1150 na.nm_txsync = iflib_netmap_txsync; 1151 na.nm_rxsync = iflib_netmap_rxsync; 1152 na.nm_register = iflib_netmap_register; 1153 na.nm_intr = iflib_netmap_intr; 1154 na.num_tx_rings = ctx->ifc_softc_ctx.isc_ntxqsets; 1155 na.num_rx_rings = ctx->ifc_softc_ctx.isc_nrxqsets; 1156 return (netmap_attach(&na)); 1157 } 1158 1159 static void 1160 iflib_netmap_txq_init(if_ctx_t ctx, iflib_txq_t txq) 1161 { 1162 struct netmap_adapter *na = NA(ctx->ifc_ifp); 1163 struct netmap_slot *slot; 1164 1165 slot = netmap_reset(na, NR_TX, txq->ift_id, 0); 1166 if (slot == NULL) 1167 return; 1168 if (txq->ift_sds.ifsd_map == NULL) 1169 return; 1170 1171 for (int i = 0; i < ctx->ifc_softc_ctx.isc_ntxd[0]; i++) { 1172 1173 /* 1174 * In netmap mode, set the map for the packet buffer. 1175 * NOTE: Some drivers (not this one) also need to set 1176 * the physical buffer address in the NIC ring. 1177 * netmap_idx_n2k() maps a nic index, i, into the corresponding 1178 * netmap slot index, si 1179 */ 1180 int si = netmap_idx_n2k(&na->tx_rings[txq->ift_id], i); 1181 netmap_load_map(na, txq->ift_desc_tag, txq->ift_sds.ifsd_map[i], NMB(na, slot + si)); 1182 } 1183 } 1184 1185 static void 1186 iflib_netmap_rxq_init(if_ctx_t ctx, iflib_rxq_t rxq) 1187 { 1188 struct netmap_adapter *na = NA(ctx->ifc_ifp); 1189 struct netmap_kring *kring = &na->rx_rings[rxq->ifr_id]; 1190 struct netmap_slot *slot; 1191 uint32_t nm_i; 1192 1193 slot = netmap_reset(na, NR_RX, rxq->ifr_id, 0); 1194 if (slot == NULL) 1195 return; 1196 nm_i = netmap_idx_n2k(kring, 0); 1197 netmap_fl_refill(rxq, kring, nm_i, true); 1198 } 1199 1200 #define iflib_netmap_detach(ifp) netmap_detach(ifp) 1201 1202 #else 1203 #define iflib_netmap_txq_init(ctx, txq) 1204 #define iflib_netmap_rxq_init(ctx, rxq) 1205 #define iflib_netmap_detach(ifp) 1206 1207 #define iflib_netmap_attach(ctx) (0) 1208 #define netmap_rx_irq(ifp, qid, budget) (0) 1209 #define netmap_tx_irq(ifp, qid) do {} while (0) 1210 1211 #endif 1212 1213 #if defined(__i386__) || defined(__amd64__) 1214 static __inline void 1215 prefetch(void *x) 1216 { 1217 __asm volatile("prefetcht0 %0" :: "m" (*(unsigned long *)x)); 1218 } 1219 static __inline void 1220 prefetch2cachelines(void *x) 1221 { 1222 __asm volatile("prefetcht0 %0" :: "m" (*(unsigned long *)x)); 1223 #if (CACHE_LINE_SIZE < 128) 1224 __asm volatile("prefetcht0 %0" :: "m" (*(((unsigned long *)x)+CACHE_LINE_SIZE/(sizeof(unsigned long))))); 1225 #endif 1226 } 1227 #else 1228 #define prefetch(x) 1229 #define prefetch2cachelines(x) 1230 #endif 1231 1232 static void 1233 iru_init(if_rxd_update_t iru, iflib_rxq_t rxq, uint8_t flid) 1234 { 1235 iflib_fl_t fl; 1236 1237 fl = &rxq->ifr_fl[flid]; 1238 iru->iru_paddrs = fl->ifl_bus_addrs; 1239 iru->iru_vaddrs = &fl->ifl_vm_addrs[0]; 1240 iru->iru_idxs = fl->ifl_rxd_idxs; 1241 iru->iru_qsidx = rxq->ifr_id; 1242 iru->iru_buf_size = fl->ifl_buf_size; 1243 iru->iru_flidx = fl->ifl_id; 1244 } 1245 1246 static void 1247 _iflib_dmamap_cb(void *arg, bus_dma_segment_t *segs, int nseg, int err) 1248 { 1249 if (err) 1250 return; 1251 *(bus_addr_t *) arg = segs[0].ds_addr; 1252 } 1253 1254 int 1255 iflib_dma_alloc(if_ctx_t ctx, int size, iflib_dma_info_t dma, int mapflags) 1256 { 1257 int err; 1258 if_shared_ctx_t sctx = ctx->ifc_sctx; 1259 device_t dev = ctx->ifc_dev; 1260 1261 KASSERT(sctx->isc_q_align != 0, ("alignment value not initialized")); 1262 1263 err = bus_dma_tag_create(bus_get_dma_tag(dev), /* parent */ 1264 sctx->isc_q_align, 0, /* alignment, bounds */ 1265 BUS_SPACE_MAXADDR, /* lowaddr */ 1266 BUS_SPACE_MAXADDR, /* highaddr */ 1267 NULL, NULL, /* filter, filterarg */ 1268 size, /* maxsize */ 1269 1, /* nsegments */ 1270 size, /* maxsegsize */ 1271 BUS_DMA_ALLOCNOW, /* flags */ 1272 NULL, /* lockfunc */ 1273 NULL, /* lockarg */ 1274 &dma->idi_tag); 1275 if (err) { 1276 device_printf(dev, 1277 "%s: bus_dma_tag_create failed: %d\n", 1278 __func__, err); 1279 goto fail_0; 1280 } 1281 1282 err = bus_dmamem_alloc(dma->idi_tag, (void**) &dma->idi_vaddr, 1283 BUS_DMA_NOWAIT | BUS_DMA_COHERENT | BUS_DMA_ZERO, &dma->idi_map); 1284 if (err) { 1285 device_printf(dev, 1286 "%s: bus_dmamem_alloc(%ju) failed: %d\n", 1287 __func__, (uintmax_t)size, err); 1288 goto fail_1; 1289 } 1290 1291 dma->idi_paddr = IF_BAD_DMA; 1292 err = bus_dmamap_load(dma->idi_tag, dma->idi_map, dma->idi_vaddr, 1293 size, _iflib_dmamap_cb, &dma->idi_paddr, mapflags | BUS_DMA_NOWAIT); 1294 if (err || dma->idi_paddr == IF_BAD_DMA) { 1295 device_printf(dev, 1296 "%s: bus_dmamap_load failed: %d\n", 1297 __func__, err); 1298 goto fail_2; 1299 } 1300 1301 dma->idi_size = size; 1302 return (0); 1303 1304 fail_2: 1305 bus_dmamem_free(dma->idi_tag, dma->idi_vaddr, dma->idi_map); 1306 fail_1: 1307 bus_dma_tag_destroy(dma->idi_tag); 1308 fail_0: 1309 dma->idi_tag = NULL; 1310 1311 return (err); 1312 } 1313 1314 int 1315 iflib_dma_alloc_multi(if_ctx_t ctx, int *sizes, iflib_dma_info_t *dmalist, int mapflags, int count) 1316 { 1317 int i, err; 1318 iflib_dma_info_t *dmaiter; 1319 1320 dmaiter = dmalist; 1321 for (i = 0; i < count; i++, dmaiter++) { 1322 if ((err = iflib_dma_alloc(ctx, sizes[i], *dmaiter, mapflags)) != 0) 1323 break; 1324 } 1325 if (err) 1326 iflib_dma_free_multi(dmalist, i); 1327 return (err); 1328 } 1329 1330 void 1331 iflib_dma_free(iflib_dma_info_t dma) 1332 { 1333 if (dma->idi_tag == NULL) 1334 return; 1335 if (dma->idi_paddr != IF_BAD_DMA) { 1336 bus_dmamap_sync(dma->idi_tag, dma->idi_map, 1337 BUS_DMASYNC_POSTREAD | BUS_DMASYNC_POSTWRITE); 1338 bus_dmamap_unload(dma->idi_tag, dma->idi_map); 1339 dma->idi_paddr = IF_BAD_DMA; 1340 } 1341 if (dma->idi_vaddr != NULL) { 1342 bus_dmamem_free(dma->idi_tag, dma->idi_vaddr, dma->idi_map); 1343 dma->idi_vaddr = NULL; 1344 } 1345 bus_dma_tag_destroy(dma->idi_tag); 1346 dma->idi_tag = NULL; 1347 } 1348 1349 void 1350 iflib_dma_free_multi(iflib_dma_info_t *dmalist, int count) 1351 { 1352 int i; 1353 iflib_dma_info_t *dmaiter = dmalist; 1354 1355 for (i = 0; i < count; i++, dmaiter++) 1356 iflib_dma_free(*dmaiter); 1357 } 1358 1359 #ifdef EARLY_AP_STARTUP 1360 static const int iflib_started = 1; 1361 #else 1362 /* 1363 * We used to abuse the smp_started flag to decide if the queues have been 1364 * fully initialized (by late taskqgroup_adjust() calls in a SYSINIT()). 1365 * That gave bad races, since the SYSINIT() runs strictly after smp_started 1366 * is set. Run a SYSINIT() strictly after that to just set a usable 1367 * completion flag. 1368 */ 1369 1370 static int iflib_started; 1371 1372 static void 1373 iflib_record_started(void *arg) 1374 { 1375 iflib_started = 1; 1376 } 1377 1378 SYSINIT(iflib_record_started, SI_SUB_SMP + 1, SI_ORDER_FIRST, 1379 iflib_record_started, NULL); 1380 #endif 1381 1382 static int 1383 iflib_fast_intr(void *arg) 1384 { 1385 iflib_filter_info_t info = arg; 1386 struct grouptask *gtask = info->ifi_task; 1387 if (!iflib_started) 1388 return (FILTER_HANDLED); 1389 1390 DBG_COUNTER_INC(fast_intrs); 1391 if (info->ifi_filter != NULL && info->ifi_filter(info->ifi_filter_arg) == FILTER_HANDLED) 1392 return (FILTER_HANDLED); 1393 1394 GROUPTASK_ENQUEUE(gtask); 1395 return (FILTER_HANDLED); 1396 } 1397 1398 static int 1399 iflib_fast_intr_rxtx(void *arg) 1400 { 1401 iflib_filter_info_t info = arg; 1402 struct grouptask *gtask = info->ifi_task; 1403 iflib_rxq_t rxq = (iflib_rxq_t)info->ifi_ctx; 1404 if_ctx_t ctx; 1405 int i, cidx; 1406 1407 if (!iflib_started) 1408 return (FILTER_HANDLED); 1409 1410 DBG_COUNTER_INC(fast_intrs); 1411 if (info->ifi_filter != NULL && info->ifi_filter(info->ifi_filter_arg) == FILTER_HANDLED) 1412 return (FILTER_HANDLED); 1413 1414 for (i = 0; i < rxq->ifr_ntxqirq; i++) { 1415 qidx_t txqid = rxq->ifr_txqid[i]; 1416 1417 ctx = rxq->ifr_ctx; 1418 1419 if (!ctx->isc_txd_credits_update(ctx->ifc_softc, txqid, false)) { 1420 IFDI_TX_QUEUE_INTR_ENABLE(ctx, txqid); 1421 continue; 1422 } 1423 GROUPTASK_ENQUEUE(&ctx->ifc_txqs[txqid].ift_task); 1424 } 1425 if (ctx->ifc_sctx->isc_flags & IFLIB_HAS_RXCQ) 1426 cidx = rxq->ifr_cq_cidx; 1427 else 1428 cidx = rxq->ifr_fl[0].ifl_cidx; 1429 if (iflib_rxd_avail(ctx, rxq, cidx, 1)) 1430 GROUPTASK_ENQUEUE(gtask); 1431 else 1432 IFDI_RX_QUEUE_INTR_ENABLE(ctx, rxq->ifr_id); 1433 return (FILTER_HANDLED); 1434 } 1435 1436 1437 static int 1438 iflib_fast_intr_ctx(void *arg) 1439 { 1440 iflib_filter_info_t info = arg; 1441 struct grouptask *gtask = info->ifi_task; 1442 1443 if (!iflib_started) 1444 return (FILTER_HANDLED); 1445 1446 DBG_COUNTER_INC(fast_intrs); 1447 if (info->ifi_filter != NULL && info->ifi_filter(info->ifi_filter_arg) == FILTER_HANDLED) 1448 return (FILTER_HANDLED); 1449 1450 GROUPTASK_ENQUEUE(gtask); 1451 return (FILTER_HANDLED); 1452 } 1453 1454 static int 1455 _iflib_irq_alloc(if_ctx_t ctx, if_irq_t irq, int rid, 1456 driver_filter_t filter, driver_intr_t handler, void *arg, 1457 char *name) 1458 { 1459 int rc, flags; 1460 struct resource *res; 1461 void *tag = NULL; 1462 device_t dev = ctx->ifc_dev; 1463 1464 flags = RF_ACTIVE; 1465 if (ctx->ifc_flags & IFC_LEGACY) 1466 flags |= RF_SHAREABLE; 1467 MPASS(rid < 512); 1468 irq->ii_rid = rid; 1469 res = bus_alloc_resource_any(dev, SYS_RES_IRQ, &irq->ii_rid, flags); 1470 if (res == NULL) { 1471 device_printf(dev, 1472 "failed to allocate IRQ for rid %d, name %s.\n", rid, name); 1473 return (ENOMEM); 1474 } 1475 irq->ii_res = res; 1476 KASSERT(filter == NULL || handler == NULL, ("filter and handler can't both be non-NULL")); 1477 rc = bus_setup_intr(dev, res, INTR_MPSAFE | INTR_TYPE_NET, 1478 filter, handler, arg, &tag); 1479 if (rc != 0) { 1480 device_printf(dev, 1481 "failed to setup interrupt for rid %d, name %s: %d\n", 1482 rid, name ? name : "unknown", rc); 1483 return (rc); 1484 } else if (name) 1485 bus_describe_intr(dev, res, tag, "%s", name); 1486 1487 irq->ii_tag = tag; 1488 return (0); 1489 } 1490 1491 1492 /********************************************************************* 1493 * 1494 * Allocate memory for tx_buffer structures. The tx_buffer stores all 1495 * the information needed to transmit a packet on the wire. This is 1496 * called only once at attach, setup is done every reset. 1497 * 1498 **********************************************************************/ 1499 1500 static int 1501 iflib_txsd_alloc(iflib_txq_t txq) 1502 { 1503 if_ctx_t ctx = txq->ift_ctx; 1504 if_shared_ctx_t sctx = ctx->ifc_sctx; 1505 if_softc_ctx_t scctx = &ctx->ifc_softc_ctx; 1506 device_t dev = ctx->ifc_dev; 1507 int err, nsegments, ntsosegments; 1508 1509 nsegments = scctx->isc_tx_nsegments; 1510 ntsosegments = scctx->isc_tx_tso_segments_max; 1511 MPASS(scctx->isc_ntxd[0] > 0); 1512 MPASS(scctx->isc_ntxd[txq->ift_br_offset] > 0); 1513 MPASS(nsegments > 0); 1514 MPASS(ntsosegments > 0); 1515 /* 1516 * Setup DMA descriptor areas. 1517 */ 1518 if ((err = bus_dma_tag_create(bus_get_dma_tag(dev), 1519 1, 0, /* alignment, bounds */ 1520 BUS_SPACE_MAXADDR, /* lowaddr */ 1521 BUS_SPACE_MAXADDR, /* highaddr */ 1522 NULL, NULL, /* filter, filterarg */ 1523 sctx->isc_tx_maxsize, /* maxsize */ 1524 nsegments, /* nsegments */ 1525 sctx->isc_tx_maxsegsize, /* maxsegsize */ 1526 0, /* flags */ 1527 NULL, /* lockfunc */ 1528 NULL, /* lockfuncarg */ 1529 &txq->ift_desc_tag))) { 1530 device_printf(dev,"Unable to allocate TX DMA tag: %d\n", err); 1531 device_printf(dev,"maxsize: %ju nsegments: %d maxsegsize: %ju\n", 1532 (uintmax_t)sctx->isc_tx_maxsize, nsegments, (uintmax_t)sctx->isc_tx_maxsegsize); 1533 goto fail; 1534 } 1535 if ((err = bus_dma_tag_create(bus_get_dma_tag(dev), 1536 1, 0, /* alignment, bounds */ 1537 BUS_SPACE_MAXADDR, /* lowaddr */ 1538 BUS_SPACE_MAXADDR, /* highaddr */ 1539 NULL, NULL, /* filter, filterarg */ 1540 scctx->isc_tx_tso_size_max, /* maxsize */ 1541 ntsosegments, /* nsegments */ 1542 scctx->isc_tx_tso_segsize_max, /* maxsegsize */ 1543 0, /* flags */ 1544 NULL, /* lockfunc */ 1545 NULL, /* lockfuncarg */ 1546 &txq->ift_tso_desc_tag))) { 1547 device_printf(dev,"Unable to allocate TX TSO DMA tag: %d\n", err); 1548 1549 goto fail; 1550 } 1551 if (!(txq->ift_sds.ifsd_flags = 1552 (uint8_t *) malloc(sizeof(uint8_t) * 1553 scctx->isc_ntxd[txq->ift_br_offset], M_IFLIB, M_NOWAIT | M_ZERO))) { 1554 device_printf(dev, "Unable to allocate tx_buffer memory\n"); 1555 err = ENOMEM; 1556 goto fail; 1557 } 1558 if (!(txq->ift_sds.ifsd_m = 1559 (struct mbuf **) malloc(sizeof(struct mbuf *) * 1560 scctx->isc_ntxd[txq->ift_br_offset], M_IFLIB, M_NOWAIT | M_ZERO))) { 1561 device_printf(dev, "Unable to allocate tx_buffer memory\n"); 1562 err = ENOMEM; 1563 goto fail; 1564 } 1565 1566 /* Create the descriptor buffer dma maps */ 1567 #if defined(ACPI_DMAR) || (! (defined(__i386__) || defined(__amd64__))) 1568 if ((ctx->ifc_flags & IFC_DMAR) == 0) 1569 return (0); 1570 1571 if (!(txq->ift_sds.ifsd_map = 1572 (bus_dmamap_t *) malloc(sizeof(bus_dmamap_t) * scctx->isc_ntxd[txq->ift_br_offset], M_IFLIB, M_NOWAIT | M_ZERO))) { 1573 device_printf(dev, "Unable to allocate tx_buffer map memory\n"); 1574 err = ENOMEM; 1575 goto fail; 1576 } 1577 1578 for (int i = 0; i < scctx->isc_ntxd[txq->ift_br_offset]; i++) { 1579 err = bus_dmamap_create(txq->ift_desc_tag, 0, &txq->ift_sds.ifsd_map[i]); 1580 if (err != 0) { 1581 device_printf(dev, "Unable to create TX DMA map\n"); 1582 goto fail; 1583 } 1584 } 1585 #endif 1586 return (0); 1587 fail: 1588 /* We free all, it handles case where we are in the middle */ 1589 iflib_tx_structures_free(ctx); 1590 return (err); 1591 } 1592 1593 static void 1594 iflib_txsd_destroy(if_ctx_t ctx, iflib_txq_t txq, int i) 1595 { 1596 bus_dmamap_t map; 1597 1598 map = NULL; 1599 if (txq->ift_sds.ifsd_map != NULL) 1600 map = txq->ift_sds.ifsd_map[i]; 1601 if (map != NULL) { 1602 bus_dmamap_unload(txq->ift_desc_tag, map); 1603 bus_dmamap_destroy(txq->ift_desc_tag, map); 1604 txq->ift_sds.ifsd_map[i] = NULL; 1605 } 1606 } 1607 1608 static void 1609 iflib_txq_destroy(iflib_txq_t txq) 1610 { 1611 if_ctx_t ctx = txq->ift_ctx; 1612 1613 for (int i = 0; i < txq->ift_size; i++) 1614 iflib_txsd_destroy(ctx, txq, i); 1615 if (txq->ift_sds.ifsd_map != NULL) { 1616 free(txq->ift_sds.ifsd_map, M_IFLIB); 1617 txq->ift_sds.ifsd_map = NULL; 1618 } 1619 if (txq->ift_sds.ifsd_m != NULL) { 1620 free(txq->ift_sds.ifsd_m, M_IFLIB); 1621 txq->ift_sds.ifsd_m = NULL; 1622 } 1623 if (txq->ift_sds.ifsd_flags != NULL) { 1624 free(txq->ift_sds.ifsd_flags, M_IFLIB); 1625 txq->ift_sds.ifsd_flags = NULL; 1626 } 1627 if (txq->ift_desc_tag != NULL) { 1628 bus_dma_tag_destroy(txq->ift_desc_tag); 1629 txq->ift_desc_tag = NULL; 1630 } 1631 if (txq->ift_tso_desc_tag != NULL) { 1632 bus_dma_tag_destroy(txq->ift_tso_desc_tag); 1633 txq->ift_tso_desc_tag = NULL; 1634 } 1635 } 1636 1637 static void 1638 iflib_txsd_free(if_ctx_t ctx, iflib_txq_t txq, int i) 1639 { 1640 struct mbuf **mp; 1641 1642 mp = &txq->ift_sds.ifsd_m[i]; 1643 if (*mp == NULL) 1644 return; 1645 1646 if (txq->ift_sds.ifsd_map != NULL) { 1647 bus_dmamap_sync(txq->ift_desc_tag, 1648 txq->ift_sds.ifsd_map[i], 1649 BUS_DMASYNC_POSTWRITE); 1650 bus_dmamap_unload(txq->ift_desc_tag, 1651 txq->ift_sds.ifsd_map[i]); 1652 } 1653 m_free(*mp); 1654 DBG_COUNTER_INC(tx_frees); 1655 *mp = NULL; 1656 } 1657 1658 static int 1659 iflib_txq_setup(iflib_txq_t txq) 1660 { 1661 if_ctx_t ctx = txq->ift_ctx; 1662 if_softc_ctx_t scctx = &ctx->ifc_softc_ctx; 1663 iflib_dma_info_t di; 1664 int i; 1665 1666 /* Set number of descriptors available */ 1667 txq->ift_qstatus = IFLIB_QUEUE_IDLE; 1668 /* XXX make configurable */ 1669 txq->ift_update_freq = IFLIB_DEFAULT_TX_UPDATE_FREQ; 1670 1671 /* Reset indices */ 1672 txq->ift_cidx_processed = 0; 1673 txq->ift_pidx = txq->ift_cidx = txq->ift_npending = 0; 1674 txq->ift_size = scctx->isc_ntxd[txq->ift_br_offset]; 1675 1676 for (i = 0, di = txq->ift_ifdi; i < ctx->ifc_nhwtxqs; i++, di++) 1677 bzero((void *)di->idi_vaddr, di->idi_size); 1678 1679 IFDI_TXQ_SETUP(ctx, txq->ift_id); 1680 for (i = 0, di = txq->ift_ifdi; i < ctx->ifc_nhwtxqs; i++, di++) 1681 bus_dmamap_sync(di->idi_tag, di->idi_map, 1682 BUS_DMASYNC_PREREAD | BUS_DMASYNC_PREWRITE); 1683 return (0); 1684 } 1685 1686 /********************************************************************* 1687 * 1688 * Allocate memory for rx_buffer structures. Since we use one 1689 * rx_buffer per received packet, the maximum number of rx_buffer's 1690 * that we'll need is equal to the number of receive descriptors 1691 * that we've allocated. 1692 * 1693 **********************************************************************/ 1694 static int 1695 iflib_rxsd_alloc(iflib_rxq_t rxq) 1696 { 1697 if_ctx_t ctx = rxq->ifr_ctx; 1698 if_shared_ctx_t sctx = ctx->ifc_sctx; 1699 if_softc_ctx_t scctx = &ctx->ifc_softc_ctx; 1700 device_t dev = ctx->ifc_dev; 1701 iflib_fl_t fl; 1702 int err; 1703 1704 MPASS(scctx->isc_nrxd[0] > 0); 1705 MPASS(scctx->isc_nrxd[rxq->ifr_fl_offset] > 0); 1706 1707 fl = rxq->ifr_fl; 1708 for (int i = 0; i < rxq->ifr_nfl; i++, fl++) { 1709 fl->ifl_size = scctx->isc_nrxd[rxq->ifr_fl_offset]; /* this isn't necessarily the same */ 1710 err = bus_dma_tag_create(bus_get_dma_tag(dev), /* parent */ 1711 1, 0, /* alignment, bounds */ 1712 BUS_SPACE_MAXADDR, /* lowaddr */ 1713 BUS_SPACE_MAXADDR, /* highaddr */ 1714 NULL, NULL, /* filter, filterarg */ 1715 sctx->isc_rx_maxsize, /* maxsize */ 1716 sctx->isc_rx_nsegments, /* nsegments */ 1717 sctx->isc_rx_maxsegsize, /* maxsegsize */ 1718 0, /* flags */ 1719 NULL, /* lockfunc */ 1720 NULL, /* lockarg */ 1721 &fl->ifl_desc_tag); 1722 if (err) { 1723 device_printf(dev, "%s: bus_dma_tag_create failed %d\n", 1724 __func__, err); 1725 goto fail; 1726 } 1727 if (!(fl->ifl_sds.ifsd_flags = 1728 (uint8_t *) malloc(sizeof(uint8_t) * 1729 scctx->isc_nrxd[rxq->ifr_fl_offset], M_IFLIB, M_NOWAIT | M_ZERO))) { 1730 device_printf(dev, "Unable to allocate tx_buffer memory\n"); 1731 err = ENOMEM; 1732 goto fail; 1733 } 1734 if (!(fl->ifl_sds.ifsd_m = 1735 (struct mbuf **) malloc(sizeof(struct mbuf *) * 1736 scctx->isc_nrxd[rxq->ifr_fl_offset], M_IFLIB, M_NOWAIT | M_ZERO))) { 1737 device_printf(dev, "Unable to allocate tx_buffer memory\n"); 1738 err = ENOMEM; 1739 goto fail; 1740 } 1741 if (!(fl->ifl_sds.ifsd_cl = 1742 (caddr_t *) malloc(sizeof(caddr_t) * 1743 scctx->isc_nrxd[rxq->ifr_fl_offset], M_IFLIB, M_NOWAIT | M_ZERO))) { 1744 device_printf(dev, "Unable to allocate tx_buffer memory\n"); 1745 err = ENOMEM; 1746 goto fail; 1747 } 1748 1749 /* Create the descriptor buffer dma maps */ 1750 #if defined(ACPI_DMAR) || (! (defined(__i386__) || defined(__amd64__))) 1751 if ((ctx->ifc_flags & IFC_DMAR) == 0) 1752 continue; 1753 1754 if (!(fl->ifl_sds.ifsd_map = 1755 (bus_dmamap_t *) malloc(sizeof(bus_dmamap_t) * scctx->isc_nrxd[rxq->ifr_fl_offset], M_IFLIB, M_NOWAIT | M_ZERO))) { 1756 device_printf(dev, "Unable to allocate tx_buffer map memory\n"); 1757 err = ENOMEM; 1758 goto fail; 1759 } 1760 1761 for (int i = 0; i < scctx->isc_nrxd[rxq->ifr_fl_offset]; i++) { 1762 err = bus_dmamap_create(fl->ifl_desc_tag, 0, &fl->ifl_sds.ifsd_map[i]); 1763 if (err != 0) { 1764 device_printf(dev, "Unable to create RX buffer DMA map\n"); 1765 goto fail; 1766 } 1767 } 1768 #endif 1769 } 1770 return (0); 1771 1772 fail: 1773 iflib_rx_structures_free(ctx); 1774 return (err); 1775 } 1776 1777 1778 /* 1779 * Internal service routines 1780 */ 1781 1782 struct rxq_refill_cb_arg { 1783 int error; 1784 bus_dma_segment_t seg; 1785 int nseg; 1786 }; 1787 1788 static void 1789 _rxq_refill_cb(void *arg, bus_dma_segment_t *segs, int nseg, int error) 1790 { 1791 struct rxq_refill_cb_arg *cb_arg = arg; 1792 1793 cb_arg->error = error; 1794 cb_arg->seg = segs[0]; 1795 cb_arg->nseg = nseg; 1796 } 1797 1798 1799 #ifdef ACPI_DMAR 1800 #define IS_DMAR(ctx) (ctx->ifc_flags & IFC_DMAR) 1801 #else 1802 #define IS_DMAR(ctx) (0) 1803 #endif 1804 1805 /** 1806 * rxq_refill - refill an rxq free-buffer list 1807 * @ctx: the iflib context 1808 * @rxq: the free-list to refill 1809 * @n: the number of new buffers to allocate 1810 * 1811 * (Re)populate an rxq free-buffer list with up to @n new packet buffers. 1812 * The caller must assure that @n does not exceed the queue's capacity. 1813 */ 1814 static void 1815 _iflib_fl_refill(if_ctx_t ctx, iflib_fl_t fl, int count) 1816 { 1817 struct mbuf *m; 1818 int idx, frag_idx = fl->ifl_fragidx; 1819 int pidx = fl->ifl_pidx; 1820 caddr_t cl, *sd_cl; 1821 struct mbuf **sd_m; 1822 uint8_t *sd_flags; 1823 struct if_rxd_update iru; 1824 bus_dmamap_t *sd_map; 1825 int n, i = 0; 1826 uint64_t bus_addr; 1827 int err; 1828 qidx_t credits; 1829 1830 sd_m = fl->ifl_sds.ifsd_m; 1831 sd_map = fl->ifl_sds.ifsd_map; 1832 sd_cl = fl->ifl_sds.ifsd_cl; 1833 sd_flags = fl->ifl_sds.ifsd_flags; 1834 idx = pidx; 1835 credits = fl->ifl_credits; 1836 1837 n = count; 1838 MPASS(n > 0); 1839 MPASS(credits + n <= fl->ifl_size); 1840 1841 if (pidx < fl->ifl_cidx) 1842 MPASS(pidx + n <= fl->ifl_cidx); 1843 if (pidx == fl->ifl_cidx && (credits < fl->ifl_size)) 1844 MPASS(fl->ifl_gen == 0); 1845 if (pidx > fl->ifl_cidx) 1846 MPASS(n <= fl->ifl_size - pidx + fl->ifl_cidx); 1847 1848 DBG_COUNTER_INC(fl_refills); 1849 if (n > 8) 1850 DBG_COUNTER_INC(fl_refills_large); 1851 iru_init(&iru, fl->ifl_rxq, fl->ifl_id); 1852 while (n--) { 1853 /* 1854 * We allocate an uninitialized mbuf + cluster, mbuf is 1855 * initialized after rx. 1856 * 1857 * If the cluster is still set then we know a minimum sized packet was received 1858 */ 1859 bit_ffc_at(fl->ifl_rx_bitmap, frag_idx, fl->ifl_size, &frag_idx); 1860 if ((frag_idx < 0) || (frag_idx >= fl->ifl_size)) 1861 bit_ffc(fl->ifl_rx_bitmap, fl->ifl_size, &frag_idx); 1862 if ((cl = sd_cl[frag_idx]) == NULL) { 1863 if ((cl = sd_cl[frag_idx] = m_cljget(NULL, M_NOWAIT, fl->ifl_buf_size)) == NULL) 1864 break; 1865 #if MEMORY_LOGGING 1866 fl->ifl_cl_enqueued++; 1867 #endif 1868 } 1869 if ((m = m_gethdr(M_NOWAIT, MT_NOINIT)) == NULL) { 1870 break; 1871 } 1872 #if MEMORY_LOGGING 1873 fl->ifl_m_enqueued++; 1874 #endif 1875 1876 DBG_COUNTER_INC(rx_allocs); 1877 #if defined(__i386__) || defined(__amd64__) 1878 if (!IS_DMAR(ctx)) { 1879 bus_addr = pmap_kextract((vm_offset_t)cl); 1880 } else 1881 #endif 1882 { 1883 struct rxq_refill_cb_arg cb_arg; 1884 iflib_rxq_t q; 1885 1886 cb_arg.error = 0; 1887 q = fl->ifl_rxq; 1888 MPASS(sd_map != NULL); 1889 MPASS(sd_map[frag_idx] != NULL); 1890 err = bus_dmamap_load(fl->ifl_desc_tag, sd_map[frag_idx], 1891 cl, fl->ifl_buf_size, _rxq_refill_cb, &cb_arg, 0); 1892 bus_dmamap_sync(fl->ifl_desc_tag, sd_map[frag_idx], 1893 BUS_DMASYNC_PREREAD); 1894 1895 if (err != 0 || cb_arg.error) { 1896 /* 1897 * !zone_pack ? 1898 */ 1899 if (fl->ifl_zone == zone_pack) 1900 uma_zfree(fl->ifl_zone, cl); 1901 m_free(m); 1902 n = 0; 1903 goto done; 1904 } 1905 bus_addr = cb_arg.seg.ds_addr; 1906 } 1907 bit_set(fl->ifl_rx_bitmap, frag_idx); 1908 sd_flags[frag_idx] |= RX_SW_DESC_INUSE; 1909 1910 MPASS(sd_m[frag_idx] == NULL); 1911 sd_cl[frag_idx] = cl; 1912 sd_m[frag_idx] = m; 1913 fl->ifl_rxd_idxs[i] = frag_idx; 1914 fl->ifl_bus_addrs[i] = bus_addr; 1915 fl->ifl_vm_addrs[i] = cl; 1916 credits++; 1917 i++; 1918 MPASS(credits <= fl->ifl_size); 1919 if (++idx == fl->ifl_size) { 1920 fl->ifl_gen = 1; 1921 idx = 0; 1922 } 1923 if (n == 0 || i == IFLIB_MAX_RX_REFRESH) { 1924 iru.iru_pidx = pidx; 1925 iru.iru_count = i; 1926 ctx->isc_rxd_refill(ctx->ifc_softc, &iru); 1927 i = 0; 1928 pidx = idx; 1929 fl->ifl_pidx = idx; 1930 fl->ifl_credits = credits; 1931 } 1932 1933 } 1934 done: 1935 if (i) { 1936 iru.iru_pidx = pidx; 1937 iru.iru_count = i; 1938 ctx->isc_rxd_refill(ctx->ifc_softc, &iru); 1939 fl->ifl_pidx = idx; 1940 fl->ifl_credits = credits; 1941 } 1942 DBG_COUNTER_INC(rxd_flush); 1943 if (fl->ifl_pidx == 0) 1944 pidx = fl->ifl_size - 1; 1945 else 1946 pidx = fl->ifl_pidx - 1; 1947 1948 if (sd_map) 1949 bus_dmamap_sync(fl->ifl_ifdi->idi_tag, fl->ifl_ifdi->idi_map, 1950 BUS_DMASYNC_PREREAD | BUS_DMASYNC_PREWRITE); 1951 ctx->isc_rxd_flush(ctx->ifc_softc, fl->ifl_rxq->ifr_id, fl->ifl_id, pidx); 1952 fl->ifl_fragidx = frag_idx; 1953 } 1954 1955 static __inline void 1956 __iflib_fl_refill_lt(if_ctx_t ctx, iflib_fl_t fl, int max) 1957 { 1958 /* we avoid allowing pidx to catch up with cidx as it confuses ixl */ 1959 int32_t reclaimable = fl->ifl_size - fl->ifl_credits - 1; 1960 #ifdef INVARIANTS 1961 int32_t delta = fl->ifl_size - get_inuse(fl->ifl_size, fl->ifl_cidx, fl->ifl_pidx, fl->ifl_gen) - 1; 1962 #endif 1963 1964 MPASS(fl->ifl_credits <= fl->ifl_size); 1965 MPASS(reclaimable == delta); 1966 1967 if (reclaimable > 0) 1968 _iflib_fl_refill(ctx, fl, min(max, reclaimable)); 1969 } 1970 1971 static void 1972 iflib_fl_bufs_free(iflib_fl_t fl) 1973 { 1974 iflib_dma_info_t idi = fl->ifl_ifdi; 1975 uint32_t i; 1976 1977 for (i = 0; i < fl->ifl_size; i++) { 1978 struct mbuf **sd_m = &fl->ifl_sds.ifsd_m[i]; 1979 uint8_t *sd_flags = &fl->ifl_sds.ifsd_flags[i]; 1980 caddr_t *sd_cl = &fl->ifl_sds.ifsd_cl[i]; 1981 1982 if (*sd_flags & RX_SW_DESC_INUSE) { 1983 if (fl->ifl_sds.ifsd_map != NULL) { 1984 bus_dmamap_t sd_map = fl->ifl_sds.ifsd_map[i]; 1985 bus_dmamap_unload(fl->ifl_desc_tag, sd_map); 1986 bus_dmamap_destroy(fl->ifl_desc_tag, sd_map); 1987 } 1988 if (*sd_m != NULL) { 1989 m_init(*sd_m, M_NOWAIT, MT_DATA, 0); 1990 uma_zfree(zone_mbuf, *sd_m); 1991 } 1992 if (*sd_cl != NULL) 1993 uma_zfree(fl->ifl_zone, *sd_cl); 1994 *sd_flags = 0; 1995 } else { 1996 MPASS(*sd_cl == NULL); 1997 MPASS(*sd_m == NULL); 1998 } 1999 #if MEMORY_LOGGING 2000 fl->ifl_m_dequeued++; 2001 fl->ifl_cl_dequeued++; 2002 #endif 2003 *sd_cl = NULL; 2004 *sd_m = NULL; 2005 } 2006 #ifdef INVARIANTS 2007 for (i = 0; i < fl->ifl_size; i++) { 2008 MPASS(fl->ifl_sds.ifsd_flags[i] == 0); 2009 MPASS(fl->ifl_sds.ifsd_cl[i] == NULL); 2010 MPASS(fl->ifl_sds.ifsd_m[i] == NULL); 2011 } 2012 #endif 2013 /* 2014 * Reset free list values 2015 */ 2016 fl->ifl_credits = fl->ifl_cidx = fl->ifl_pidx = fl->ifl_gen = fl->ifl_fragidx = 0; 2017 bzero(idi->idi_vaddr, idi->idi_size); 2018 } 2019 2020 /********************************************************************* 2021 * 2022 * Initialize a receive ring and its buffers. 2023 * 2024 **********************************************************************/ 2025 static int 2026 iflib_fl_setup(iflib_fl_t fl) 2027 { 2028 iflib_rxq_t rxq = fl->ifl_rxq; 2029 if_ctx_t ctx = rxq->ifr_ctx; 2030 if_softc_ctx_t sctx = &ctx->ifc_softc_ctx; 2031 2032 bit_nclear(fl->ifl_rx_bitmap, 0, fl->ifl_size - 1); 2033 /* 2034 ** Free current RX buffer structs and their mbufs 2035 */ 2036 iflib_fl_bufs_free(fl); 2037 /* Now replenish the mbufs */ 2038 MPASS(fl->ifl_credits == 0); 2039 /* 2040 * XXX don't set the max_frame_size to larger 2041 * than the hardware can handle 2042 */ 2043 if (sctx->isc_max_frame_size <= 2048) 2044 fl->ifl_buf_size = MCLBYTES; 2045 #ifndef CONTIGMALLOC_WORKS 2046 else 2047 fl->ifl_buf_size = MJUMPAGESIZE; 2048 #else 2049 else if (sctx->isc_max_frame_size <= 4096) 2050 fl->ifl_buf_size = MJUMPAGESIZE; 2051 else if (sctx->isc_max_frame_size <= 9216) 2052 fl->ifl_buf_size = MJUM9BYTES; 2053 else 2054 fl->ifl_buf_size = MJUM16BYTES; 2055 #endif 2056 if (fl->ifl_buf_size > ctx->ifc_max_fl_buf_size) 2057 ctx->ifc_max_fl_buf_size = fl->ifl_buf_size; 2058 fl->ifl_cltype = m_gettype(fl->ifl_buf_size); 2059 fl->ifl_zone = m_getzone(fl->ifl_buf_size); 2060 2061 2062 /* avoid pre-allocating zillions of clusters to an idle card 2063 * potentially speeding up attach 2064 */ 2065 _iflib_fl_refill(ctx, fl, min(128, fl->ifl_size)); 2066 MPASS(min(128, fl->ifl_size) == fl->ifl_credits); 2067 if (min(128, fl->ifl_size) != fl->ifl_credits) 2068 return (ENOBUFS); 2069 /* 2070 * handle failure 2071 */ 2072 MPASS(rxq != NULL); 2073 MPASS(fl->ifl_ifdi != NULL); 2074 bus_dmamap_sync(fl->ifl_ifdi->idi_tag, fl->ifl_ifdi->idi_map, 2075 BUS_DMASYNC_PREREAD | BUS_DMASYNC_PREWRITE); 2076 return (0); 2077 } 2078 2079 /********************************************************************* 2080 * 2081 * Free receive ring data structures 2082 * 2083 **********************************************************************/ 2084 static void 2085 iflib_rx_sds_free(iflib_rxq_t rxq) 2086 { 2087 iflib_fl_t fl; 2088 int i; 2089 2090 if (rxq->ifr_fl != NULL) { 2091 for (i = 0; i < rxq->ifr_nfl; i++) { 2092 fl = &rxq->ifr_fl[i]; 2093 if (fl->ifl_desc_tag != NULL) { 2094 bus_dma_tag_destroy(fl->ifl_desc_tag); 2095 fl->ifl_desc_tag = NULL; 2096 } 2097 free(fl->ifl_sds.ifsd_m, M_IFLIB); 2098 free(fl->ifl_sds.ifsd_cl, M_IFLIB); 2099 /* XXX destroy maps first */ 2100 free(fl->ifl_sds.ifsd_map, M_IFLIB); 2101 fl->ifl_sds.ifsd_m = NULL; 2102 fl->ifl_sds.ifsd_cl = NULL; 2103 fl->ifl_sds.ifsd_map = NULL; 2104 } 2105 free(rxq->ifr_fl, M_IFLIB); 2106 rxq->ifr_fl = NULL; 2107 rxq->ifr_cq_gen = rxq->ifr_cq_cidx = rxq->ifr_cq_pidx = 0; 2108 } 2109 } 2110 2111 /* 2112 * MI independent logic 2113 * 2114 */ 2115 static void 2116 iflib_timer(void *arg) 2117 { 2118 iflib_txq_t txq = arg; 2119 if_ctx_t ctx = txq->ift_ctx; 2120 if_softc_ctx_t sctx = &ctx->ifc_softc_ctx; 2121 2122 if (!(if_getdrvflags(ctx->ifc_ifp) & IFF_DRV_RUNNING)) 2123 return; 2124 /* 2125 ** Check on the state of the TX queue(s), this 2126 ** can be done without the lock because its RO 2127 ** and the HUNG state will be static if set. 2128 */ 2129 IFDI_TIMER(ctx, txq->ift_id); 2130 if ((txq->ift_qstatus == IFLIB_QUEUE_HUNG) && 2131 ((txq->ift_cleaned_prev == txq->ift_cleaned) || 2132 (sctx->isc_pause_frames == 0))) 2133 goto hung; 2134 2135 if (ifmp_ring_is_stalled(txq->ift_br)) 2136 txq->ift_qstatus = IFLIB_QUEUE_HUNG; 2137 txq->ift_cleaned_prev = txq->ift_cleaned; 2138 /* handle any laggards */ 2139 if (txq->ift_db_pending) 2140 GROUPTASK_ENQUEUE(&txq->ift_task); 2141 2142 sctx->isc_pause_frames = 0; 2143 if (if_getdrvflags(ctx->ifc_ifp) & IFF_DRV_RUNNING) 2144 callout_reset_on(&txq->ift_timer, hz/2, iflib_timer, txq, txq->ift_timer.c_cpu); 2145 return; 2146 hung: 2147 CTX_LOCK(ctx); 2148 if_setdrvflagbits(ctx->ifc_ifp, IFF_DRV_OACTIVE, IFF_DRV_RUNNING); 2149 device_printf(ctx->ifc_dev, "TX(%d) desc avail = %d, pidx = %d\n", 2150 txq->ift_id, TXQ_AVAIL(txq), txq->ift_pidx); 2151 2152 IFDI_WATCHDOG_RESET(ctx); 2153 ctx->ifc_watchdog_events++; 2154 2155 ctx->ifc_flags |= IFC_DO_RESET; 2156 iflib_admin_intr_deferred(ctx); 2157 CTX_UNLOCK(ctx); 2158 } 2159 2160 static void 2161 iflib_init_locked(if_ctx_t ctx) 2162 { 2163 if_softc_ctx_t sctx = &ctx->ifc_softc_ctx; 2164 if_softc_ctx_t scctx = &ctx->ifc_softc_ctx; 2165 if_t ifp = ctx->ifc_ifp; 2166 iflib_fl_t fl; 2167 iflib_txq_t txq; 2168 iflib_rxq_t rxq; 2169 int i, j, tx_ip_csum_flags, tx_ip6_csum_flags; 2170 2171 2172 if_setdrvflagbits(ifp, IFF_DRV_OACTIVE, IFF_DRV_RUNNING); 2173 IFDI_INTR_DISABLE(ctx); 2174 2175 tx_ip_csum_flags = scctx->isc_tx_csum_flags & (CSUM_IP | CSUM_TCP | CSUM_UDP | CSUM_SCTP); 2176 tx_ip6_csum_flags = scctx->isc_tx_csum_flags & (CSUM_IP6_TCP | CSUM_IP6_UDP | CSUM_IP6_SCTP); 2177 /* Set hardware offload abilities */ 2178 if_clearhwassist(ifp); 2179 if (if_getcapenable(ifp) & IFCAP_TXCSUM) 2180 if_sethwassistbits(ifp, tx_ip_csum_flags, 0); 2181 if (if_getcapenable(ifp) & IFCAP_TXCSUM_IPV6) 2182 if_sethwassistbits(ifp, tx_ip6_csum_flags, 0); 2183 if (if_getcapenable(ifp) & IFCAP_TSO4) 2184 if_sethwassistbits(ifp, CSUM_IP_TSO, 0); 2185 if (if_getcapenable(ifp) & IFCAP_TSO6) 2186 if_sethwassistbits(ifp, CSUM_IP6_TSO, 0); 2187 2188 for (i = 0, txq = ctx->ifc_txqs; i < sctx->isc_ntxqsets; i++, txq++) { 2189 CALLOUT_LOCK(txq); 2190 callout_stop(&txq->ift_timer); 2191 CALLOUT_UNLOCK(txq); 2192 iflib_netmap_txq_init(ctx, txq); 2193 } 2194 #ifdef INVARIANTS 2195 i = if_getdrvflags(ifp); 2196 #endif 2197 IFDI_INIT(ctx); 2198 MPASS(if_getdrvflags(ifp) == i); 2199 for (i = 0, rxq = ctx->ifc_rxqs; i < sctx->isc_nrxqsets; i++, rxq++) { 2200 /* XXX this should really be done on a per-queue basis */ 2201 if (if_getcapenable(ifp) & IFCAP_NETMAP) { 2202 MPASS(rxq->ifr_id == i); 2203 iflib_netmap_rxq_init(ctx, rxq); 2204 continue; 2205 } 2206 for (j = 0, fl = rxq->ifr_fl; j < rxq->ifr_nfl; j++, fl++) { 2207 if (iflib_fl_setup(fl)) { 2208 device_printf(ctx->ifc_dev, "freelist setup failed - check cluster settings\n"); 2209 goto done; 2210 } 2211 } 2212 } 2213 done: 2214 if_setdrvflagbits(ctx->ifc_ifp, IFF_DRV_RUNNING, IFF_DRV_OACTIVE); 2215 IFDI_INTR_ENABLE(ctx); 2216 txq = ctx->ifc_txqs; 2217 for (i = 0; i < sctx->isc_ntxqsets; i++, txq++) 2218 callout_reset_on(&txq->ift_timer, hz/2, iflib_timer, txq, 2219 txq->ift_timer.c_cpu); 2220 } 2221 2222 static int 2223 iflib_media_change(if_t ifp) 2224 { 2225 if_ctx_t ctx = if_getsoftc(ifp); 2226 int err; 2227 2228 CTX_LOCK(ctx); 2229 if ((err = IFDI_MEDIA_CHANGE(ctx)) == 0) 2230 iflib_init_locked(ctx); 2231 CTX_UNLOCK(ctx); 2232 return (err); 2233 } 2234 2235 static void 2236 iflib_media_status(if_t ifp, struct ifmediareq *ifmr) 2237 { 2238 if_ctx_t ctx = if_getsoftc(ifp); 2239 2240 CTX_LOCK(ctx); 2241 IFDI_UPDATE_ADMIN_STATUS(ctx); 2242 IFDI_MEDIA_STATUS(ctx, ifmr); 2243 CTX_UNLOCK(ctx); 2244 } 2245 2246 static void 2247 iflib_stop(if_ctx_t ctx) 2248 { 2249 iflib_txq_t txq = ctx->ifc_txqs; 2250 iflib_rxq_t rxq = ctx->ifc_rxqs; 2251 if_softc_ctx_t scctx = &ctx->ifc_softc_ctx; 2252 iflib_dma_info_t di; 2253 iflib_fl_t fl; 2254 int i, j; 2255 2256 /* Tell the stack that the interface is no longer active */ 2257 if_setdrvflagbits(ctx->ifc_ifp, IFF_DRV_OACTIVE, IFF_DRV_RUNNING); 2258 2259 IFDI_INTR_DISABLE(ctx); 2260 DELAY(1000); 2261 IFDI_STOP(ctx); 2262 DELAY(1000); 2263 2264 iflib_debug_reset(); 2265 /* Wait for current tx queue users to exit to disarm watchdog timer. */ 2266 for (i = 0; i < scctx->isc_ntxqsets; i++, txq++) { 2267 /* make sure all transmitters have completed before proceeding XXX */ 2268 2269 /* clean any enqueued buffers */ 2270 iflib_ifmp_purge(txq); 2271 /* Free any existing tx buffers. */ 2272 for (j = 0; j < txq->ift_size; j++) { 2273 iflib_txsd_free(ctx, txq, j); 2274 } 2275 txq->ift_processed = txq->ift_cleaned = txq->ift_cidx_processed = 0; 2276 txq->ift_in_use = txq->ift_gen = txq->ift_cidx = txq->ift_pidx = txq->ift_no_desc_avail = 0; 2277 txq->ift_closed = txq->ift_mbuf_defrag = txq->ift_mbuf_defrag_failed = 0; 2278 txq->ift_no_tx_dma_setup = txq->ift_txd_encap_efbig = txq->ift_map_failed = 0; 2279 txq->ift_pullups = 0; 2280 ifmp_ring_reset_stats(txq->ift_br); 2281 for (j = 0, di = txq->ift_ifdi; j < ctx->ifc_nhwtxqs; j++, di++) 2282 bzero((void *)di->idi_vaddr, di->idi_size); 2283 } 2284 for (i = 0; i < scctx->isc_nrxqsets; i++, rxq++) { 2285 /* make sure all transmitters have completed before proceeding XXX */ 2286 2287 for (j = 0, di = txq->ift_ifdi; j < ctx->ifc_nhwrxqs; j++, di++) 2288 bzero((void *)di->idi_vaddr, di->idi_size); 2289 /* also resets the free lists pidx/cidx */ 2290 for (j = 0, fl = rxq->ifr_fl; j < rxq->ifr_nfl; j++, fl++) 2291 iflib_fl_bufs_free(fl); 2292 } 2293 } 2294 2295 static inline caddr_t 2296 calc_next_rxd(iflib_fl_t fl, int cidx) 2297 { 2298 qidx_t size; 2299 int nrxd; 2300 caddr_t start, end, cur, next; 2301 2302 nrxd = fl->ifl_size; 2303 size = fl->ifl_rxd_size; 2304 start = fl->ifl_ifdi->idi_vaddr; 2305 2306 if (__predict_false(size == 0)) 2307 return (start); 2308 cur = start + size*cidx; 2309 end = start + size*nrxd; 2310 next = CACHE_PTR_NEXT(cur); 2311 return (next < end ? next : start); 2312 } 2313 2314 static inline void 2315 prefetch_pkts(iflib_fl_t fl, int cidx) 2316 { 2317 int nextptr; 2318 int nrxd = fl->ifl_size; 2319 caddr_t next_rxd; 2320 2321 2322 nextptr = (cidx + CACHE_PTR_INCREMENT) & (nrxd-1); 2323 prefetch(&fl->ifl_sds.ifsd_m[nextptr]); 2324 prefetch(&fl->ifl_sds.ifsd_cl[nextptr]); 2325 next_rxd = calc_next_rxd(fl, cidx); 2326 prefetch(next_rxd); 2327 prefetch(fl->ifl_sds.ifsd_m[(cidx + 1) & (nrxd-1)]); 2328 prefetch(fl->ifl_sds.ifsd_m[(cidx + 2) & (nrxd-1)]); 2329 prefetch(fl->ifl_sds.ifsd_m[(cidx + 3) & (nrxd-1)]); 2330 prefetch(fl->ifl_sds.ifsd_m[(cidx + 4) & (nrxd-1)]); 2331 prefetch(fl->ifl_sds.ifsd_cl[(cidx + 1) & (nrxd-1)]); 2332 prefetch(fl->ifl_sds.ifsd_cl[(cidx + 2) & (nrxd-1)]); 2333 prefetch(fl->ifl_sds.ifsd_cl[(cidx + 3) & (nrxd-1)]); 2334 prefetch(fl->ifl_sds.ifsd_cl[(cidx + 4) & (nrxd-1)]); 2335 } 2336 2337 static void 2338 rxd_frag_to_sd(iflib_rxq_t rxq, if_rxd_frag_t irf, int unload, if_rxsd_t sd) 2339 { 2340 int flid, cidx; 2341 bus_dmamap_t map; 2342 iflib_fl_t fl; 2343 iflib_dma_info_t di; 2344 int next; 2345 2346 map = NULL; 2347 flid = irf->irf_flid; 2348 cidx = irf->irf_idx; 2349 fl = &rxq->ifr_fl[flid]; 2350 sd->ifsd_fl = fl; 2351 sd->ifsd_cidx = cidx; 2352 sd->ifsd_m = &fl->ifl_sds.ifsd_m[cidx]; 2353 sd->ifsd_cl = &fl->ifl_sds.ifsd_cl[cidx]; 2354 fl->ifl_credits--; 2355 #if MEMORY_LOGGING 2356 fl->ifl_m_dequeued++; 2357 #endif 2358 if (rxq->ifr_ctx->ifc_flags & IFC_PREFETCH) 2359 prefetch_pkts(fl, cidx); 2360 if (fl->ifl_sds.ifsd_map != NULL) { 2361 next = (cidx + CACHE_PTR_INCREMENT) & (fl->ifl_size-1); 2362 prefetch(&fl->ifl_sds.ifsd_map[next]); 2363 map = fl->ifl_sds.ifsd_map[cidx]; 2364 di = fl->ifl_ifdi; 2365 next = (cidx + CACHE_LINE_SIZE) & (fl->ifl_size-1); 2366 prefetch(&fl->ifl_sds.ifsd_flags[next]); 2367 bus_dmamap_sync(di->idi_tag, di->idi_map, 2368 BUS_DMASYNC_POSTREAD | BUS_DMASYNC_POSTWRITE); 2369 2370 /* not valid assert if bxe really does SGE from non-contiguous elements */ 2371 MPASS(fl->ifl_cidx == cidx); 2372 if (unload) 2373 bus_dmamap_unload(fl->ifl_desc_tag, map); 2374 } 2375 fl->ifl_cidx = (fl->ifl_cidx + 1) & (fl->ifl_size-1); 2376 if (__predict_false(fl->ifl_cidx == 0)) 2377 fl->ifl_gen = 0; 2378 if (map != NULL) 2379 bus_dmamap_sync(fl->ifl_ifdi->idi_tag, fl->ifl_ifdi->idi_map, 2380 BUS_DMASYNC_PREREAD | BUS_DMASYNC_PREWRITE); 2381 bit_clear(fl->ifl_rx_bitmap, cidx); 2382 } 2383 2384 static struct mbuf * 2385 assemble_segments(iflib_rxq_t rxq, if_rxd_info_t ri, if_rxsd_t sd) 2386 { 2387 int i, padlen , flags; 2388 struct mbuf *m, *mh, *mt; 2389 caddr_t cl; 2390 2391 i = 0; 2392 mh = NULL; 2393 do { 2394 rxd_frag_to_sd(rxq, &ri->iri_frags[i], TRUE, sd); 2395 2396 MPASS(*sd->ifsd_cl != NULL); 2397 MPASS(*sd->ifsd_m != NULL); 2398 2399 /* Don't include zero-length frags */ 2400 if (ri->iri_frags[i].irf_len == 0) { 2401 /* XXX we can save the cluster here, but not the mbuf */ 2402 m_init(*sd->ifsd_m, M_NOWAIT, MT_DATA, 0); 2403 m_free(*sd->ifsd_m); 2404 *sd->ifsd_m = NULL; 2405 continue; 2406 } 2407 m = *sd->ifsd_m; 2408 *sd->ifsd_m = NULL; 2409 if (mh == NULL) { 2410 flags = M_PKTHDR|M_EXT; 2411 mh = mt = m; 2412 padlen = ri->iri_pad; 2413 } else { 2414 flags = M_EXT; 2415 mt->m_next = m; 2416 mt = m; 2417 /* assuming padding is only on the first fragment */ 2418 padlen = 0; 2419 } 2420 cl = *sd->ifsd_cl; 2421 *sd->ifsd_cl = NULL; 2422 2423 /* Can these two be made one ? */ 2424 m_init(m, M_NOWAIT, MT_DATA, flags); 2425 m_cljset(m, cl, sd->ifsd_fl->ifl_cltype); 2426 /* 2427 * These must follow m_init and m_cljset 2428 */ 2429 m->m_data += padlen; 2430 ri->iri_len -= padlen; 2431 m->m_len = ri->iri_frags[i].irf_len; 2432 } while (++i < ri->iri_nfrags); 2433 2434 return (mh); 2435 } 2436 2437 /* 2438 * Process one software descriptor 2439 */ 2440 static struct mbuf * 2441 iflib_rxd_pkt_get(iflib_rxq_t rxq, if_rxd_info_t ri) 2442 { 2443 struct if_rxsd sd; 2444 struct mbuf *m; 2445 2446 /* should I merge this back in now that the two paths are basically duplicated? */ 2447 if (ri->iri_nfrags == 1 && 2448 ri->iri_frags[0].irf_len <= IFLIB_RX_COPY_THRESH) { 2449 rxd_frag_to_sd(rxq, &ri->iri_frags[0], FALSE, &sd); 2450 m = *sd.ifsd_m; 2451 *sd.ifsd_m = NULL; 2452 m_init(m, M_NOWAIT, MT_DATA, M_PKTHDR); 2453 #ifndef __NO_STRICT_ALIGNMENT 2454 if (!IP_ALIGNED(m)) 2455 m->m_data += 2; 2456 #endif 2457 memcpy(m->m_data, *sd.ifsd_cl, ri->iri_len); 2458 m->m_len = ri->iri_frags[0].irf_len; 2459 } else { 2460 m = assemble_segments(rxq, ri, &sd); 2461 } 2462 m->m_pkthdr.len = ri->iri_len; 2463 m->m_pkthdr.rcvif = ri->iri_ifp; 2464 m->m_flags |= ri->iri_flags; 2465 m->m_pkthdr.ether_vtag = ri->iri_vtag; 2466 m->m_pkthdr.flowid = ri->iri_flowid; 2467 M_HASHTYPE_SET(m, ri->iri_rsstype); 2468 m->m_pkthdr.csum_flags = ri->iri_csum_flags; 2469 m->m_pkthdr.csum_data = ri->iri_csum_data; 2470 return (m); 2471 } 2472 2473 #if defined(INET6) || defined(INET) 2474 static void 2475 iflib_get_ip_forwarding(struct lro_ctrl *lc, bool *v4, bool *v6) 2476 { 2477 CURVNET_SET(lc->ifp->if_vnet); 2478 #if defined(INET6) 2479 *v6 = VNET(ip6_forwarding); 2480 #endif 2481 #if defined(INET) 2482 *v4 = VNET(ipforwarding); 2483 #endif 2484 CURVNET_RESTORE(); 2485 } 2486 2487 /* 2488 * Returns true if it's possible this packet could be LROed. 2489 * if it returns false, it is guaranteed that tcp_lro_rx() 2490 * would not return zero. 2491 */ 2492 static bool 2493 iflib_check_lro_possible(struct mbuf *m, bool v4_forwarding, bool v6_forwarding) 2494 { 2495 struct ether_header *eh; 2496 uint16_t eh_type; 2497 2498 eh = mtod(m, struct ether_header *); 2499 eh_type = ntohs(eh->ether_type); 2500 switch (eh_type) { 2501 #if defined(INET6) 2502 case ETHERTYPE_IPV6: 2503 return !v6_forwarding; 2504 #endif 2505 #if defined (INET) 2506 case ETHERTYPE_IP: 2507 return !v4_forwarding; 2508 #endif 2509 } 2510 2511 return false; 2512 } 2513 #else 2514 static void 2515 iflib_get_ip_forwarding(struct lro_ctrl *lc __unused, bool *v4 __unused, bool *v6 __unused) 2516 { 2517 } 2518 #endif 2519 2520 static bool 2521 iflib_rxeof(iflib_rxq_t rxq, qidx_t budget) 2522 { 2523 if_ctx_t ctx = rxq->ifr_ctx; 2524 if_shared_ctx_t sctx = ctx->ifc_sctx; 2525 if_softc_ctx_t scctx = &ctx->ifc_softc_ctx; 2526 int avail, i; 2527 qidx_t *cidxp; 2528 struct if_rxd_info ri; 2529 int err, budget_left, rx_bytes, rx_pkts; 2530 iflib_fl_t fl; 2531 struct ifnet *ifp; 2532 int lro_enabled; 2533 bool lro_possible = false; 2534 bool v4_forwarding, v6_forwarding; 2535 2536 /* 2537 * XXX early demux data packets so that if_input processing only handles 2538 * acks in interrupt context 2539 */ 2540 struct mbuf *m, *mh, *mt, *mf; 2541 2542 ifp = ctx->ifc_ifp; 2543 mh = mt = NULL; 2544 MPASS(budget > 0); 2545 rx_pkts = rx_bytes = 0; 2546 if (sctx->isc_flags & IFLIB_HAS_RXCQ) 2547 cidxp = &rxq->ifr_cq_cidx; 2548 else 2549 cidxp = &rxq->ifr_fl[0].ifl_cidx; 2550 if ((avail = iflib_rxd_avail(ctx, rxq, *cidxp, budget)) == 0) { 2551 for (i = 0, fl = &rxq->ifr_fl[0]; i < sctx->isc_nfl; i++, fl++) 2552 __iflib_fl_refill_lt(ctx, fl, budget + 8); 2553 DBG_COUNTER_INC(rx_unavail); 2554 return (false); 2555 } 2556 2557 for (budget_left = budget; (budget_left > 0) && (avail > 0); budget_left--, avail--) { 2558 if (__predict_false(!CTX_ACTIVE(ctx))) { 2559 DBG_COUNTER_INC(rx_ctx_inactive); 2560 break; 2561 } 2562 /* 2563 * Reset client set fields to their default values 2564 */ 2565 rxd_info_zero(&ri); 2566 ri.iri_qsidx = rxq->ifr_id; 2567 ri.iri_cidx = *cidxp; 2568 ri.iri_ifp = ifp; 2569 ri.iri_frags = rxq->ifr_frags; 2570 err = ctx->isc_rxd_pkt_get(ctx->ifc_softc, &ri); 2571 2572 if (err) 2573 goto err; 2574 if (sctx->isc_flags & IFLIB_HAS_RXCQ) { 2575 *cidxp = ri.iri_cidx; 2576 /* Update our consumer index */ 2577 /* XXX NB: shurd - check if this is still safe */ 2578 while (rxq->ifr_cq_cidx >= scctx->isc_nrxd[0]) { 2579 rxq->ifr_cq_cidx -= scctx->isc_nrxd[0]; 2580 rxq->ifr_cq_gen = 0; 2581 } 2582 /* was this only a completion queue message? */ 2583 if (__predict_false(ri.iri_nfrags == 0)) 2584 continue; 2585 } 2586 MPASS(ri.iri_nfrags != 0); 2587 MPASS(ri.iri_len != 0); 2588 2589 /* will advance the cidx on the corresponding free lists */ 2590 m = iflib_rxd_pkt_get(rxq, &ri); 2591 if (avail == 0 && budget_left) 2592 avail = iflib_rxd_avail(ctx, rxq, *cidxp, budget_left); 2593 2594 if (__predict_false(m == NULL)) { 2595 DBG_COUNTER_INC(rx_mbuf_null); 2596 continue; 2597 } 2598 /* imm_pkt: -- cxgb */ 2599 if (mh == NULL) 2600 mh = mt = m; 2601 else { 2602 mt->m_nextpkt = m; 2603 mt = m; 2604 } 2605 } 2606 /* make sure that we can refill faster than drain */ 2607 for (i = 0, fl = &rxq->ifr_fl[0]; i < sctx->isc_nfl; i++, fl++) 2608 __iflib_fl_refill_lt(ctx, fl, budget + 8); 2609 2610 lro_enabled = (if_getcapenable(ifp) & IFCAP_LRO); 2611 if (lro_enabled) 2612 iflib_get_ip_forwarding(&rxq->ifr_lc, &v4_forwarding, &v6_forwarding); 2613 mt = mf = NULL; 2614 while (mh != NULL) { 2615 m = mh; 2616 mh = mh->m_nextpkt; 2617 m->m_nextpkt = NULL; 2618 #ifndef __NO_STRICT_ALIGNMENT 2619 if (!IP_ALIGNED(m) && (m = iflib_fixup_rx(m)) == NULL) 2620 continue; 2621 #endif 2622 rx_bytes += m->m_pkthdr.len; 2623 rx_pkts++; 2624 #if defined(INET6) || defined(INET) 2625 if (lro_enabled) { 2626 if (!lro_possible) { 2627 lro_possible = iflib_check_lro_possible(m, v4_forwarding, v6_forwarding); 2628 if (lro_possible && mf != NULL) { 2629 ifp->if_input(ifp, mf); 2630 DBG_COUNTER_INC(rx_if_input); 2631 mt = mf = NULL; 2632 } 2633 } 2634 if (lro_possible && tcp_lro_rx(&rxq->ifr_lc, m, 0) == 0) 2635 continue; 2636 } 2637 #endif 2638 if (lro_possible) { 2639 ifp->if_input(ifp, m); 2640 DBG_COUNTER_INC(rx_if_input); 2641 continue; 2642 } 2643 2644 if (mf == NULL) 2645 mf = m; 2646 if (mt != NULL) 2647 mt->m_nextpkt = m; 2648 mt = m; 2649 } 2650 if (mf != NULL) { 2651 ifp->if_input(ifp, mf); 2652 DBG_COUNTER_INC(rx_if_input); 2653 } 2654 2655 if_inc_counter(ifp, IFCOUNTER_IBYTES, rx_bytes); 2656 if_inc_counter(ifp, IFCOUNTER_IPACKETS, rx_pkts); 2657 2658 /* 2659 * Flush any outstanding LRO work 2660 */ 2661 #if defined(INET6) || defined(INET) 2662 tcp_lro_flush_all(&rxq->ifr_lc); 2663 #endif 2664 if (avail) 2665 return true; 2666 return (iflib_rxd_avail(ctx, rxq, *cidxp, 1)); 2667 err: 2668 CTX_LOCK(ctx); 2669 ctx->ifc_flags |= IFC_DO_RESET; 2670 iflib_admin_intr_deferred(ctx); 2671 CTX_UNLOCK(ctx); 2672 return (false); 2673 } 2674 2675 #define TXD_NOTIFY_COUNT(txq) (((txq)->ift_size / (txq)->ift_update_freq)-1) 2676 static inline qidx_t 2677 txq_max_db_deferred(iflib_txq_t txq, qidx_t in_use) 2678 { 2679 qidx_t notify_count = TXD_NOTIFY_COUNT(txq); 2680 qidx_t minthresh = txq->ift_size / 8; 2681 if (in_use > 4*minthresh) 2682 return (notify_count); 2683 if (in_use > 2*minthresh) 2684 return (notify_count >> 1); 2685 if (in_use > minthresh) 2686 return (notify_count >> 3); 2687 return (0); 2688 } 2689 2690 static inline qidx_t 2691 txq_max_rs_deferred(iflib_txq_t txq) 2692 { 2693 qidx_t notify_count = TXD_NOTIFY_COUNT(txq); 2694 qidx_t minthresh = txq->ift_size / 8; 2695 if (txq->ift_in_use > 4*minthresh) 2696 return (notify_count); 2697 if (txq->ift_in_use > 2*minthresh) 2698 return (notify_count >> 1); 2699 if (txq->ift_in_use > minthresh) 2700 return (notify_count >> 2); 2701 return (2); 2702 } 2703 2704 #define M_CSUM_FLAGS(m) ((m)->m_pkthdr.csum_flags) 2705 #define M_HAS_VLANTAG(m) (m->m_flags & M_VLANTAG) 2706 2707 #define TXQ_MAX_DB_DEFERRED(txq, in_use) txq_max_db_deferred((txq), (in_use)) 2708 #define TXQ_MAX_RS_DEFERRED(txq) txq_max_rs_deferred(txq) 2709 #define TXQ_MAX_DB_CONSUMED(size) (size >> 4) 2710 2711 /* forward compatibility for cxgb */ 2712 #define FIRST_QSET(ctx) 0 2713 #define NTXQSETS(ctx) ((ctx)->ifc_softc_ctx.isc_ntxqsets) 2714 #define NRXQSETS(ctx) ((ctx)->ifc_softc_ctx.isc_nrxqsets) 2715 #define QIDX(ctx, m) ((((m)->m_pkthdr.flowid & ctx->ifc_softc_ctx.isc_rss_table_mask) % NTXQSETS(ctx)) + FIRST_QSET(ctx)) 2716 #define DESC_RECLAIMABLE(q) ((int)((q)->ift_processed - (q)->ift_cleaned - (q)->ift_ctx->ifc_softc_ctx.isc_tx_nsegments)) 2717 2718 /* XXX we should be setting this to something other than zero */ 2719 #define RECLAIM_THRESH(ctx) ((ctx)->ifc_sctx->isc_tx_reclaim_thresh) 2720 #define MAX_TX_DESC(ctx) ((ctx)->ifc_softc_ctx.isc_tx_tso_segments_max) 2721 2722 static inline bool 2723 iflib_txd_db_check(if_ctx_t ctx, iflib_txq_t txq, int ring, qidx_t in_use) 2724 { 2725 qidx_t dbval, max; 2726 bool rang; 2727 2728 rang = false; 2729 max = TXQ_MAX_DB_DEFERRED(txq, in_use); 2730 if (ring || txq->ift_db_pending >= max) { 2731 dbval = txq->ift_npending ? txq->ift_npending : txq->ift_pidx; 2732 ctx->isc_txd_flush(ctx->ifc_softc, txq->ift_id, dbval); 2733 txq->ift_db_pending = txq->ift_npending = 0; 2734 rang = true; 2735 } 2736 return (rang); 2737 } 2738 2739 #ifdef PKT_DEBUG 2740 static void 2741 print_pkt(if_pkt_info_t pi) 2742 { 2743 printf("pi len: %d qsidx: %d nsegs: %d ndescs: %d flags: %x pidx: %d\n", 2744 pi->ipi_len, pi->ipi_qsidx, pi->ipi_nsegs, pi->ipi_ndescs, pi->ipi_flags, pi->ipi_pidx); 2745 printf("pi new_pidx: %d csum_flags: %lx tso_segsz: %d mflags: %x vtag: %d\n", 2746 pi->ipi_new_pidx, pi->ipi_csum_flags, pi->ipi_tso_segsz, pi->ipi_mflags, pi->ipi_vtag); 2747 printf("pi etype: %d ehdrlen: %d ip_hlen: %d ipproto: %d\n", 2748 pi->ipi_etype, pi->ipi_ehdrlen, pi->ipi_ip_hlen, pi->ipi_ipproto); 2749 } 2750 #endif 2751 2752 #define IS_TSO4(pi) ((pi)->ipi_csum_flags & CSUM_IP_TSO) 2753 #define IS_TSO6(pi) ((pi)->ipi_csum_flags & CSUM_IP6_TSO) 2754 2755 static int 2756 iflib_parse_header(iflib_txq_t txq, if_pkt_info_t pi, struct mbuf **mp) 2757 { 2758 if_shared_ctx_t sctx = txq->ift_ctx->ifc_sctx; 2759 struct ether_vlan_header *eh; 2760 struct mbuf *m, *n; 2761 2762 n = m = *mp; 2763 if ((sctx->isc_flags & IFLIB_NEED_SCRATCH) && 2764 M_WRITABLE(m) == 0) { 2765 if ((m = m_dup(m, M_NOWAIT)) == NULL) { 2766 return (ENOMEM); 2767 } else { 2768 m_freem(*mp); 2769 n = *mp = m; 2770 } 2771 } 2772 2773 /* 2774 * Determine where frame payload starts. 2775 * Jump over vlan headers if already present, 2776 * helpful for QinQ too. 2777 */ 2778 if (__predict_false(m->m_len < sizeof(*eh))) { 2779 txq->ift_pullups++; 2780 if (__predict_false((m = m_pullup(m, sizeof(*eh))) == NULL)) 2781 return (ENOMEM); 2782 } 2783 eh = mtod(m, struct ether_vlan_header *); 2784 if (eh->evl_encap_proto == htons(ETHERTYPE_VLAN)) { 2785 pi->ipi_etype = ntohs(eh->evl_proto); 2786 pi->ipi_ehdrlen = ETHER_HDR_LEN + ETHER_VLAN_ENCAP_LEN; 2787 } else { 2788 pi->ipi_etype = ntohs(eh->evl_encap_proto); 2789 pi->ipi_ehdrlen = ETHER_HDR_LEN; 2790 } 2791 2792 switch (pi->ipi_etype) { 2793 #ifdef INET 2794 case ETHERTYPE_IP: 2795 { 2796 struct ip *ip = NULL; 2797 struct tcphdr *th = NULL; 2798 int minthlen; 2799 2800 minthlen = min(m->m_pkthdr.len, pi->ipi_ehdrlen + sizeof(*ip) + sizeof(*th)); 2801 if (__predict_false(m->m_len < minthlen)) { 2802 /* 2803 * if this code bloat is causing too much of a hit 2804 * move it to a separate function and mark it noinline 2805 */ 2806 if (m->m_len == pi->ipi_ehdrlen) { 2807 n = m->m_next; 2808 MPASS(n); 2809 if (n->m_len >= sizeof(*ip)) { 2810 ip = (struct ip *)n->m_data; 2811 if (n->m_len >= (ip->ip_hl << 2) + sizeof(*th)) 2812 th = (struct tcphdr *)((caddr_t)ip + (ip->ip_hl << 2)); 2813 } else { 2814 txq->ift_pullups++; 2815 if (__predict_false((m = m_pullup(m, minthlen)) == NULL)) 2816 return (ENOMEM); 2817 ip = (struct ip *)(m->m_data + pi->ipi_ehdrlen); 2818 } 2819 } else { 2820 txq->ift_pullups++; 2821 if (__predict_false((m = m_pullup(m, minthlen)) == NULL)) 2822 return (ENOMEM); 2823 ip = (struct ip *)(m->m_data + pi->ipi_ehdrlen); 2824 if (m->m_len >= (ip->ip_hl << 2) + sizeof(*th)) 2825 th = (struct tcphdr *)((caddr_t)ip + (ip->ip_hl << 2)); 2826 } 2827 } else { 2828 ip = (struct ip *)(m->m_data + pi->ipi_ehdrlen); 2829 if (m->m_len >= (ip->ip_hl << 2) + sizeof(*th)) 2830 th = (struct tcphdr *)((caddr_t)ip + (ip->ip_hl << 2)); 2831 } 2832 pi->ipi_ip_hlen = ip->ip_hl << 2; 2833 pi->ipi_ipproto = ip->ip_p; 2834 pi->ipi_flags |= IPI_TX_IPV4; 2835 2836 if ((sctx->isc_flags & IFLIB_NEED_ZERO_CSUM) && (pi->ipi_csum_flags & CSUM_IP)) 2837 ip->ip_sum = 0; 2838 2839 if (IS_TSO4(pi)) { 2840 if (pi->ipi_ipproto == IPPROTO_TCP) { 2841 if (__predict_false(th == NULL)) { 2842 txq->ift_pullups++; 2843 if (__predict_false((m = m_pullup(m, (ip->ip_hl << 2) + sizeof(*th))) == NULL)) 2844 return (ENOMEM); 2845 th = (struct tcphdr *)((caddr_t)ip + pi->ipi_ip_hlen); 2846 } 2847 pi->ipi_tcp_hflags = th->th_flags; 2848 pi->ipi_tcp_hlen = th->th_off << 2; 2849 pi->ipi_tcp_seq = th->th_seq; 2850 } 2851 if (__predict_false(ip->ip_p != IPPROTO_TCP)) 2852 return (ENXIO); 2853 th->th_sum = in_pseudo(ip->ip_src.s_addr, 2854 ip->ip_dst.s_addr, htons(IPPROTO_TCP)); 2855 pi->ipi_tso_segsz = m->m_pkthdr.tso_segsz; 2856 if (sctx->isc_flags & IFLIB_TSO_INIT_IP) { 2857 ip->ip_sum = 0; 2858 ip->ip_len = htons(pi->ipi_ip_hlen + pi->ipi_tcp_hlen + pi->ipi_tso_segsz); 2859 } 2860 } 2861 break; 2862 } 2863 #endif 2864 #ifdef INET6 2865 case ETHERTYPE_IPV6: 2866 { 2867 struct ip6_hdr *ip6 = (struct ip6_hdr *)(m->m_data + pi->ipi_ehdrlen); 2868 struct tcphdr *th; 2869 pi->ipi_ip_hlen = sizeof(struct ip6_hdr); 2870 2871 if (__predict_false(m->m_len < pi->ipi_ehdrlen + sizeof(struct ip6_hdr))) { 2872 if (__predict_false((m = m_pullup(m, pi->ipi_ehdrlen + sizeof(struct ip6_hdr))) == NULL)) 2873 return (ENOMEM); 2874 } 2875 th = (struct tcphdr *)((caddr_t)ip6 + pi->ipi_ip_hlen); 2876 2877 /* XXX-BZ this will go badly in case of ext hdrs. */ 2878 pi->ipi_ipproto = ip6->ip6_nxt; 2879 pi->ipi_flags |= IPI_TX_IPV6; 2880 2881 if (IS_TSO6(pi)) { 2882 if (pi->ipi_ipproto == IPPROTO_TCP) { 2883 if (__predict_false(m->m_len < pi->ipi_ehdrlen + sizeof(struct ip6_hdr) + sizeof(struct tcphdr))) { 2884 if (__predict_false((m = m_pullup(m, pi->ipi_ehdrlen + sizeof(struct ip6_hdr) + sizeof(struct tcphdr))) == NULL)) 2885 return (ENOMEM); 2886 } 2887 pi->ipi_tcp_hflags = th->th_flags; 2888 pi->ipi_tcp_hlen = th->th_off << 2; 2889 } 2890 2891 if (__predict_false(ip6->ip6_nxt != IPPROTO_TCP)) 2892 return (ENXIO); 2893 /* 2894 * The corresponding flag is set by the stack in the IPv4 2895 * TSO case, but not in IPv6 (at least in FreeBSD 10.2). 2896 * So, set it here because the rest of the flow requires it. 2897 */ 2898 pi->ipi_csum_flags |= CSUM_TCP_IPV6; 2899 th->th_sum = in6_cksum_pseudo(ip6, 0, IPPROTO_TCP, 0); 2900 pi->ipi_tso_segsz = m->m_pkthdr.tso_segsz; 2901 } 2902 break; 2903 } 2904 #endif 2905 default: 2906 pi->ipi_csum_flags &= ~CSUM_OFFLOAD; 2907 pi->ipi_ip_hlen = 0; 2908 break; 2909 } 2910 *mp = m; 2911 2912 return (0); 2913 } 2914 2915 static __noinline struct mbuf * 2916 collapse_pkthdr(struct mbuf *m0) 2917 { 2918 struct mbuf *m, *m_next, *tmp; 2919 2920 m = m0; 2921 m_next = m->m_next; 2922 while (m_next != NULL && m_next->m_len == 0) { 2923 m = m_next; 2924 m->m_next = NULL; 2925 m_free(m); 2926 m_next = m_next->m_next; 2927 } 2928 m = m0; 2929 m->m_next = m_next; 2930 if ((m_next->m_flags & M_EXT) == 0) { 2931 m = m_defrag(m, M_NOWAIT); 2932 } else { 2933 tmp = m_next->m_next; 2934 memcpy(m_next, m, MPKTHSIZE); 2935 m = m_next; 2936 m->m_next = tmp; 2937 } 2938 return (m); 2939 } 2940 2941 /* 2942 * If dodgy hardware rejects the scatter gather chain we've handed it 2943 * we'll need to remove the mbuf chain from ifsg_m[] before we can add the 2944 * m_defrag'd mbufs 2945 */ 2946 static __noinline struct mbuf * 2947 iflib_remove_mbuf(iflib_txq_t txq) 2948 { 2949 int ntxd, i, pidx; 2950 struct mbuf *m, *mh, **ifsd_m; 2951 2952 pidx = txq->ift_pidx; 2953 ifsd_m = txq->ift_sds.ifsd_m; 2954 ntxd = txq->ift_size; 2955 mh = m = ifsd_m[pidx]; 2956 ifsd_m[pidx] = NULL; 2957 #if MEMORY_LOGGING 2958 txq->ift_dequeued++; 2959 #endif 2960 i = 1; 2961 2962 while (m) { 2963 ifsd_m[(pidx + i) & (ntxd -1)] = NULL; 2964 #if MEMORY_LOGGING 2965 txq->ift_dequeued++; 2966 #endif 2967 m = m->m_next; 2968 i++; 2969 } 2970 return (mh); 2971 } 2972 2973 static int 2974 iflib_busdma_load_mbuf_sg(iflib_txq_t txq, bus_dma_tag_t tag, bus_dmamap_t map, 2975 struct mbuf **m0, bus_dma_segment_t *segs, int *nsegs, 2976 int max_segs, int flags) 2977 { 2978 if_ctx_t ctx; 2979 if_shared_ctx_t sctx; 2980 if_softc_ctx_t scctx; 2981 int i, next, pidx, err, ntxd, count; 2982 struct mbuf *m, *tmp, **ifsd_m; 2983 2984 m = *m0; 2985 2986 /* 2987 * Please don't ever do this 2988 */ 2989 if (__predict_false(m->m_len == 0)) 2990 *m0 = m = collapse_pkthdr(m); 2991 2992 ctx = txq->ift_ctx; 2993 sctx = ctx->ifc_sctx; 2994 scctx = &ctx->ifc_softc_ctx; 2995 ifsd_m = txq->ift_sds.ifsd_m; 2996 ntxd = txq->ift_size; 2997 pidx = txq->ift_pidx; 2998 if (map != NULL) { 2999 uint8_t *ifsd_flags = txq->ift_sds.ifsd_flags; 3000 3001 err = bus_dmamap_load_mbuf_sg(tag, map, 3002 *m0, segs, nsegs, BUS_DMA_NOWAIT); 3003 if (err) 3004 return (err); 3005 ifsd_flags[pidx] |= TX_SW_DESC_MAPPED; 3006 count = 0; 3007 m = *m0; 3008 do { 3009 if (__predict_false(m->m_len <= 0)) { 3010 tmp = m; 3011 m = m->m_next; 3012 tmp->m_next = NULL; 3013 m_free(tmp); 3014 continue; 3015 } 3016 m = m->m_next; 3017 count++; 3018 } while (m != NULL); 3019 if (count > *nsegs) { 3020 ifsd_m[pidx] = *m0; 3021 ifsd_m[pidx]->m_flags |= M_TOOBIG; 3022 return (0); 3023 } 3024 m = *m0; 3025 count = 0; 3026 do { 3027 next = (pidx + count) & (ntxd-1); 3028 MPASS(ifsd_m[next] == NULL); 3029 ifsd_m[next] = m; 3030 count++; 3031 tmp = m; 3032 m = m->m_next; 3033 } while (m != NULL); 3034 } else { 3035 int buflen, sgsize, maxsegsz, max_sgsize; 3036 vm_offset_t vaddr; 3037 vm_paddr_t curaddr; 3038 3039 count = i = 0; 3040 m = *m0; 3041 if (m->m_pkthdr.csum_flags & CSUM_TSO) 3042 maxsegsz = scctx->isc_tx_tso_segsize_max; 3043 else 3044 maxsegsz = sctx->isc_tx_maxsegsize; 3045 3046 do { 3047 if (__predict_false(m->m_len <= 0)) { 3048 tmp = m; 3049 m = m->m_next; 3050 tmp->m_next = NULL; 3051 m_free(tmp); 3052 continue; 3053 } 3054 buflen = m->m_len; 3055 vaddr = (vm_offset_t)m->m_data; 3056 /* 3057 * see if we can't be smarter about physically 3058 * contiguous mappings 3059 */ 3060 next = (pidx + count) & (ntxd-1); 3061 MPASS(ifsd_m[next] == NULL); 3062 #if MEMORY_LOGGING 3063 txq->ift_enqueued++; 3064 #endif 3065 ifsd_m[next] = m; 3066 while (buflen > 0) { 3067 if (i >= max_segs) 3068 goto err; 3069 max_sgsize = MIN(buflen, maxsegsz); 3070 curaddr = pmap_kextract(vaddr); 3071 sgsize = PAGE_SIZE - (curaddr & PAGE_MASK); 3072 sgsize = MIN(sgsize, max_sgsize); 3073 segs[i].ds_addr = curaddr; 3074 segs[i].ds_len = sgsize; 3075 vaddr += sgsize; 3076 buflen -= sgsize; 3077 i++; 3078 } 3079 count++; 3080 tmp = m; 3081 m = m->m_next; 3082 } while (m != NULL); 3083 *nsegs = i; 3084 } 3085 return (0); 3086 err: 3087 *m0 = iflib_remove_mbuf(txq); 3088 return (EFBIG); 3089 } 3090 3091 static inline caddr_t 3092 calc_next_txd(iflib_txq_t txq, int cidx, uint8_t qid) 3093 { 3094 qidx_t size; 3095 int ntxd; 3096 caddr_t start, end, cur, next; 3097 3098 ntxd = txq->ift_size; 3099 size = txq->ift_txd_size[qid]; 3100 start = txq->ift_ifdi[qid].idi_vaddr; 3101 3102 if (__predict_false(size == 0)) 3103 return (start); 3104 cur = start + size*cidx; 3105 end = start + size*ntxd; 3106 next = CACHE_PTR_NEXT(cur); 3107 return (next < end ? next : start); 3108 } 3109 3110 /* 3111 * Pad an mbuf to ensure a minimum ethernet frame size. 3112 * min_frame_size is the frame size (less CRC) to pad the mbuf to 3113 */ 3114 static __noinline int 3115 iflib_ether_pad(device_t dev, struct mbuf **m_head, uint16_t min_frame_size) 3116 { 3117 /* 3118 * 18 is enough bytes to pad an ARP packet to 46 bytes, and 3119 * and ARP message is the smallest common payload I can think of 3120 */ 3121 static char pad[18]; /* just zeros */ 3122 int n; 3123 struct mbuf *new_head; 3124 3125 if (!M_WRITABLE(*m_head)) { 3126 new_head = m_dup(*m_head, M_NOWAIT); 3127 if (new_head == NULL) { 3128 m_freem(*m_head); 3129 device_printf(dev, "cannot pad short frame, m_dup() failed"); 3130 return ENOMEM; 3131 } 3132 m_freem(*m_head); 3133 *m_head = new_head; 3134 } 3135 3136 for (n = min_frame_size - (*m_head)->m_pkthdr.len; 3137 n > 0; n -= sizeof(pad)) 3138 if (!m_append(*m_head, min(n, sizeof(pad)), pad)) 3139 break; 3140 3141 if (n > 0) { 3142 m_freem(*m_head); 3143 device_printf(dev, "cannot pad short frame\n"); 3144 DBG_COUNTER_INC(encap_pad_mbuf_fail); 3145 return (ENOBUFS); 3146 } 3147 3148 return 0; 3149 } 3150 3151 static int 3152 iflib_encap(iflib_txq_t txq, struct mbuf **m_headp) 3153 { 3154 if_ctx_t ctx; 3155 if_shared_ctx_t sctx; 3156 if_softc_ctx_t scctx; 3157 bus_dma_segment_t *segs; 3158 struct mbuf *m_head; 3159 void *next_txd; 3160 bus_dmamap_t map; 3161 struct if_pkt_info pi; 3162 int remap = 0; 3163 int err, nsegs, ndesc, max_segs, pidx, cidx, next, ntxd; 3164 bus_dma_tag_t desc_tag; 3165 3166 segs = txq->ift_segs; 3167 ctx = txq->ift_ctx; 3168 sctx = ctx->ifc_sctx; 3169 scctx = &ctx->ifc_softc_ctx; 3170 segs = txq->ift_segs; 3171 ntxd = txq->ift_size; 3172 m_head = *m_headp; 3173 map = NULL; 3174 3175 /* 3176 * If we're doing TSO the next descriptor to clean may be quite far ahead 3177 */ 3178 cidx = txq->ift_cidx; 3179 pidx = txq->ift_pidx; 3180 if (ctx->ifc_flags & IFC_PREFETCH) { 3181 next = (cidx + CACHE_PTR_INCREMENT) & (ntxd-1); 3182 if (!(ctx->ifc_flags & IFLIB_HAS_TXCQ)) { 3183 next_txd = calc_next_txd(txq, cidx, 0); 3184 prefetch(next_txd); 3185 } 3186 3187 /* prefetch the next cache line of mbuf pointers and flags */ 3188 prefetch(&txq->ift_sds.ifsd_m[next]); 3189 if (txq->ift_sds.ifsd_map != NULL) { 3190 prefetch(&txq->ift_sds.ifsd_map[next]); 3191 next = (cidx + CACHE_LINE_SIZE) & (ntxd-1); 3192 prefetch(&txq->ift_sds.ifsd_flags[next]); 3193 } 3194 } else if (txq->ift_sds.ifsd_map != NULL) 3195 map = txq->ift_sds.ifsd_map[pidx]; 3196 3197 if (m_head->m_pkthdr.csum_flags & CSUM_TSO) { 3198 desc_tag = txq->ift_tso_desc_tag; 3199 max_segs = scctx->isc_tx_tso_segments_max; 3200 } else { 3201 desc_tag = txq->ift_desc_tag; 3202 max_segs = scctx->isc_tx_nsegments; 3203 } 3204 if ((sctx->isc_flags & IFLIB_NEED_ETHER_PAD) && 3205 __predict_false(m_head->m_pkthdr.len < scctx->isc_min_frame_size)) { 3206 err = iflib_ether_pad(ctx->ifc_dev, m_headp, scctx->isc_min_frame_size); 3207 if (err) 3208 return err; 3209 } 3210 m_head = *m_headp; 3211 3212 pkt_info_zero(&pi); 3213 pi.ipi_mflags = (m_head->m_flags & (M_VLANTAG|M_BCAST|M_MCAST)); 3214 pi.ipi_pidx = pidx; 3215 pi.ipi_qsidx = txq->ift_id; 3216 pi.ipi_len = m_head->m_pkthdr.len; 3217 pi.ipi_csum_flags = m_head->m_pkthdr.csum_flags; 3218 pi.ipi_vtag = (m_head->m_flags & M_VLANTAG) ? m_head->m_pkthdr.ether_vtag : 0; 3219 3220 /* deliberate bitwise OR to make one condition */ 3221 if (__predict_true((pi.ipi_csum_flags | pi.ipi_vtag))) { 3222 if (__predict_false((err = iflib_parse_header(txq, &pi, m_headp)) != 0)) 3223 return (err); 3224 m_head = *m_headp; 3225 } 3226 3227 retry: 3228 err = iflib_busdma_load_mbuf_sg(txq, desc_tag, map, m_headp, segs, &nsegs, max_segs, BUS_DMA_NOWAIT); 3229 defrag: 3230 if (__predict_false(err)) { 3231 switch (err) { 3232 case EFBIG: 3233 /* try collapse once and defrag once */ 3234 if (remap == 0) 3235 m_head = m_collapse(*m_headp, M_NOWAIT, max_segs); 3236 if (remap == 1) 3237 m_head = m_defrag(*m_headp, M_NOWAIT); 3238 remap++; 3239 if (__predict_false(m_head == NULL)) 3240 goto defrag_failed; 3241 txq->ift_mbuf_defrag++; 3242 *m_headp = m_head; 3243 goto retry; 3244 break; 3245 case ENOMEM: 3246 txq->ift_no_tx_dma_setup++; 3247 break; 3248 default: 3249 txq->ift_no_tx_dma_setup++; 3250 m_freem(*m_headp); 3251 DBG_COUNTER_INC(tx_frees); 3252 *m_headp = NULL; 3253 break; 3254 } 3255 txq->ift_map_failed++; 3256 DBG_COUNTER_INC(encap_load_mbuf_fail); 3257 return (err); 3258 } 3259 3260 /* 3261 * XXX assumes a 1 to 1 relationship between segments and 3262 * descriptors - this does not hold true on all drivers, e.g. 3263 * cxgb 3264 */ 3265 if (__predict_false(nsegs + 2 > TXQ_AVAIL(txq))) { 3266 txq->ift_no_desc_avail++; 3267 if (map != NULL) 3268 bus_dmamap_unload(desc_tag, map); 3269 DBG_COUNTER_INC(encap_txq_avail_fail); 3270 if ((txq->ift_task.gt_task.ta_flags & TASK_ENQUEUED) == 0) 3271 GROUPTASK_ENQUEUE(&txq->ift_task); 3272 return (ENOBUFS); 3273 } 3274 /* 3275 * On Intel cards we can greatly reduce the number of TX interrupts 3276 * we see by only setting report status on every Nth descriptor. 3277 * However, this also means that the driver will need to keep track 3278 * of the descriptors that RS was set on to check them for the DD bit. 3279 */ 3280 txq->ift_rs_pending += nsegs + 1; 3281 if (txq->ift_rs_pending > TXQ_MAX_RS_DEFERRED(txq) || 3282 iflib_no_tx_batch || (TXQ_AVAIL(txq) - nsegs - 1) <= MAX_TX_DESC(ctx)) { 3283 pi.ipi_flags |= IPI_TX_INTR; 3284 txq->ift_rs_pending = 0; 3285 } 3286 3287 pi.ipi_segs = segs; 3288 pi.ipi_nsegs = nsegs; 3289 3290 MPASS(pidx >= 0 && pidx < txq->ift_size); 3291 #ifdef PKT_DEBUG 3292 print_pkt(&pi); 3293 #endif 3294 if (map != NULL) 3295 bus_dmamap_sync(desc_tag, map, BUS_DMASYNC_PREWRITE); 3296 if ((err = ctx->isc_txd_encap(ctx->ifc_softc, &pi)) == 0) { 3297 if (map != NULL) 3298 bus_dmamap_sync(txq->ift_ifdi->idi_tag, txq->ift_ifdi->idi_map, 3299 BUS_DMASYNC_PREREAD | BUS_DMASYNC_PREWRITE); 3300 DBG_COUNTER_INC(tx_encap); 3301 MPASS(pi.ipi_new_pidx < txq->ift_size); 3302 3303 ndesc = pi.ipi_new_pidx - pi.ipi_pidx; 3304 if (pi.ipi_new_pidx < pi.ipi_pidx) { 3305 ndesc += txq->ift_size; 3306 txq->ift_gen = 1; 3307 } 3308 /* 3309 * drivers can need as many as 3310 * two sentinels 3311 */ 3312 MPASS(ndesc <= pi.ipi_nsegs + 2); 3313 MPASS(pi.ipi_new_pidx != pidx); 3314 MPASS(ndesc > 0); 3315 txq->ift_in_use += ndesc; 3316 3317 /* 3318 * We update the last software descriptor again here because there may 3319 * be a sentinel and/or there may be more mbufs than segments 3320 */ 3321 txq->ift_pidx = pi.ipi_new_pidx; 3322 txq->ift_npending += pi.ipi_ndescs; 3323 } else if (__predict_false(err == EFBIG && remap < 2)) { 3324 *m_headp = m_head = iflib_remove_mbuf(txq); 3325 remap = 1; 3326 txq->ift_txd_encap_efbig++; 3327 goto defrag; 3328 } else 3329 DBG_COUNTER_INC(encap_txd_encap_fail); 3330 return (err); 3331 3332 defrag_failed: 3333 txq->ift_mbuf_defrag_failed++; 3334 txq->ift_map_failed++; 3335 m_freem(*m_headp); 3336 DBG_COUNTER_INC(tx_frees); 3337 *m_headp = NULL; 3338 return (ENOMEM); 3339 } 3340 3341 static void 3342 iflib_tx_desc_free(iflib_txq_t txq, int n) 3343 { 3344 int hasmap; 3345 uint32_t qsize, cidx, mask, gen; 3346 struct mbuf *m, **ifsd_m; 3347 uint8_t *ifsd_flags; 3348 bus_dmamap_t *ifsd_map; 3349 bool do_prefetch; 3350 3351 cidx = txq->ift_cidx; 3352 gen = txq->ift_gen; 3353 qsize = txq->ift_size; 3354 mask = qsize-1; 3355 hasmap = txq->ift_sds.ifsd_map != NULL; 3356 ifsd_flags = txq->ift_sds.ifsd_flags; 3357 ifsd_m = txq->ift_sds.ifsd_m; 3358 ifsd_map = txq->ift_sds.ifsd_map; 3359 do_prefetch = (txq->ift_ctx->ifc_flags & IFC_PREFETCH); 3360 3361 while (n--) { 3362 if (do_prefetch) { 3363 prefetch(ifsd_m[(cidx + 3) & mask]); 3364 prefetch(ifsd_m[(cidx + 4) & mask]); 3365 } 3366 if (ifsd_m[cidx] != NULL) { 3367 prefetch(&ifsd_m[(cidx + CACHE_PTR_INCREMENT) & mask]); 3368 prefetch(&ifsd_flags[(cidx + CACHE_PTR_INCREMENT) & mask]); 3369 if (hasmap && (ifsd_flags[cidx] & TX_SW_DESC_MAPPED)) { 3370 /* 3371 * does it matter if it's not the TSO tag? If so we'll 3372 * have to add the type to flags 3373 */ 3374 bus_dmamap_unload(txq->ift_desc_tag, ifsd_map[cidx]); 3375 ifsd_flags[cidx] &= ~TX_SW_DESC_MAPPED; 3376 } 3377 if ((m = ifsd_m[cidx]) != NULL) { 3378 /* XXX we don't support any drivers that batch packets yet */ 3379 MPASS(m->m_nextpkt == NULL); 3380 /* if the number of clusters exceeds the number of segments 3381 * there won't be space on the ring to save a pointer to each 3382 * cluster so we simply free the list here 3383 */ 3384 if (m->m_flags & M_TOOBIG) { 3385 m_freem(m); 3386 } else { 3387 m_free(m); 3388 } 3389 ifsd_m[cidx] = NULL; 3390 #if MEMORY_LOGGING 3391 txq->ift_dequeued++; 3392 #endif 3393 DBG_COUNTER_INC(tx_frees); 3394 } 3395 } 3396 if (__predict_false(++cidx == qsize)) { 3397 cidx = 0; 3398 gen = 0; 3399 } 3400 } 3401 txq->ift_cidx = cidx; 3402 txq->ift_gen = gen; 3403 } 3404 3405 static __inline int 3406 iflib_completed_tx_reclaim(iflib_txq_t txq, int thresh) 3407 { 3408 int reclaim; 3409 if_ctx_t ctx = txq->ift_ctx; 3410 3411 KASSERT(thresh >= 0, ("invalid threshold to reclaim")); 3412 MPASS(thresh /*+ MAX_TX_DESC(txq->ift_ctx) */ < txq->ift_size); 3413 3414 /* 3415 * Need a rate-limiting check so that this isn't called every time 3416 */ 3417 iflib_tx_credits_update(ctx, txq); 3418 reclaim = DESC_RECLAIMABLE(txq); 3419 3420 if (reclaim <= thresh /* + MAX_TX_DESC(txq->ift_ctx) */) { 3421 #ifdef INVARIANTS 3422 if (iflib_verbose_debug) { 3423 printf("%s processed=%ju cleaned=%ju tx_nsegments=%d reclaim=%d thresh=%d\n", __FUNCTION__, 3424 txq->ift_processed, txq->ift_cleaned, txq->ift_ctx->ifc_softc_ctx.isc_tx_nsegments, 3425 reclaim, thresh); 3426 3427 } 3428 #endif 3429 return (0); 3430 } 3431 iflib_tx_desc_free(txq, reclaim); 3432 txq->ift_cleaned += reclaim; 3433 txq->ift_in_use -= reclaim; 3434 3435 return (reclaim); 3436 } 3437 3438 static struct mbuf ** 3439 _ring_peek_one(struct ifmp_ring *r, int cidx, int offset, int remaining) 3440 { 3441 int next, size; 3442 struct mbuf **items; 3443 3444 size = r->size; 3445 next = (cidx + CACHE_PTR_INCREMENT) & (size-1); 3446 items = __DEVOLATILE(struct mbuf **, &r->items[0]); 3447 3448 prefetch(items[(cidx + offset) & (size-1)]); 3449 if (remaining > 1) { 3450 prefetch2cachelines(&items[next]); 3451 prefetch2cachelines(items[(cidx + offset + 1) & (size-1)]); 3452 prefetch2cachelines(items[(cidx + offset + 2) & (size-1)]); 3453 prefetch2cachelines(items[(cidx + offset + 3) & (size-1)]); 3454 } 3455 return (__DEVOLATILE(struct mbuf **, &r->items[(cidx + offset) & (size-1)])); 3456 } 3457 3458 static void 3459 iflib_txq_check_drain(iflib_txq_t txq, int budget) 3460 { 3461 3462 ifmp_ring_check_drainage(txq->ift_br, budget); 3463 } 3464 3465 static uint32_t 3466 iflib_txq_can_drain(struct ifmp_ring *r) 3467 { 3468 iflib_txq_t txq = r->cookie; 3469 if_ctx_t ctx = txq->ift_ctx; 3470 3471 return ((TXQ_AVAIL(txq) > MAX_TX_DESC(ctx) + 2) || 3472 ctx->isc_txd_credits_update(ctx->ifc_softc, txq->ift_id, false)); 3473 } 3474 3475 static uint32_t 3476 iflib_txq_drain(struct ifmp_ring *r, uint32_t cidx, uint32_t pidx) 3477 { 3478 iflib_txq_t txq = r->cookie; 3479 if_ctx_t ctx = txq->ift_ctx; 3480 struct ifnet *ifp = ctx->ifc_ifp; 3481 struct mbuf **mp, *m; 3482 int i, count, consumed, pkt_sent, bytes_sent, mcast_sent, avail; 3483 int reclaimed, err, in_use_prev, desc_used; 3484 bool do_prefetch, ring, rang; 3485 3486 if (__predict_false(!(if_getdrvflags(ifp) & IFF_DRV_RUNNING) || 3487 !LINK_ACTIVE(ctx))) { 3488 DBG_COUNTER_INC(txq_drain_notready); 3489 return (0); 3490 } 3491 reclaimed = iflib_completed_tx_reclaim(txq, RECLAIM_THRESH(ctx)); 3492 rang = iflib_txd_db_check(ctx, txq, reclaimed, txq->ift_in_use); 3493 avail = IDXDIFF(pidx, cidx, r->size); 3494 if (__predict_false(ctx->ifc_flags & IFC_QFLUSH)) { 3495 DBG_COUNTER_INC(txq_drain_flushing); 3496 for (i = 0; i < avail; i++) { 3497 m_free(r->items[(cidx + i) & (r->size-1)]); 3498 r->items[(cidx + i) & (r->size-1)] = NULL; 3499 } 3500 return (avail); 3501 } 3502 3503 if (__predict_false(if_getdrvflags(ctx->ifc_ifp) & IFF_DRV_OACTIVE)) { 3504 txq->ift_qstatus = IFLIB_QUEUE_IDLE; 3505 CALLOUT_LOCK(txq); 3506 callout_stop(&txq->ift_timer); 3507 CALLOUT_UNLOCK(txq); 3508 DBG_COUNTER_INC(txq_drain_oactive); 3509 return (0); 3510 } 3511 if (reclaimed) 3512 txq->ift_qstatus = IFLIB_QUEUE_IDLE; 3513 consumed = mcast_sent = bytes_sent = pkt_sent = 0; 3514 count = MIN(avail, TX_BATCH_SIZE); 3515 #ifdef INVARIANTS 3516 if (iflib_verbose_debug) 3517 printf("%s avail=%d ifc_flags=%x txq_avail=%d ", __FUNCTION__, 3518 avail, ctx->ifc_flags, TXQ_AVAIL(txq)); 3519 #endif 3520 do_prefetch = (ctx->ifc_flags & IFC_PREFETCH); 3521 avail = TXQ_AVAIL(txq); 3522 for (desc_used = i = 0; i < count && avail > MAX_TX_DESC(ctx) + 2; i++) { 3523 int pidx_prev, rem = do_prefetch ? count - i : 0; 3524 3525 mp = _ring_peek_one(r, cidx, i, rem); 3526 MPASS(mp != NULL && *mp != NULL); 3527 if (__predict_false(*mp == (struct mbuf *)txq)) { 3528 consumed++; 3529 reclaimed++; 3530 continue; 3531 } 3532 in_use_prev = txq->ift_in_use; 3533 pidx_prev = txq->ift_pidx; 3534 err = iflib_encap(txq, mp); 3535 if (__predict_false(err)) { 3536 DBG_COUNTER_INC(txq_drain_encapfail); 3537 /* no room - bail out */ 3538 if (err == ENOBUFS) 3539 break; 3540 consumed++; 3541 DBG_COUNTER_INC(txq_drain_encapfail); 3542 /* we can't send this packet - skip it */ 3543 continue; 3544 } 3545 consumed++; 3546 pkt_sent++; 3547 m = *mp; 3548 DBG_COUNTER_INC(tx_sent); 3549 bytes_sent += m->m_pkthdr.len; 3550 mcast_sent += !!(m->m_flags & M_MCAST); 3551 avail = TXQ_AVAIL(txq); 3552 3553 txq->ift_db_pending += (txq->ift_in_use - in_use_prev); 3554 desc_used += (txq->ift_in_use - in_use_prev); 3555 ETHER_BPF_MTAP(ifp, m); 3556 if (__predict_false(!(ifp->if_drv_flags & IFF_DRV_RUNNING))) 3557 break; 3558 rang = iflib_txd_db_check(ctx, txq, false, in_use_prev); 3559 } 3560 3561 /* deliberate use of bitwise or to avoid gratuitous short-circuit */ 3562 ring = rang ? false : (iflib_min_tx_latency | err) || (TXQ_AVAIL(txq) < MAX_TX_DESC(ctx)); 3563 iflib_txd_db_check(ctx, txq, ring, txq->ift_in_use); 3564 if_inc_counter(ifp, IFCOUNTER_OBYTES, bytes_sent); 3565 if_inc_counter(ifp, IFCOUNTER_OPACKETS, pkt_sent); 3566 if (mcast_sent) 3567 if_inc_counter(ifp, IFCOUNTER_OMCASTS, mcast_sent); 3568 #ifdef INVARIANTS 3569 if (iflib_verbose_debug) 3570 printf("consumed=%d\n", consumed); 3571 #endif 3572 return (consumed); 3573 } 3574 3575 static uint32_t 3576 iflib_txq_drain_always(struct ifmp_ring *r) 3577 { 3578 return (1); 3579 } 3580 3581 static uint32_t 3582 iflib_txq_drain_free(struct ifmp_ring *r, uint32_t cidx, uint32_t pidx) 3583 { 3584 int i, avail; 3585 struct mbuf **mp; 3586 iflib_txq_t txq; 3587 3588 txq = r->cookie; 3589 3590 txq->ift_qstatus = IFLIB_QUEUE_IDLE; 3591 CALLOUT_LOCK(txq); 3592 callout_stop(&txq->ift_timer); 3593 CALLOUT_UNLOCK(txq); 3594 3595 avail = IDXDIFF(pidx, cidx, r->size); 3596 for (i = 0; i < avail; i++) { 3597 mp = _ring_peek_one(r, cidx, i, avail - i); 3598 if (__predict_false(*mp == (struct mbuf *)txq)) 3599 continue; 3600 m_freem(*mp); 3601 } 3602 MPASS(ifmp_ring_is_stalled(r) == 0); 3603 return (avail); 3604 } 3605 3606 static void 3607 iflib_ifmp_purge(iflib_txq_t txq) 3608 { 3609 struct ifmp_ring *r; 3610 3611 r = txq->ift_br; 3612 r->drain = iflib_txq_drain_free; 3613 r->can_drain = iflib_txq_drain_always; 3614 3615 ifmp_ring_check_drainage(r, r->size); 3616 3617 r->drain = iflib_txq_drain; 3618 r->can_drain = iflib_txq_can_drain; 3619 } 3620 3621 static void 3622 _task_fn_tx(void *context) 3623 { 3624 iflib_txq_t txq = context; 3625 if_ctx_t ctx = txq->ift_ctx; 3626 struct ifnet *ifp = ctx->ifc_ifp; 3627 int rc; 3628 3629 #ifdef IFLIB_DIAGNOSTICS 3630 txq->ift_cpu_exec_count[curcpu]++; 3631 #endif 3632 if (!(if_getdrvflags(ctx->ifc_ifp) & IFF_DRV_RUNNING)) 3633 return; 3634 if (if_getcapenable(ifp) & IFCAP_NETMAP) { 3635 if (ctx->isc_txd_credits_update(ctx->ifc_softc, txq->ift_id, false)) 3636 netmap_tx_irq(ifp, txq->ift_id); 3637 IFDI_TX_QUEUE_INTR_ENABLE(ctx, txq->ift_id); 3638 return; 3639 } 3640 if (txq->ift_db_pending) 3641 ifmp_ring_enqueue(txq->ift_br, (void **)&txq, 1, TX_BATCH_SIZE); 3642 ifmp_ring_check_drainage(txq->ift_br, TX_BATCH_SIZE); 3643 if (ctx->ifc_flags & IFC_LEGACY) 3644 IFDI_INTR_ENABLE(ctx); 3645 else { 3646 rc = IFDI_TX_QUEUE_INTR_ENABLE(ctx, txq->ift_id); 3647 KASSERT(rc != ENOTSUP, ("MSI-X support requires queue_intr_enable, but not implemented in driver")); 3648 } 3649 } 3650 3651 static void 3652 _task_fn_rx(void *context) 3653 { 3654 iflib_rxq_t rxq = context; 3655 if_ctx_t ctx = rxq->ifr_ctx; 3656 bool more; 3657 int rc; 3658 uint16_t budget; 3659 3660 #ifdef IFLIB_DIAGNOSTICS 3661 rxq->ifr_cpu_exec_count[curcpu]++; 3662 #endif 3663 DBG_COUNTER_INC(task_fn_rxs); 3664 if (__predict_false(!(if_getdrvflags(ctx->ifc_ifp) & IFF_DRV_RUNNING))) 3665 return; 3666 more = true; 3667 #ifdef DEV_NETMAP 3668 if (if_getcapenable(ctx->ifc_ifp) & IFCAP_NETMAP) { 3669 u_int work = 0; 3670 if (netmap_rx_irq(ctx->ifc_ifp, rxq->ifr_id, &work)) { 3671 more = false; 3672 } 3673 } 3674 #endif 3675 budget = ctx->ifc_sysctl_rx_budget; 3676 if (budget == 0) 3677 budget = 16; /* XXX */ 3678 if (more == false || (more = iflib_rxeof(rxq, budget)) == false) { 3679 if (ctx->ifc_flags & IFC_LEGACY) 3680 IFDI_INTR_ENABLE(ctx); 3681 else { 3682 DBG_COUNTER_INC(rx_intr_enables); 3683 rc = IFDI_RX_QUEUE_INTR_ENABLE(ctx, rxq->ifr_id); 3684 KASSERT(rc != ENOTSUP, ("MSI-X support requires queue_intr_enable, but not implemented in driver")); 3685 } 3686 } 3687 if (__predict_false(!(if_getdrvflags(ctx->ifc_ifp) & IFF_DRV_RUNNING))) 3688 return; 3689 if (more) 3690 GROUPTASK_ENQUEUE(&rxq->ifr_task); 3691 } 3692 3693 static void 3694 _task_fn_admin(void *context) 3695 { 3696 if_ctx_t ctx = context; 3697 if_softc_ctx_t sctx = &ctx->ifc_softc_ctx; 3698 iflib_txq_t txq; 3699 int i; 3700 3701 if (!(if_getdrvflags(ctx->ifc_ifp) & IFF_DRV_RUNNING)) { 3702 if (!(if_getdrvflags(ctx->ifc_ifp) & IFF_DRV_OACTIVE)) { 3703 return; 3704 } 3705 } 3706 3707 CTX_LOCK(ctx); 3708 for (txq = ctx->ifc_txqs, i = 0; i < sctx->isc_ntxqsets; i++, txq++) { 3709 CALLOUT_LOCK(txq); 3710 callout_stop(&txq->ift_timer); 3711 CALLOUT_UNLOCK(txq); 3712 } 3713 IFDI_UPDATE_ADMIN_STATUS(ctx); 3714 for (txq = ctx->ifc_txqs, i = 0; i < sctx->isc_ntxqsets; i++, txq++) 3715 callout_reset_on(&txq->ift_timer, hz/2, iflib_timer, txq, txq->ift_timer.c_cpu); 3716 IFDI_LINK_INTR_ENABLE(ctx); 3717 if (ctx->ifc_flags & IFC_DO_RESET) { 3718 ctx->ifc_flags &= ~IFC_DO_RESET; 3719 iflib_if_init_locked(ctx); 3720 } 3721 CTX_UNLOCK(ctx); 3722 3723 if (LINK_ACTIVE(ctx) == 0) 3724 return; 3725 for (txq = ctx->ifc_txqs, i = 0; i < sctx->isc_ntxqsets; i++, txq++) 3726 iflib_txq_check_drain(txq, IFLIB_RESTART_BUDGET); 3727 } 3728 3729 3730 static void 3731 _task_fn_iov(void *context) 3732 { 3733 if_ctx_t ctx = context; 3734 3735 if (!(if_getdrvflags(ctx->ifc_ifp) & IFF_DRV_RUNNING)) 3736 return; 3737 3738 CTX_LOCK(ctx); 3739 IFDI_VFLR_HANDLE(ctx); 3740 CTX_UNLOCK(ctx); 3741 } 3742 3743 static int 3744 iflib_sysctl_int_delay(SYSCTL_HANDLER_ARGS) 3745 { 3746 int err; 3747 if_int_delay_info_t info; 3748 if_ctx_t ctx; 3749 3750 info = (if_int_delay_info_t)arg1; 3751 ctx = info->iidi_ctx; 3752 info->iidi_req = req; 3753 info->iidi_oidp = oidp; 3754 CTX_LOCK(ctx); 3755 err = IFDI_SYSCTL_INT_DELAY(ctx, info); 3756 CTX_UNLOCK(ctx); 3757 return (err); 3758 } 3759 3760 /********************************************************************* 3761 * 3762 * IFNET FUNCTIONS 3763 * 3764 **********************************************************************/ 3765 3766 static void 3767 iflib_if_init_locked(if_ctx_t ctx) 3768 { 3769 iflib_stop(ctx); 3770 iflib_init_locked(ctx); 3771 } 3772 3773 3774 static void 3775 iflib_if_init(void *arg) 3776 { 3777 if_ctx_t ctx = arg; 3778 3779 CTX_LOCK(ctx); 3780 iflib_if_init_locked(ctx); 3781 CTX_UNLOCK(ctx); 3782 } 3783 3784 static int 3785 iflib_if_transmit(if_t ifp, struct mbuf *m) 3786 { 3787 if_ctx_t ctx = if_getsoftc(ifp); 3788 3789 iflib_txq_t txq; 3790 int err, qidx; 3791 3792 if (__predict_false((ifp->if_drv_flags & IFF_DRV_RUNNING) == 0 || !LINK_ACTIVE(ctx))) { 3793 DBG_COUNTER_INC(tx_frees); 3794 m_freem(m); 3795 return (ENOBUFS); 3796 } 3797 3798 MPASS(m->m_nextpkt == NULL); 3799 qidx = 0; 3800 if ((NTXQSETS(ctx) > 1) && M_HASHTYPE_GET(m)) 3801 qidx = QIDX(ctx, m); 3802 /* 3803 * XXX calculate buf_ring based on flowid (divvy up bits?) 3804 */ 3805 txq = &ctx->ifc_txqs[qidx]; 3806 3807 #ifdef DRIVER_BACKPRESSURE 3808 if (txq->ift_closed) { 3809 while (m != NULL) { 3810 next = m->m_nextpkt; 3811 m->m_nextpkt = NULL; 3812 m_freem(m); 3813 m = next; 3814 } 3815 return (ENOBUFS); 3816 } 3817 #endif 3818 #ifdef notyet 3819 qidx = count = 0; 3820 mp = marr; 3821 next = m; 3822 do { 3823 count++; 3824 next = next->m_nextpkt; 3825 } while (next != NULL); 3826 3827 if (count > nitems(marr)) 3828 if ((mp = malloc(count*sizeof(struct mbuf *), M_IFLIB, M_NOWAIT)) == NULL) { 3829 /* XXX check nextpkt */ 3830 m_freem(m); 3831 /* XXX simplify for now */ 3832 DBG_COUNTER_INC(tx_frees); 3833 return (ENOBUFS); 3834 } 3835 for (next = m, i = 0; next != NULL; i++) { 3836 mp[i] = next; 3837 next = next->m_nextpkt; 3838 mp[i]->m_nextpkt = NULL; 3839 } 3840 #endif 3841 DBG_COUNTER_INC(tx_seen); 3842 err = ifmp_ring_enqueue(txq->ift_br, (void **)&m, 1, TX_BATCH_SIZE); 3843 3844 GROUPTASK_ENQUEUE(&txq->ift_task); 3845 if (err) { 3846 /* support forthcoming later */ 3847 #ifdef DRIVER_BACKPRESSURE 3848 txq->ift_closed = TRUE; 3849 #endif 3850 ifmp_ring_check_drainage(txq->ift_br, TX_BATCH_SIZE); 3851 m_freem(m); 3852 } 3853 3854 return (err); 3855 } 3856 3857 static void 3858 iflib_if_qflush(if_t ifp) 3859 { 3860 if_ctx_t ctx = if_getsoftc(ifp); 3861 iflib_txq_t txq = ctx->ifc_txqs; 3862 int i; 3863 3864 CTX_LOCK(ctx); 3865 ctx->ifc_flags |= IFC_QFLUSH; 3866 CTX_UNLOCK(ctx); 3867 for (i = 0; i < NTXQSETS(ctx); i++, txq++) 3868 while (!(ifmp_ring_is_idle(txq->ift_br) || ifmp_ring_is_stalled(txq->ift_br))) 3869 iflib_txq_check_drain(txq, 0); 3870 CTX_LOCK(ctx); 3871 ctx->ifc_flags &= ~IFC_QFLUSH; 3872 CTX_UNLOCK(ctx); 3873 3874 if_qflush(ifp); 3875 } 3876 3877 3878 #define IFCAP_FLAGS (IFCAP_TXCSUM_IPV6 | IFCAP_RXCSUM_IPV6 | IFCAP_HWCSUM | IFCAP_LRO | \ 3879 IFCAP_TSO4 | IFCAP_TSO6 | IFCAP_VLAN_HWTAGGING | IFCAP_HWSTATS | \ 3880 IFCAP_VLAN_MTU | IFCAP_VLAN_HWFILTER | IFCAP_VLAN_HWTSO) 3881 3882 static int 3883 iflib_if_ioctl(if_t ifp, u_long command, caddr_t data) 3884 { 3885 if_ctx_t ctx = if_getsoftc(ifp); 3886 struct ifreq *ifr = (struct ifreq *)data; 3887 #if defined(INET) || defined(INET6) 3888 struct ifaddr *ifa = (struct ifaddr *)data; 3889 #endif 3890 bool avoid_reset = FALSE; 3891 int err = 0, reinit = 0, bits; 3892 3893 switch (command) { 3894 case SIOCSIFADDR: 3895 #ifdef INET 3896 if (ifa->ifa_addr->sa_family == AF_INET) 3897 avoid_reset = TRUE; 3898 #endif 3899 #ifdef INET6 3900 if (ifa->ifa_addr->sa_family == AF_INET6) 3901 avoid_reset = TRUE; 3902 #endif 3903 /* 3904 ** Calling init results in link renegotiation, 3905 ** so we avoid doing it when possible. 3906 */ 3907 if (avoid_reset) { 3908 if_setflagbits(ifp, IFF_UP,0); 3909 if (!(if_getdrvflags(ifp)& IFF_DRV_RUNNING)) 3910 reinit = 1; 3911 #ifdef INET 3912 if (!(if_getflags(ifp) & IFF_NOARP)) 3913 arp_ifinit(ifp, ifa); 3914 #endif 3915 } else 3916 err = ether_ioctl(ifp, command, data); 3917 break; 3918 case SIOCSIFMTU: 3919 CTX_LOCK(ctx); 3920 if (ifr->ifr_mtu == if_getmtu(ifp)) { 3921 CTX_UNLOCK(ctx); 3922 break; 3923 } 3924 bits = if_getdrvflags(ifp); 3925 /* stop the driver and free any clusters before proceeding */ 3926 iflib_stop(ctx); 3927 3928 if ((err = IFDI_MTU_SET(ctx, ifr->ifr_mtu)) == 0) { 3929 if (ifr->ifr_mtu > ctx->ifc_max_fl_buf_size) 3930 ctx->ifc_flags |= IFC_MULTISEG; 3931 else 3932 ctx->ifc_flags &= ~IFC_MULTISEG; 3933 err = if_setmtu(ifp, ifr->ifr_mtu); 3934 } 3935 iflib_init_locked(ctx); 3936 if_setdrvflags(ifp, bits); 3937 CTX_UNLOCK(ctx); 3938 break; 3939 case SIOCSIFFLAGS: 3940 CTX_LOCK(ctx); 3941 if (if_getflags(ifp) & IFF_UP) { 3942 if (if_getdrvflags(ifp) & IFF_DRV_RUNNING) { 3943 if ((if_getflags(ifp) ^ ctx->ifc_if_flags) & 3944 (IFF_PROMISC | IFF_ALLMULTI)) { 3945 err = IFDI_PROMISC_SET(ctx, if_getflags(ifp)); 3946 } 3947 } else 3948 reinit = 1; 3949 } else if (if_getdrvflags(ifp) & IFF_DRV_RUNNING) { 3950 iflib_stop(ctx); 3951 } 3952 ctx->ifc_if_flags = if_getflags(ifp); 3953 CTX_UNLOCK(ctx); 3954 break; 3955 case SIOCADDMULTI: 3956 case SIOCDELMULTI: 3957 if (if_getdrvflags(ifp) & IFF_DRV_RUNNING) { 3958 CTX_LOCK(ctx); 3959 IFDI_INTR_DISABLE(ctx); 3960 IFDI_MULTI_SET(ctx); 3961 IFDI_INTR_ENABLE(ctx); 3962 CTX_UNLOCK(ctx); 3963 } 3964 break; 3965 case SIOCSIFMEDIA: 3966 CTX_LOCK(ctx); 3967 IFDI_MEDIA_SET(ctx); 3968 CTX_UNLOCK(ctx); 3969 /* falls thru */ 3970 case SIOCGIFMEDIA: 3971 case SIOCGIFXMEDIA: 3972 err = ifmedia_ioctl(ifp, ifr, &ctx->ifc_media, command); 3973 break; 3974 case SIOCGI2C: 3975 { 3976 struct ifi2creq i2c; 3977 3978 err = copyin(ifr->ifr_data, &i2c, sizeof(i2c)); 3979 if (err != 0) 3980 break; 3981 if (i2c.dev_addr != 0xA0 && i2c.dev_addr != 0xA2) { 3982 err = EINVAL; 3983 break; 3984 } 3985 if (i2c.len > sizeof(i2c.data)) { 3986 err = EINVAL; 3987 break; 3988 } 3989 3990 if ((err = IFDI_I2C_REQ(ctx, &i2c)) == 0) 3991 err = copyout(&i2c, ifr->ifr_data, sizeof(i2c)); 3992 break; 3993 } 3994 case SIOCSIFCAP: 3995 { 3996 int mask, setmask; 3997 3998 mask = ifr->ifr_reqcap ^ if_getcapenable(ifp); 3999 setmask = 0; 4000 #ifdef TCP_OFFLOAD 4001 setmask |= mask & (IFCAP_TOE4|IFCAP_TOE6); 4002 #endif 4003 setmask |= (mask & IFCAP_FLAGS); 4004 4005 if (setmask & (IFCAP_RXCSUM | IFCAP_RXCSUM_IPV6)) 4006 setmask |= (IFCAP_RXCSUM | IFCAP_RXCSUM_IPV6); 4007 if ((mask & IFCAP_WOL) && 4008 (if_getcapabilities(ifp) & IFCAP_WOL) != 0) 4009 setmask |= (mask & (IFCAP_WOL_MCAST|IFCAP_WOL_MAGIC)); 4010 if_vlancap(ifp); 4011 /* 4012 * want to ensure that traffic has stopped before we change any of the flags 4013 */ 4014 if (setmask) { 4015 CTX_LOCK(ctx); 4016 bits = if_getdrvflags(ifp); 4017 if (bits & IFF_DRV_RUNNING) 4018 iflib_stop(ctx); 4019 if_togglecapenable(ifp, setmask); 4020 if (bits & IFF_DRV_RUNNING) 4021 iflib_init_locked(ctx); 4022 if_setdrvflags(ifp, bits); 4023 CTX_UNLOCK(ctx); 4024 } 4025 break; 4026 } 4027 case SIOCGPRIVATE_0: 4028 case SIOCSDRVSPEC: 4029 case SIOCGDRVSPEC: 4030 CTX_LOCK(ctx); 4031 err = IFDI_PRIV_IOCTL(ctx, command, data); 4032 CTX_UNLOCK(ctx); 4033 break; 4034 default: 4035 err = ether_ioctl(ifp, command, data); 4036 break; 4037 } 4038 if (reinit) 4039 iflib_if_init(ctx); 4040 return (err); 4041 } 4042 4043 static uint64_t 4044 iflib_if_get_counter(if_t ifp, ift_counter cnt) 4045 { 4046 if_ctx_t ctx = if_getsoftc(ifp); 4047 4048 return (IFDI_GET_COUNTER(ctx, cnt)); 4049 } 4050 4051 /********************************************************************* 4052 * 4053 * OTHER FUNCTIONS EXPORTED TO THE STACK 4054 * 4055 **********************************************************************/ 4056 4057 static void 4058 iflib_vlan_register(void *arg, if_t ifp, uint16_t vtag) 4059 { 4060 if_ctx_t ctx = if_getsoftc(ifp); 4061 4062 if ((void *)ctx != arg) 4063 return; 4064 4065 if ((vtag == 0) || (vtag > 4095)) 4066 return; 4067 4068 CTX_LOCK(ctx); 4069 IFDI_VLAN_REGISTER(ctx, vtag); 4070 /* Re-init to load the changes */ 4071 if (if_getcapenable(ifp) & IFCAP_VLAN_HWFILTER) 4072 iflib_if_init_locked(ctx); 4073 CTX_UNLOCK(ctx); 4074 } 4075 4076 static void 4077 iflib_vlan_unregister(void *arg, if_t ifp, uint16_t vtag) 4078 { 4079 if_ctx_t ctx = if_getsoftc(ifp); 4080 4081 if ((void *)ctx != arg) 4082 return; 4083 4084 if ((vtag == 0) || (vtag > 4095)) 4085 return; 4086 4087 CTX_LOCK(ctx); 4088 IFDI_VLAN_UNREGISTER(ctx, vtag); 4089 /* Re-init to load the changes */ 4090 if (if_getcapenable(ifp) & IFCAP_VLAN_HWFILTER) 4091 iflib_if_init_locked(ctx); 4092 CTX_UNLOCK(ctx); 4093 } 4094 4095 static void 4096 iflib_led_func(void *arg, int onoff) 4097 { 4098 if_ctx_t ctx = arg; 4099 4100 CTX_LOCK(ctx); 4101 IFDI_LED_FUNC(ctx, onoff); 4102 CTX_UNLOCK(ctx); 4103 } 4104 4105 /********************************************************************* 4106 * 4107 * BUS FUNCTION DEFINITIONS 4108 * 4109 **********************************************************************/ 4110 4111 int 4112 iflib_device_probe(device_t dev) 4113 { 4114 pci_vendor_info_t *ent; 4115 4116 uint16_t pci_vendor_id, pci_device_id; 4117 uint16_t pci_subvendor_id, pci_subdevice_id; 4118 uint16_t pci_rev_id; 4119 if_shared_ctx_t sctx; 4120 4121 if ((sctx = DEVICE_REGISTER(dev)) == NULL || sctx->isc_magic != IFLIB_MAGIC) 4122 return (ENOTSUP); 4123 4124 pci_vendor_id = pci_get_vendor(dev); 4125 pci_device_id = pci_get_device(dev); 4126 pci_subvendor_id = pci_get_subvendor(dev); 4127 pci_subdevice_id = pci_get_subdevice(dev); 4128 pci_rev_id = pci_get_revid(dev); 4129 if (sctx->isc_parse_devinfo != NULL) 4130 sctx->isc_parse_devinfo(&pci_device_id, &pci_subvendor_id, &pci_subdevice_id, &pci_rev_id); 4131 4132 ent = sctx->isc_vendor_info; 4133 while (ent->pvi_vendor_id != 0) { 4134 if (pci_vendor_id != ent->pvi_vendor_id) { 4135 ent++; 4136 continue; 4137 } 4138 if ((pci_device_id == ent->pvi_device_id) && 4139 ((pci_subvendor_id == ent->pvi_subvendor_id) || 4140 (ent->pvi_subvendor_id == 0)) && 4141 ((pci_subdevice_id == ent->pvi_subdevice_id) || 4142 (ent->pvi_subdevice_id == 0)) && 4143 ((pci_rev_id == ent->pvi_rev_id) || 4144 (ent->pvi_rev_id == 0))) { 4145 4146 device_set_desc_copy(dev, ent->pvi_name); 4147 /* this needs to be changed to zero if the bus probing code 4148 * ever stops re-probing on best match because the sctx 4149 * may have its values over written by register calls 4150 * in subsequent probes 4151 */ 4152 return (BUS_PROBE_DEFAULT); 4153 } 4154 ent++; 4155 } 4156 return (ENXIO); 4157 } 4158 4159 int 4160 iflib_device_register(device_t dev, void *sc, if_shared_ctx_t sctx, if_ctx_t *ctxp) 4161 { 4162 int err, rid, msix, msix_bar; 4163 if_ctx_t ctx; 4164 if_t ifp; 4165 if_softc_ctx_t scctx; 4166 int i; 4167 uint16_t main_txq; 4168 uint16_t main_rxq; 4169 4170 4171 ctx = malloc(sizeof(* ctx), M_IFLIB, M_WAITOK|M_ZERO); 4172 4173 if (sc == NULL) { 4174 sc = malloc(sctx->isc_driver->size, M_IFLIB, M_WAITOK|M_ZERO); 4175 device_set_softc(dev, ctx); 4176 ctx->ifc_flags |= IFC_SC_ALLOCATED; 4177 } 4178 4179 ctx->ifc_sctx = sctx; 4180 ctx->ifc_dev = dev; 4181 ctx->ifc_softc = sc; 4182 4183 if ((err = iflib_register(ctx)) != 0) { 4184 device_printf(dev, "iflib_register failed %d\n", err); 4185 return (err); 4186 } 4187 iflib_add_device_sysctl_pre(ctx); 4188 4189 scctx = &ctx->ifc_softc_ctx; 4190 ifp = ctx->ifc_ifp; 4191 4192 /* 4193 * XXX sanity check that ntxd & nrxd are a power of 2 4194 */ 4195 if (ctx->ifc_sysctl_ntxqs != 0) 4196 scctx->isc_ntxqsets = ctx->ifc_sysctl_ntxqs; 4197 if (ctx->ifc_sysctl_nrxqs != 0) 4198 scctx->isc_nrxqsets = ctx->ifc_sysctl_nrxqs; 4199 4200 for (i = 0; i < sctx->isc_ntxqs; i++) { 4201 if (ctx->ifc_sysctl_ntxds[i] != 0) 4202 scctx->isc_ntxd[i] = ctx->ifc_sysctl_ntxds[i]; 4203 else 4204 scctx->isc_ntxd[i] = sctx->isc_ntxd_default[i]; 4205 } 4206 4207 for (i = 0; i < sctx->isc_nrxqs; i++) { 4208 if (ctx->ifc_sysctl_nrxds[i] != 0) 4209 scctx->isc_nrxd[i] = ctx->ifc_sysctl_nrxds[i]; 4210 else 4211 scctx->isc_nrxd[i] = sctx->isc_nrxd_default[i]; 4212 } 4213 4214 for (i = 0; i < sctx->isc_nrxqs; i++) { 4215 if (scctx->isc_nrxd[i] < sctx->isc_nrxd_min[i]) { 4216 device_printf(dev, "nrxd%d: %d less than nrxd_min %d - resetting to min\n", 4217 i, scctx->isc_nrxd[i], sctx->isc_nrxd_min[i]); 4218 scctx->isc_nrxd[i] = sctx->isc_nrxd_min[i]; 4219 } 4220 if (scctx->isc_nrxd[i] > sctx->isc_nrxd_max[i]) { 4221 device_printf(dev, "nrxd%d: %d greater than nrxd_max %d - resetting to max\n", 4222 i, scctx->isc_nrxd[i], sctx->isc_nrxd_max[i]); 4223 scctx->isc_nrxd[i] = sctx->isc_nrxd_max[i]; 4224 } 4225 } 4226 4227 for (i = 0; i < sctx->isc_ntxqs; i++) { 4228 if (scctx->isc_ntxd[i] < sctx->isc_ntxd_min[i]) { 4229 device_printf(dev, "ntxd%d: %d less than ntxd_min %d - resetting to min\n", 4230 i, scctx->isc_ntxd[i], sctx->isc_ntxd_min[i]); 4231 scctx->isc_ntxd[i] = sctx->isc_ntxd_min[i]; 4232 } 4233 if (scctx->isc_ntxd[i] > sctx->isc_ntxd_max[i]) { 4234 device_printf(dev, "ntxd%d: %d greater than ntxd_max %d - resetting to max\n", 4235 i, scctx->isc_ntxd[i], sctx->isc_ntxd_max[i]); 4236 scctx->isc_ntxd[i] = sctx->isc_ntxd_max[i]; 4237 } 4238 } 4239 4240 if ((err = IFDI_ATTACH_PRE(ctx)) != 0) { 4241 device_printf(dev, "IFDI_ATTACH_PRE failed %d\n", err); 4242 return (err); 4243 } 4244 _iflib_pre_assert(scctx); 4245 ctx->ifc_txrx = *scctx->isc_txrx; 4246 4247 #ifdef INVARIANTS 4248 MPASS(scctx->isc_capenable); 4249 if (scctx->isc_capenable & IFCAP_TXCSUM) 4250 MPASS(scctx->isc_tx_csum_flags); 4251 #endif 4252 4253 if_setcapabilities(ifp, scctx->isc_capenable | IFCAP_HWSTATS); 4254 if_setcapenable(ifp, scctx->isc_capenable | IFCAP_HWSTATS); 4255 4256 if (scctx->isc_ntxqsets == 0 || (scctx->isc_ntxqsets_max && scctx->isc_ntxqsets_max < scctx->isc_ntxqsets)) 4257 scctx->isc_ntxqsets = scctx->isc_ntxqsets_max; 4258 if (scctx->isc_nrxqsets == 0 || (scctx->isc_nrxqsets_max && scctx->isc_nrxqsets_max < scctx->isc_nrxqsets)) 4259 scctx->isc_nrxqsets = scctx->isc_nrxqsets_max; 4260 4261 #ifdef ACPI_DMAR 4262 if (dmar_get_dma_tag(device_get_parent(dev), dev) != NULL) 4263 ctx->ifc_flags |= IFC_DMAR; 4264 #elif !(defined(__i386__) || defined(__amd64__)) 4265 /* set unconditionally for !x86 */ 4266 ctx->ifc_flags |= IFC_DMAR; 4267 #endif 4268 4269 msix_bar = scctx->isc_msix_bar; 4270 main_txq = (sctx->isc_flags & IFLIB_HAS_TXCQ) ? 1 : 0; 4271 main_rxq = (sctx->isc_flags & IFLIB_HAS_RXCQ) ? 1 : 0; 4272 4273 /* XXX change for per-queue sizes */ 4274 device_printf(dev, "using %d tx descriptors and %d rx descriptors\n", 4275 scctx->isc_ntxd[main_txq], scctx->isc_nrxd[main_rxq]); 4276 for (i = 0; i < sctx->isc_nrxqs; i++) { 4277 if (!powerof2(scctx->isc_nrxd[i])) { 4278 /* round down instead? */ 4279 device_printf(dev, "# rx descriptors must be a power of 2\n"); 4280 err = EINVAL; 4281 goto fail; 4282 } 4283 } 4284 for (i = 0; i < sctx->isc_ntxqs; i++) { 4285 if (!powerof2(scctx->isc_ntxd[i])) { 4286 device_printf(dev, 4287 "# tx descriptors must be a power of 2"); 4288 err = EINVAL; 4289 goto fail; 4290 } 4291 } 4292 4293 if (scctx->isc_tx_nsegments > scctx->isc_ntxd[main_txq] / 4294 MAX_SINGLE_PACKET_FRACTION) 4295 scctx->isc_tx_nsegments = max(1, scctx->isc_ntxd[main_txq] / 4296 MAX_SINGLE_PACKET_FRACTION); 4297 if (scctx->isc_tx_tso_segments_max > scctx->isc_ntxd[main_txq] / 4298 MAX_SINGLE_PACKET_FRACTION) 4299 scctx->isc_tx_tso_segments_max = max(1, 4300 scctx->isc_ntxd[main_txq] / MAX_SINGLE_PACKET_FRACTION); 4301 4302 /* 4303 * Protect the stack against modern hardware 4304 */ 4305 if (scctx->isc_tx_tso_size_max > FREEBSD_TSO_SIZE_MAX) 4306 scctx->isc_tx_tso_size_max = FREEBSD_TSO_SIZE_MAX; 4307 4308 /* TSO parameters - dig these out of the data sheet - simply correspond to tag setup */ 4309 ifp->if_hw_tsomaxsegcount = scctx->isc_tx_tso_segments_max; 4310 ifp->if_hw_tsomax = scctx->isc_tx_tso_size_max; 4311 ifp->if_hw_tsomaxsegsize = scctx->isc_tx_tso_segsize_max; 4312 if (scctx->isc_rss_table_size == 0) 4313 scctx->isc_rss_table_size = 64; 4314 scctx->isc_rss_table_mask = scctx->isc_rss_table_size-1; 4315 4316 GROUPTASK_INIT(&ctx->ifc_admin_task, 0, _task_fn_admin, ctx); 4317 /* XXX format name */ 4318 taskqgroup_attach(qgroup_if_config_tqg, &ctx->ifc_admin_task, ctx, -1, "admin"); 4319 4320 /* Set up cpu set. If it fails, use the set of all CPUs. */ 4321 if (bus_get_cpus(dev, INTR_CPUS, sizeof(ctx->ifc_cpus), &ctx->ifc_cpus) != 0) { 4322 device_printf(dev, "Unable to fetch CPU list\n"); 4323 CPU_COPY(&all_cpus, &ctx->ifc_cpus); 4324 } 4325 MPASS(CPU_COUNT(&ctx->ifc_cpus) > 0); 4326 4327 /* 4328 ** Now setup MSI or MSI/X, should 4329 ** return us the number of supported 4330 ** vectors. (Will be 1 for MSI) 4331 */ 4332 if (sctx->isc_flags & IFLIB_SKIP_MSIX) { 4333 msix = scctx->isc_vectors; 4334 } else if (scctx->isc_msix_bar != 0) 4335 /* 4336 * The simple fact that isc_msix_bar is not 0 does not mean we 4337 * we have a good value there that is known to work. 4338 */ 4339 msix = iflib_msix_init(ctx); 4340 else { 4341 scctx->isc_vectors = 1; 4342 scctx->isc_ntxqsets = 1; 4343 scctx->isc_nrxqsets = 1; 4344 scctx->isc_intr = IFLIB_INTR_LEGACY; 4345 msix = 0; 4346 } 4347 /* Get memory for the station queues */ 4348 if ((err = iflib_queues_alloc(ctx))) { 4349 device_printf(dev, "Unable to allocate queue memory\n"); 4350 goto fail; 4351 } 4352 4353 if ((err = iflib_qset_structures_setup(ctx))) { 4354 device_printf(dev, "qset structure setup failed %d\n", err); 4355 goto fail_queues; 4356 } 4357 4358 /* 4359 * Group taskqueues aren't properly set up until SMP is started, 4360 * so we disable interrupts until we can handle them post 4361 * SI_SUB_SMP. 4362 * 4363 * XXX: disabling interrupts doesn't actually work, at least for 4364 * the non-MSI case. When they occur before SI_SUB_SMP completes, 4365 * we do null handling and depend on this not causing too large an 4366 * interrupt storm. 4367 */ 4368 IFDI_INTR_DISABLE(ctx); 4369 if (msix > 1 && (err = IFDI_MSIX_INTR_ASSIGN(ctx, msix)) != 0) { 4370 device_printf(dev, "IFDI_MSIX_INTR_ASSIGN failed %d\n", err); 4371 goto fail_intr_free; 4372 } 4373 if (msix <= 1) { 4374 rid = 0; 4375 if (scctx->isc_intr == IFLIB_INTR_MSI) { 4376 MPASS(msix == 1); 4377 rid = 1; 4378 } 4379 if ((err = iflib_legacy_setup(ctx, ctx->isc_legacy_intr, ctx->ifc_softc, &rid, "irq0")) != 0) { 4380 device_printf(dev, "iflib_legacy_setup failed %d\n", err); 4381 goto fail_intr_free; 4382 } 4383 } 4384 ether_ifattach(ctx->ifc_ifp, ctx->ifc_mac); 4385 if ((err = IFDI_ATTACH_POST(ctx)) != 0) { 4386 device_printf(dev, "IFDI_ATTACH_POST failed %d\n", err); 4387 goto fail_detach; 4388 } 4389 if ((err = iflib_netmap_attach(ctx))) { 4390 device_printf(ctx->ifc_dev, "netmap attach failed: %d\n", err); 4391 goto fail_detach; 4392 } 4393 *ctxp = ctx; 4394 4395 if_setgetcounterfn(ctx->ifc_ifp, iflib_if_get_counter); 4396 iflib_add_device_sysctl_post(ctx); 4397 ctx->ifc_flags |= IFC_INIT_DONE; 4398 return (0); 4399 fail_detach: 4400 ether_ifdetach(ctx->ifc_ifp); 4401 fail_intr_free: 4402 if (scctx->isc_intr == IFLIB_INTR_MSIX || scctx->isc_intr == IFLIB_INTR_MSI) 4403 pci_release_msi(ctx->ifc_dev); 4404 fail_queues: 4405 /* XXX free queues */ 4406 fail: 4407 IFDI_DETACH(ctx); 4408 return (err); 4409 } 4410 4411 int 4412 iflib_device_attach(device_t dev) 4413 { 4414 if_ctx_t ctx; 4415 if_shared_ctx_t sctx; 4416 4417 if ((sctx = DEVICE_REGISTER(dev)) == NULL || sctx->isc_magic != IFLIB_MAGIC) 4418 return (ENOTSUP); 4419 4420 pci_enable_busmaster(dev); 4421 4422 return (iflib_device_register(dev, NULL, sctx, &ctx)); 4423 } 4424 4425 int 4426 iflib_device_deregister(if_ctx_t ctx) 4427 { 4428 if_t ifp = ctx->ifc_ifp; 4429 iflib_txq_t txq; 4430 iflib_rxq_t rxq; 4431 device_t dev = ctx->ifc_dev; 4432 int i, j; 4433 struct taskqgroup *tqg; 4434 iflib_fl_t fl; 4435 4436 /* Make sure VLANS are not using driver */ 4437 if (if_vlantrunkinuse(ifp)) { 4438 device_printf(dev,"Vlan in use, detach first\n"); 4439 return (EBUSY); 4440 } 4441 4442 CTX_LOCK(ctx); 4443 ctx->ifc_in_detach = 1; 4444 iflib_stop(ctx); 4445 CTX_UNLOCK(ctx); 4446 4447 /* Unregister VLAN events */ 4448 if (ctx->ifc_vlan_attach_event != NULL) 4449 EVENTHANDLER_DEREGISTER(vlan_config, ctx->ifc_vlan_attach_event); 4450 if (ctx->ifc_vlan_detach_event != NULL) 4451 EVENTHANDLER_DEREGISTER(vlan_unconfig, ctx->ifc_vlan_detach_event); 4452 4453 iflib_netmap_detach(ifp); 4454 ether_ifdetach(ifp); 4455 /* ether_ifdetach calls if_qflush - lock must be destroy afterwards*/ 4456 CTX_LOCK_DESTROY(ctx); 4457 if (ctx->ifc_led_dev != NULL) 4458 led_destroy(ctx->ifc_led_dev); 4459 /* XXX drain any dependent tasks */ 4460 tqg = qgroup_if_io_tqg; 4461 for (txq = ctx->ifc_txqs, i = 0; i < NTXQSETS(ctx); i++, txq++) { 4462 callout_drain(&txq->ift_timer); 4463 if (txq->ift_task.gt_uniq != NULL) 4464 taskqgroup_detach(tqg, &txq->ift_task); 4465 } 4466 for (i = 0, rxq = ctx->ifc_rxqs; i < NRXQSETS(ctx); i++, rxq++) { 4467 if (rxq->ifr_task.gt_uniq != NULL) 4468 taskqgroup_detach(tqg, &rxq->ifr_task); 4469 4470 for (j = 0, fl = rxq->ifr_fl; j < rxq->ifr_nfl; j++, fl++) 4471 free(fl->ifl_rx_bitmap, M_IFLIB); 4472 4473 } 4474 tqg = qgroup_if_config_tqg; 4475 if (ctx->ifc_admin_task.gt_uniq != NULL) 4476 taskqgroup_detach(tqg, &ctx->ifc_admin_task); 4477 if (ctx->ifc_vflr_task.gt_uniq != NULL) 4478 taskqgroup_detach(tqg, &ctx->ifc_vflr_task); 4479 4480 IFDI_DETACH(ctx); 4481 device_set_softc(ctx->ifc_dev, NULL); 4482 if (ctx->ifc_softc_ctx.isc_intr != IFLIB_INTR_LEGACY) { 4483 pci_release_msi(dev); 4484 } 4485 if (ctx->ifc_softc_ctx.isc_intr != IFLIB_INTR_MSIX) { 4486 iflib_irq_free(ctx, &ctx->ifc_legacy_irq); 4487 } 4488 if (ctx->ifc_msix_mem != NULL) { 4489 bus_release_resource(ctx->ifc_dev, SYS_RES_MEMORY, 4490 ctx->ifc_softc_ctx.isc_msix_bar, ctx->ifc_msix_mem); 4491 ctx->ifc_msix_mem = NULL; 4492 } 4493 4494 bus_generic_detach(dev); 4495 if_free(ifp); 4496 4497 iflib_tx_structures_free(ctx); 4498 iflib_rx_structures_free(ctx); 4499 if (ctx->ifc_flags & IFC_SC_ALLOCATED) 4500 free(ctx->ifc_softc, M_IFLIB); 4501 free(ctx, M_IFLIB); 4502 return (0); 4503 } 4504 4505 4506 int 4507 iflib_device_detach(device_t dev) 4508 { 4509 if_ctx_t ctx = device_get_softc(dev); 4510 4511 return (iflib_device_deregister(ctx)); 4512 } 4513 4514 int 4515 iflib_device_suspend(device_t dev) 4516 { 4517 if_ctx_t ctx = device_get_softc(dev); 4518 4519 CTX_LOCK(ctx); 4520 IFDI_SUSPEND(ctx); 4521 CTX_UNLOCK(ctx); 4522 4523 return bus_generic_suspend(dev); 4524 } 4525 int 4526 iflib_device_shutdown(device_t dev) 4527 { 4528 if_ctx_t ctx = device_get_softc(dev); 4529 4530 CTX_LOCK(ctx); 4531 IFDI_SHUTDOWN(ctx); 4532 CTX_UNLOCK(ctx); 4533 4534 return bus_generic_suspend(dev); 4535 } 4536 4537 4538 int 4539 iflib_device_resume(device_t dev) 4540 { 4541 if_ctx_t ctx = device_get_softc(dev); 4542 iflib_txq_t txq = ctx->ifc_txqs; 4543 4544 CTX_LOCK(ctx); 4545 IFDI_RESUME(ctx); 4546 iflib_init_locked(ctx); 4547 CTX_UNLOCK(ctx); 4548 for (int i = 0; i < NTXQSETS(ctx); i++, txq++) 4549 iflib_txq_check_drain(txq, IFLIB_RESTART_BUDGET); 4550 4551 return (bus_generic_resume(dev)); 4552 } 4553 4554 int 4555 iflib_device_iov_init(device_t dev, uint16_t num_vfs, const nvlist_t *params) 4556 { 4557 int error; 4558 if_ctx_t ctx = device_get_softc(dev); 4559 4560 CTX_LOCK(ctx); 4561 error = IFDI_IOV_INIT(ctx, num_vfs, params); 4562 CTX_UNLOCK(ctx); 4563 4564 return (error); 4565 } 4566 4567 void 4568 iflib_device_iov_uninit(device_t dev) 4569 { 4570 if_ctx_t ctx = device_get_softc(dev); 4571 4572 CTX_LOCK(ctx); 4573 IFDI_IOV_UNINIT(ctx); 4574 CTX_UNLOCK(ctx); 4575 } 4576 4577 int 4578 iflib_device_iov_add_vf(device_t dev, uint16_t vfnum, const nvlist_t *params) 4579 { 4580 int error; 4581 if_ctx_t ctx = device_get_softc(dev); 4582 4583 CTX_LOCK(ctx); 4584 error = IFDI_IOV_VF_ADD(ctx, vfnum, params); 4585 CTX_UNLOCK(ctx); 4586 4587 return (error); 4588 } 4589 4590 /********************************************************************* 4591 * 4592 * MODULE FUNCTION DEFINITIONS 4593 * 4594 **********************************************************************/ 4595 4596 /* 4597 * - Start a fast taskqueue thread for each core 4598 * - Start a taskqueue for control operations 4599 */ 4600 static int 4601 iflib_module_init(void) 4602 { 4603 return (0); 4604 } 4605 4606 static int 4607 iflib_module_event_handler(module_t mod, int what, void *arg) 4608 { 4609 int err; 4610 4611 switch (what) { 4612 case MOD_LOAD: 4613 if ((err = iflib_module_init()) != 0) 4614 return (err); 4615 break; 4616 case MOD_UNLOAD: 4617 return (EBUSY); 4618 default: 4619 return (EOPNOTSUPP); 4620 } 4621 4622 return (0); 4623 } 4624 4625 /********************************************************************* 4626 * 4627 * PUBLIC FUNCTION DEFINITIONS 4628 * ordered as in iflib.h 4629 * 4630 **********************************************************************/ 4631 4632 4633 static void 4634 _iflib_assert(if_shared_ctx_t sctx) 4635 { 4636 MPASS(sctx->isc_tx_maxsize); 4637 MPASS(sctx->isc_tx_maxsegsize); 4638 4639 MPASS(sctx->isc_rx_maxsize); 4640 MPASS(sctx->isc_rx_nsegments); 4641 MPASS(sctx->isc_rx_maxsegsize); 4642 4643 MPASS(sctx->isc_nrxd_min[0]); 4644 MPASS(sctx->isc_nrxd_max[0]); 4645 MPASS(sctx->isc_nrxd_default[0]); 4646 MPASS(sctx->isc_ntxd_min[0]); 4647 MPASS(sctx->isc_ntxd_max[0]); 4648 MPASS(sctx->isc_ntxd_default[0]); 4649 } 4650 4651 static void 4652 _iflib_pre_assert(if_softc_ctx_t scctx) 4653 { 4654 4655 MPASS(scctx->isc_txrx->ift_txd_encap); 4656 MPASS(scctx->isc_txrx->ift_txd_flush); 4657 MPASS(scctx->isc_txrx->ift_txd_credits_update); 4658 MPASS(scctx->isc_txrx->ift_rxd_available); 4659 MPASS(scctx->isc_txrx->ift_rxd_pkt_get); 4660 MPASS(scctx->isc_txrx->ift_rxd_refill); 4661 MPASS(scctx->isc_txrx->ift_rxd_flush); 4662 } 4663 4664 static int 4665 iflib_register(if_ctx_t ctx) 4666 { 4667 if_shared_ctx_t sctx = ctx->ifc_sctx; 4668 driver_t *driver = sctx->isc_driver; 4669 device_t dev = ctx->ifc_dev; 4670 if_t ifp; 4671 4672 _iflib_assert(sctx); 4673 4674 CTX_LOCK_INIT(ctx, device_get_nameunit(ctx->ifc_dev)); 4675 4676 ifp = ctx->ifc_ifp = if_gethandle(IFT_ETHER); 4677 if (ifp == NULL) { 4678 device_printf(dev, "can not allocate ifnet structure\n"); 4679 return (ENOMEM); 4680 } 4681 4682 /* 4683 * Initialize our context's device specific methods 4684 */ 4685 kobj_init((kobj_t) ctx, (kobj_class_t) driver); 4686 kobj_class_compile((kobj_class_t) driver); 4687 driver->refs++; 4688 4689 if_initname(ifp, device_get_name(dev), device_get_unit(dev)); 4690 if_setsoftc(ifp, ctx); 4691 if_setdev(ifp, dev); 4692 if_setinitfn(ifp, iflib_if_init); 4693 if_setioctlfn(ifp, iflib_if_ioctl); 4694 if_settransmitfn(ifp, iflib_if_transmit); 4695 if_setqflushfn(ifp, iflib_if_qflush); 4696 if_setflags(ifp, IFF_BROADCAST | IFF_SIMPLEX | IFF_MULTICAST); 4697 4698 ctx->ifc_vlan_attach_event = 4699 EVENTHANDLER_REGISTER(vlan_config, iflib_vlan_register, ctx, 4700 EVENTHANDLER_PRI_FIRST); 4701 ctx->ifc_vlan_detach_event = 4702 EVENTHANDLER_REGISTER(vlan_unconfig, iflib_vlan_unregister, ctx, 4703 EVENTHANDLER_PRI_FIRST); 4704 4705 ifmedia_init(&ctx->ifc_media, IFM_IMASK, 4706 iflib_media_change, iflib_media_status); 4707 4708 return (0); 4709 } 4710 4711 4712 static int 4713 iflib_queues_alloc(if_ctx_t ctx) 4714 { 4715 if_shared_ctx_t sctx = ctx->ifc_sctx; 4716 if_softc_ctx_t scctx = &ctx->ifc_softc_ctx; 4717 device_t dev = ctx->ifc_dev; 4718 int nrxqsets = scctx->isc_nrxqsets; 4719 int ntxqsets = scctx->isc_ntxqsets; 4720 iflib_txq_t txq; 4721 iflib_rxq_t rxq; 4722 iflib_fl_t fl = NULL; 4723 int i, j, cpu, err, txconf, rxconf; 4724 iflib_dma_info_t ifdip; 4725 uint32_t *rxqsizes = scctx->isc_rxqsizes; 4726 uint32_t *txqsizes = scctx->isc_txqsizes; 4727 uint8_t nrxqs = sctx->isc_nrxqs; 4728 uint8_t ntxqs = sctx->isc_ntxqs; 4729 int nfree_lists = sctx->isc_nfl ? sctx->isc_nfl : 1; 4730 caddr_t *vaddrs; 4731 uint64_t *paddrs; 4732 struct ifmp_ring **brscp; 4733 4734 KASSERT(ntxqs > 0, ("number of queues per qset must be at least 1")); 4735 KASSERT(nrxqs > 0, ("number of queues per qset must be at least 1")); 4736 4737 brscp = NULL; 4738 txq = NULL; 4739 rxq = NULL; 4740 4741 /* Allocate the TX ring struct memory */ 4742 if (!(txq = 4743 (iflib_txq_t) malloc(sizeof(struct iflib_txq) * 4744 ntxqsets, M_IFLIB, M_NOWAIT | M_ZERO))) { 4745 device_printf(dev, "Unable to allocate TX ring memory\n"); 4746 err = ENOMEM; 4747 goto fail; 4748 } 4749 4750 /* Now allocate the RX */ 4751 if (!(rxq = 4752 (iflib_rxq_t) malloc(sizeof(struct iflib_rxq) * 4753 nrxqsets, M_IFLIB, M_NOWAIT | M_ZERO))) { 4754 device_printf(dev, "Unable to allocate RX ring memory\n"); 4755 err = ENOMEM; 4756 goto rx_fail; 4757 } 4758 4759 ctx->ifc_txqs = txq; 4760 ctx->ifc_rxqs = rxq; 4761 4762 /* 4763 * XXX handle allocation failure 4764 */ 4765 for (txconf = i = 0, cpu = CPU_FIRST(); i < ntxqsets; i++, txconf++, txq++, cpu = CPU_NEXT(cpu)) { 4766 /* Set up some basics */ 4767 4768 if ((ifdip = malloc(sizeof(struct iflib_dma_info) * ntxqs, M_IFLIB, M_WAITOK|M_ZERO)) == NULL) { 4769 device_printf(dev, "failed to allocate iflib_dma_info\n"); 4770 err = ENOMEM; 4771 goto err_tx_desc; 4772 } 4773 txq->ift_ifdi = ifdip; 4774 for (j = 0; j < ntxqs; j++, ifdip++) { 4775 if (iflib_dma_alloc(ctx, txqsizes[j], ifdip, BUS_DMA_NOWAIT)) { 4776 device_printf(dev, "Unable to allocate Descriptor memory\n"); 4777 err = ENOMEM; 4778 goto err_tx_desc; 4779 } 4780 txq->ift_txd_size[j] = scctx->isc_txd_size[j]; 4781 bzero((void *)ifdip->idi_vaddr, txqsizes[j]); 4782 } 4783 txq->ift_ctx = ctx; 4784 txq->ift_id = i; 4785 if (sctx->isc_flags & IFLIB_HAS_TXCQ) { 4786 txq->ift_br_offset = 1; 4787 } else { 4788 txq->ift_br_offset = 0; 4789 } 4790 /* XXX fix this */ 4791 txq->ift_timer.c_cpu = cpu; 4792 4793 if (iflib_txsd_alloc(txq)) { 4794 device_printf(dev, "Critical Failure setting up TX buffers\n"); 4795 err = ENOMEM; 4796 goto err_tx_desc; 4797 } 4798 4799 /* Initialize the TX lock */ 4800 snprintf(txq->ift_mtx_name, MTX_NAME_LEN, "%s:tx(%d):callout", 4801 device_get_nameunit(dev), txq->ift_id); 4802 mtx_init(&txq->ift_mtx, txq->ift_mtx_name, NULL, MTX_DEF); 4803 callout_init_mtx(&txq->ift_timer, &txq->ift_mtx, 0); 4804 4805 snprintf(txq->ift_db_mtx_name, MTX_NAME_LEN, "%s:tx(%d):db", 4806 device_get_nameunit(dev), txq->ift_id); 4807 4808 err = ifmp_ring_alloc(&txq->ift_br, 2048, txq, iflib_txq_drain, 4809 iflib_txq_can_drain, M_IFLIB, M_WAITOK); 4810 if (err) { 4811 /* XXX free any allocated rings */ 4812 device_printf(dev, "Unable to allocate buf_ring\n"); 4813 goto err_tx_desc; 4814 } 4815 } 4816 4817 for (rxconf = i = 0; i < nrxqsets; i++, rxconf++, rxq++) { 4818 /* Set up some basics */ 4819 4820 if ((ifdip = malloc(sizeof(struct iflib_dma_info) * nrxqs, M_IFLIB, M_WAITOK|M_ZERO)) == NULL) { 4821 device_printf(dev, "failed to allocate iflib_dma_info\n"); 4822 err = ENOMEM; 4823 goto err_tx_desc; 4824 } 4825 4826 rxq->ifr_ifdi = ifdip; 4827 /* XXX this needs to be changed if #rx queues != #tx queues */ 4828 rxq->ifr_ntxqirq = 1; 4829 rxq->ifr_txqid[0] = i; 4830 for (j = 0; j < nrxqs; j++, ifdip++) { 4831 if (iflib_dma_alloc(ctx, rxqsizes[j], ifdip, BUS_DMA_NOWAIT)) { 4832 device_printf(dev, "Unable to allocate Descriptor memory\n"); 4833 err = ENOMEM; 4834 goto err_tx_desc; 4835 } 4836 bzero((void *)ifdip->idi_vaddr, rxqsizes[j]); 4837 } 4838 rxq->ifr_ctx = ctx; 4839 rxq->ifr_id = i; 4840 if (sctx->isc_flags & IFLIB_HAS_RXCQ) { 4841 rxq->ifr_fl_offset = 1; 4842 } else { 4843 rxq->ifr_fl_offset = 0; 4844 } 4845 rxq->ifr_nfl = nfree_lists; 4846 if (!(fl = 4847 (iflib_fl_t) malloc(sizeof(struct iflib_fl) * nfree_lists, M_IFLIB, M_NOWAIT | M_ZERO))) { 4848 device_printf(dev, "Unable to allocate free list memory\n"); 4849 err = ENOMEM; 4850 goto err_tx_desc; 4851 } 4852 rxq->ifr_fl = fl; 4853 for (j = 0; j < nfree_lists; j++) { 4854 fl[j].ifl_rxq = rxq; 4855 fl[j].ifl_id = j; 4856 fl[j].ifl_ifdi = &rxq->ifr_ifdi[j + rxq->ifr_fl_offset]; 4857 fl[j].ifl_rxd_size = scctx->isc_rxd_size[j]; 4858 } 4859 /* Allocate receive buffers for the ring*/ 4860 if (iflib_rxsd_alloc(rxq)) { 4861 device_printf(dev, 4862 "Critical Failure setting up receive buffers\n"); 4863 err = ENOMEM; 4864 goto err_rx_desc; 4865 } 4866 4867 for (j = 0, fl = rxq->ifr_fl; j < rxq->ifr_nfl; j++, fl++) 4868 fl->ifl_rx_bitmap = bit_alloc(fl->ifl_size, M_IFLIB, M_WAITOK|M_ZERO); 4869 } 4870 4871 /* TXQs */ 4872 vaddrs = malloc(sizeof(caddr_t)*ntxqsets*ntxqs, M_IFLIB, M_WAITOK); 4873 paddrs = malloc(sizeof(uint64_t)*ntxqsets*ntxqs, M_IFLIB, M_WAITOK); 4874 for (i = 0; i < ntxqsets; i++) { 4875 iflib_dma_info_t di = ctx->ifc_txqs[i].ift_ifdi; 4876 4877 for (j = 0; j < ntxqs; j++, di++) { 4878 vaddrs[i*ntxqs + j] = di->idi_vaddr; 4879 paddrs[i*ntxqs + j] = di->idi_paddr; 4880 } 4881 } 4882 if ((err = IFDI_TX_QUEUES_ALLOC(ctx, vaddrs, paddrs, ntxqs, ntxqsets)) != 0) { 4883 device_printf(ctx->ifc_dev, "device queue allocation failed\n"); 4884 iflib_tx_structures_free(ctx); 4885 free(vaddrs, M_IFLIB); 4886 free(paddrs, M_IFLIB); 4887 goto err_rx_desc; 4888 } 4889 free(vaddrs, M_IFLIB); 4890 free(paddrs, M_IFLIB); 4891 4892 /* RXQs */ 4893 vaddrs = malloc(sizeof(caddr_t)*nrxqsets*nrxqs, M_IFLIB, M_WAITOK); 4894 paddrs = malloc(sizeof(uint64_t)*nrxqsets*nrxqs, M_IFLIB, M_WAITOK); 4895 for (i = 0; i < nrxqsets; i++) { 4896 iflib_dma_info_t di = ctx->ifc_rxqs[i].ifr_ifdi; 4897 4898 for (j = 0; j < nrxqs; j++, di++) { 4899 vaddrs[i*nrxqs + j] = di->idi_vaddr; 4900 paddrs[i*nrxqs + j] = di->idi_paddr; 4901 } 4902 } 4903 if ((err = IFDI_RX_QUEUES_ALLOC(ctx, vaddrs, paddrs, nrxqs, nrxqsets)) != 0) { 4904 device_printf(ctx->ifc_dev, "device queue allocation failed\n"); 4905 iflib_tx_structures_free(ctx); 4906 free(vaddrs, M_IFLIB); 4907 free(paddrs, M_IFLIB); 4908 goto err_rx_desc; 4909 } 4910 free(vaddrs, M_IFLIB); 4911 free(paddrs, M_IFLIB); 4912 4913 return (0); 4914 4915 /* XXX handle allocation failure changes */ 4916 err_rx_desc: 4917 err_tx_desc: 4918 if (ctx->ifc_rxqs != NULL) 4919 free(ctx->ifc_rxqs, M_IFLIB); 4920 ctx->ifc_rxqs = NULL; 4921 if (ctx->ifc_txqs != NULL) 4922 free(ctx->ifc_txqs, M_IFLIB); 4923 ctx->ifc_txqs = NULL; 4924 rx_fail: 4925 if (brscp != NULL) 4926 free(brscp, M_IFLIB); 4927 if (rxq != NULL) 4928 free(rxq, M_IFLIB); 4929 if (txq != NULL) 4930 free(txq, M_IFLIB); 4931 fail: 4932 return (err); 4933 } 4934 4935 static int 4936 iflib_tx_structures_setup(if_ctx_t ctx) 4937 { 4938 iflib_txq_t txq = ctx->ifc_txqs; 4939 int i; 4940 4941 for (i = 0; i < NTXQSETS(ctx); i++, txq++) 4942 iflib_txq_setup(txq); 4943 4944 return (0); 4945 } 4946 4947 static void 4948 iflib_tx_structures_free(if_ctx_t ctx) 4949 { 4950 iflib_txq_t txq = ctx->ifc_txqs; 4951 int i, j; 4952 4953 for (i = 0; i < NTXQSETS(ctx); i++, txq++) { 4954 iflib_txq_destroy(txq); 4955 for (j = 0; j < ctx->ifc_nhwtxqs; j++) 4956 iflib_dma_free(&txq->ift_ifdi[j]); 4957 } 4958 free(ctx->ifc_txqs, M_IFLIB); 4959 ctx->ifc_txqs = NULL; 4960 IFDI_QUEUES_FREE(ctx); 4961 } 4962 4963 /********************************************************************* 4964 * 4965 * Initialize all receive rings. 4966 * 4967 **********************************************************************/ 4968 static int 4969 iflib_rx_structures_setup(if_ctx_t ctx) 4970 { 4971 iflib_rxq_t rxq = ctx->ifc_rxqs; 4972 int q; 4973 #if defined(INET6) || defined(INET) 4974 int i, err; 4975 #endif 4976 4977 for (q = 0; q < ctx->ifc_softc_ctx.isc_nrxqsets; q++, rxq++) { 4978 #if defined(INET6) || defined(INET) 4979 tcp_lro_free(&rxq->ifr_lc); 4980 if ((err = tcp_lro_init_args(&rxq->ifr_lc, ctx->ifc_ifp, 4981 TCP_LRO_ENTRIES, min(1024, 4982 ctx->ifc_softc_ctx.isc_nrxd[rxq->ifr_fl_offset]))) != 0) { 4983 device_printf(ctx->ifc_dev, "LRO Initialization failed!\n"); 4984 goto fail; 4985 } 4986 rxq->ifr_lro_enabled = TRUE; 4987 #endif 4988 IFDI_RXQ_SETUP(ctx, rxq->ifr_id); 4989 } 4990 return (0); 4991 #if defined(INET6) || defined(INET) 4992 fail: 4993 /* 4994 * Free RX software descriptors allocated so far, we will only handle 4995 * the rings that completed, the failing case will have 4996 * cleaned up for itself. 'q' failed, so its the terminus. 4997 */ 4998 rxq = ctx->ifc_rxqs; 4999 for (i = 0; i < q; ++i, rxq++) { 5000 iflib_rx_sds_free(rxq); 5001 rxq->ifr_cq_gen = rxq->ifr_cq_cidx = rxq->ifr_cq_pidx = 0; 5002 } 5003 return (err); 5004 #endif 5005 } 5006 5007 /********************************************************************* 5008 * 5009 * Free all receive rings. 5010 * 5011 **********************************************************************/ 5012 static void 5013 iflib_rx_structures_free(if_ctx_t ctx) 5014 { 5015 iflib_rxq_t rxq = ctx->ifc_rxqs; 5016 5017 for (int i = 0; i < ctx->ifc_softc_ctx.isc_nrxqsets; i++, rxq++) { 5018 iflib_rx_sds_free(rxq); 5019 } 5020 } 5021 5022 static int 5023 iflib_qset_structures_setup(if_ctx_t ctx) 5024 { 5025 int err; 5026 5027 if ((err = iflib_tx_structures_setup(ctx)) != 0) 5028 return (err); 5029 5030 if ((err = iflib_rx_structures_setup(ctx)) != 0) { 5031 device_printf(ctx->ifc_dev, "iflib_rx_structures_setup failed: %d\n", err); 5032 iflib_tx_structures_free(ctx); 5033 iflib_rx_structures_free(ctx); 5034 } 5035 return (err); 5036 } 5037 5038 int 5039 iflib_irq_alloc(if_ctx_t ctx, if_irq_t irq, int rid, 5040 driver_filter_t filter, void *filter_arg, driver_intr_t handler, void *arg, char *name) 5041 { 5042 5043 return (_iflib_irq_alloc(ctx, irq, rid, filter, handler, arg, name)); 5044 } 5045 5046 static int 5047 find_nth(if_ctx_t ctx, cpuset_t *cpus, int qid) 5048 { 5049 int i, cpuid, eqid, count; 5050 5051 CPU_COPY(&ctx->ifc_cpus, cpus); 5052 count = CPU_COUNT(cpus); 5053 eqid = qid % count; 5054 /* clear up to the qid'th bit */ 5055 for (i = 0; i < eqid; i++) { 5056 cpuid = CPU_FFS(cpus); 5057 MPASS(cpuid != 0); 5058 CPU_CLR(cpuid-1, cpus); 5059 } 5060 cpuid = CPU_FFS(cpus); 5061 MPASS(cpuid != 0); 5062 return (cpuid-1); 5063 } 5064 5065 int 5066 iflib_irq_alloc_generic(if_ctx_t ctx, if_irq_t irq, int rid, 5067 iflib_intr_type_t type, driver_filter_t *filter, 5068 void *filter_arg, int qid, char *name) 5069 { 5070 struct grouptask *gtask; 5071 struct taskqgroup *tqg; 5072 iflib_filter_info_t info; 5073 cpuset_t cpus; 5074 gtask_fn_t *fn; 5075 int tqrid, err, cpuid; 5076 driver_filter_t *intr_fast; 5077 void *q; 5078 5079 info = &ctx->ifc_filter_info; 5080 tqrid = rid; 5081 5082 switch (type) { 5083 /* XXX merge tx/rx for netmap? */ 5084 case IFLIB_INTR_TX: 5085 q = &ctx->ifc_txqs[qid]; 5086 info = &ctx->ifc_txqs[qid].ift_filter_info; 5087 gtask = &ctx->ifc_txqs[qid].ift_task; 5088 tqg = qgroup_if_io_tqg; 5089 fn = _task_fn_tx; 5090 intr_fast = iflib_fast_intr; 5091 GROUPTASK_INIT(gtask, 0, fn, q); 5092 break; 5093 case IFLIB_INTR_RX: 5094 q = &ctx->ifc_rxqs[qid]; 5095 info = &ctx->ifc_rxqs[qid].ifr_filter_info; 5096 gtask = &ctx->ifc_rxqs[qid].ifr_task; 5097 tqg = qgroup_if_io_tqg; 5098 fn = _task_fn_rx; 5099 intr_fast = iflib_fast_intr; 5100 GROUPTASK_INIT(gtask, 0, fn, q); 5101 break; 5102 case IFLIB_INTR_RXTX: 5103 q = &ctx->ifc_rxqs[qid]; 5104 info = &ctx->ifc_rxqs[qid].ifr_filter_info; 5105 gtask = &ctx->ifc_rxqs[qid].ifr_task; 5106 tqg = qgroup_if_io_tqg; 5107 fn = _task_fn_rx; 5108 intr_fast = iflib_fast_intr_rxtx; 5109 GROUPTASK_INIT(gtask, 0, fn, q); 5110 break; 5111 case IFLIB_INTR_ADMIN: 5112 q = ctx; 5113 tqrid = -1; 5114 info = &ctx->ifc_filter_info; 5115 gtask = &ctx->ifc_admin_task; 5116 tqg = qgroup_if_config_tqg; 5117 fn = _task_fn_admin; 5118 intr_fast = iflib_fast_intr_ctx; 5119 break; 5120 default: 5121 panic("unknown net intr type"); 5122 } 5123 5124 info->ifi_filter = filter; 5125 info->ifi_filter_arg = filter_arg; 5126 info->ifi_task = gtask; 5127 info->ifi_ctx = q; 5128 5129 err = _iflib_irq_alloc(ctx, irq, rid, intr_fast, NULL, info, name); 5130 if (err != 0) { 5131 device_printf(ctx->ifc_dev, "_iflib_irq_alloc failed %d\n", err); 5132 return (err); 5133 } 5134 if (type == IFLIB_INTR_ADMIN) 5135 return (0); 5136 5137 if (tqrid != -1) { 5138 cpuid = find_nth(ctx, &cpus, qid); 5139 taskqgroup_attach_cpu(tqg, gtask, q, cpuid, rman_get_start(irq->ii_res), name); 5140 } else { 5141 taskqgroup_attach(tqg, gtask, q, rman_get_start(irq->ii_res), name); 5142 } 5143 5144 return (0); 5145 } 5146 5147 void 5148 iflib_softirq_alloc_generic(if_ctx_t ctx, if_irq_t irq, iflib_intr_type_t type, void *arg, int qid, char *name) 5149 { 5150 struct grouptask *gtask; 5151 struct taskqgroup *tqg; 5152 gtask_fn_t *fn; 5153 void *q; 5154 int irq_num = -1; 5155 5156 switch (type) { 5157 case IFLIB_INTR_TX: 5158 q = &ctx->ifc_txqs[qid]; 5159 gtask = &ctx->ifc_txqs[qid].ift_task; 5160 tqg = qgroup_if_io_tqg; 5161 fn = _task_fn_tx; 5162 if (irq != NULL) 5163 irq_num = rman_get_start(irq->ii_res); 5164 break; 5165 case IFLIB_INTR_RX: 5166 q = &ctx->ifc_rxqs[qid]; 5167 gtask = &ctx->ifc_rxqs[qid].ifr_task; 5168 tqg = qgroup_if_io_tqg; 5169 fn = _task_fn_rx; 5170 if (irq != NULL) 5171 irq_num = rman_get_start(irq->ii_res); 5172 break; 5173 case IFLIB_INTR_IOV: 5174 q = ctx; 5175 gtask = &ctx->ifc_vflr_task; 5176 tqg = qgroup_if_config_tqg; 5177 fn = _task_fn_iov; 5178 break; 5179 default: 5180 panic("unknown net intr type"); 5181 } 5182 GROUPTASK_INIT(gtask, 0, fn, q); 5183 taskqgroup_attach(tqg, gtask, q, irq_num, name); 5184 } 5185 5186 void 5187 iflib_irq_free(if_ctx_t ctx, if_irq_t irq) 5188 { 5189 if (irq->ii_tag) 5190 bus_teardown_intr(ctx->ifc_dev, irq->ii_res, irq->ii_tag); 5191 5192 if (irq->ii_res) 5193 bus_release_resource(ctx->ifc_dev, SYS_RES_IRQ, irq->ii_rid, irq->ii_res); 5194 } 5195 5196 static int 5197 iflib_legacy_setup(if_ctx_t ctx, driver_filter_t filter, void *filter_arg, int *rid, char *name) 5198 { 5199 iflib_txq_t txq = ctx->ifc_txqs; 5200 iflib_rxq_t rxq = ctx->ifc_rxqs; 5201 if_irq_t irq = &ctx->ifc_legacy_irq; 5202 iflib_filter_info_t info; 5203 struct grouptask *gtask; 5204 struct taskqgroup *tqg; 5205 gtask_fn_t *fn; 5206 int tqrid; 5207 void *q; 5208 int err; 5209 5210 q = &ctx->ifc_rxqs[0]; 5211 info = &rxq[0].ifr_filter_info; 5212 gtask = &rxq[0].ifr_task; 5213 tqg = qgroup_if_io_tqg; 5214 tqrid = irq->ii_rid = *rid; 5215 fn = _task_fn_rx; 5216 5217 ctx->ifc_flags |= IFC_LEGACY; 5218 info->ifi_filter = filter; 5219 info->ifi_filter_arg = filter_arg; 5220 info->ifi_task = gtask; 5221 info->ifi_ctx = ctx; 5222 5223 /* We allocate a single interrupt resource */ 5224 if ((err = _iflib_irq_alloc(ctx, irq, tqrid, iflib_fast_intr_ctx, NULL, info, name)) != 0) 5225 return (err); 5226 GROUPTASK_INIT(gtask, 0, fn, q); 5227 taskqgroup_attach(tqg, gtask, q, tqrid, name); 5228 5229 GROUPTASK_INIT(&txq->ift_task, 0, _task_fn_tx, txq); 5230 taskqgroup_attach(qgroup_if_io_tqg, &txq->ift_task, txq, tqrid, "tx"); 5231 return (0); 5232 } 5233 5234 void 5235 iflib_led_create(if_ctx_t ctx) 5236 { 5237 5238 ctx->ifc_led_dev = led_create(iflib_led_func, ctx, 5239 device_get_nameunit(ctx->ifc_dev)); 5240 } 5241 5242 void 5243 iflib_tx_intr_deferred(if_ctx_t ctx, int txqid) 5244 { 5245 5246 GROUPTASK_ENQUEUE(&ctx->ifc_txqs[txqid].ift_task); 5247 } 5248 5249 void 5250 iflib_rx_intr_deferred(if_ctx_t ctx, int rxqid) 5251 { 5252 5253 GROUPTASK_ENQUEUE(&ctx->ifc_rxqs[rxqid].ifr_task); 5254 } 5255 5256 void 5257 iflib_admin_intr_deferred(if_ctx_t ctx) 5258 { 5259 #ifdef INVARIANTS 5260 struct grouptask *gtask; 5261 5262 gtask = &ctx->ifc_admin_task; 5263 MPASS(gtask != NULL && gtask->gt_taskqueue != NULL); 5264 #endif 5265 5266 GROUPTASK_ENQUEUE(&ctx->ifc_admin_task); 5267 } 5268 5269 void 5270 iflib_iov_intr_deferred(if_ctx_t ctx) 5271 { 5272 5273 GROUPTASK_ENQUEUE(&ctx->ifc_vflr_task); 5274 } 5275 5276 void 5277 iflib_io_tqg_attach(struct grouptask *gt, void *uniq, int cpu, char *name) 5278 { 5279 5280 taskqgroup_attach_cpu(qgroup_if_io_tqg, gt, uniq, cpu, -1, name); 5281 } 5282 5283 void 5284 iflib_config_gtask_init(if_ctx_t ctx, struct grouptask *gtask, gtask_fn_t *fn, 5285 char *name) 5286 { 5287 5288 GROUPTASK_INIT(gtask, 0, fn, ctx); 5289 taskqgroup_attach(qgroup_if_config_tqg, gtask, gtask, -1, name); 5290 } 5291 5292 void 5293 iflib_config_gtask_deinit(struct grouptask *gtask) 5294 { 5295 5296 taskqgroup_detach(qgroup_if_config_tqg, gtask); 5297 } 5298 5299 void 5300 iflib_link_state_change(if_ctx_t ctx, int link_state, uint64_t baudrate) 5301 { 5302 if_t ifp = ctx->ifc_ifp; 5303 iflib_txq_t txq = ctx->ifc_txqs; 5304 5305 if_setbaudrate(ifp, baudrate); 5306 if (baudrate >= IF_Gbps(10)) 5307 ctx->ifc_flags |= IFC_PREFETCH; 5308 5309 /* If link down, disable watchdog */ 5310 if ((ctx->ifc_link_state == LINK_STATE_UP) && (link_state == LINK_STATE_DOWN)) { 5311 for (int i = 0; i < ctx->ifc_softc_ctx.isc_ntxqsets; i++, txq++) 5312 txq->ift_qstatus = IFLIB_QUEUE_IDLE; 5313 } 5314 ctx->ifc_link_state = link_state; 5315 if_link_state_change(ifp, link_state); 5316 } 5317 5318 static int 5319 iflib_tx_credits_update(if_ctx_t ctx, iflib_txq_t txq) 5320 { 5321 int credits; 5322 #ifdef INVARIANTS 5323 int credits_pre = txq->ift_cidx_processed; 5324 #endif 5325 5326 if (ctx->isc_txd_credits_update == NULL) 5327 return (0); 5328 5329 if ((credits = ctx->isc_txd_credits_update(ctx->ifc_softc, txq->ift_id, true)) == 0) 5330 return (0); 5331 5332 txq->ift_processed += credits; 5333 txq->ift_cidx_processed += credits; 5334 5335 MPASS(credits_pre + credits == txq->ift_cidx_processed); 5336 if (txq->ift_cidx_processed >= txq->ift_size) 5337 txq->ift_cidx_processed -= txq->ift_size; 5338 return (credits); 5339 } 5340 5341 static int 5342 iflib_rxd_avail(if_ctx_t ctx, iflib_rxq_t rxq, qidx_t cidx, qidx_t budget) 5343 { 5344 5345 return (ctx->isc_rxd_available(ctx->ifc_softc, rxq->ifr_id, cidx, 5346 budget)); 5347 } 5348 5349 void 5350 iflib_add_int_delay_sysctl(if_ctx_t ctx, const char *name, 5351 const char *description, if_int_delay_info_t info, 5352 int offset, int value) 5353 { 5354 info->iidi_ctx = ctx; 5355 info->iidi_offset = offset; 5356 info->iidi_value = value; 5357 SYSCTL_ADD_PROC(device_get_sysctl_ctx(ctx->ifc_dev), 5358 SYSCTL_CHILDREN(device_get_sysctl_tree(ctx->ifc_dev)), 5359 OID_AUTO, name, CTLTYPE_INT|CTLFLAG_RW, 5360 info, 0, iflib_sysctl_int_delay, "I", description); 5361 } 5362 5363 struct mtx * 5364 iflib_ctx_lock_get(if_ctx_t ctx) 5365 { 5366 5367 return (&ctx->ifc_mtx); 5368 } 5369 5370 static int 5371 iflib_msix_init(if_ctx_t ctx) 5372 { 5373 device_t dev = ctx->ifc_dev; 5374 if_shared_ctx_t sctx = ctx->ifc_sctx; 5375 if_softc_ctx_t scctx = &ctx->ifc_softc_ctx; 5376 int vectors, queues, rx_queues, tx_queues, queuemsgs, msgs; 5377 int iflib_num_tx_queues, iflib_num_rx_queues; 5378 int err, admincnt, bar; 5379 5380 iflib_num_tx_queues = ctx->ifc_sysctl_ntxqs; 5381 iflib_num_rx_queues = ctx->ifc_sysctl_nrxqs; 5382 5383 device_printf(dev, "msix_init qsets capped at %d\n", imax(scctx->isc_ntxqsets, scctx->isc_nrxqsets)); 5384 5385 bar = ctx->ifc_softc_ctx.isc_msix_bar; 5386 admincnt = sctx->isc_admin_intrcnt; 5387 /* Override by global tuneable */ 5388 { 5389 int i; 5390 size_t len = sizeof(i); 5391 err = kernel_sysctlbyname(curthread, "hw.pci.enable_msix", &i, &len, NULL, 0, NULL, 0); 5392 if (err == 0) { 5393 if (i == 0) 5394 goto msi; 5395 } 5396 else { 5397 device_printf(dev, "unable to read hw.pci.enable_msix."); 5398 } 5399 } 5400 /* Override by tuneable */ 5401 if (scctx->isc_disable_msix) 5402 goto msi; 5403 5404 /* 5405 ** When used in a virtualized environment 5406 ** PCI BUSMASTER capability may not be set 5407 ** so explicity set it here and rewrite 5408 ** the ENABLE in the MSIX control register 5409 ** at this point to cause the host to 5410 ** successfully initialize us. 5411 */ 5412 { 5413 int msix_ctrl, rid; 5414 5415 pci_enable_busmaster(dev); 5416 rid = 0; 5417 if (pci_find_cap(dev, PCIY_MSIX, &rid) == 0 && rid != 0) { 5418 rid += PCIR_MSIX_CTRL; 5419 msix_ctrl = pci_read_config(dev, rid, 2); 5420 msix_ctrl |= PCIM_MSIXCTRL_MSIX_ENABLE; 5421 pci_write_config(dev, rid, msix_ctrl, 2); 5422 } else { 5423 device_printf(dev, "PCIY_MSIX capability not found; " 5424 "or rid %d == 0.\n", rid); 5425 goto msi; 5426 } 5427 } 5428 5429 /* 5430 * bar == -1 => "trust me I know what I'm doing" 5431 * Some drivers are for hardware that is so shoddily 5432 * documented that no one knows which bars are which 5433 * so the developer has to map all bars. This hack 5434 * allows shoddy garbage to use msix in this framework. 5435 */ 5436 if (bar != -1) { 5437 ctx->ifc_msix_mem = bus_alloc_resource_any(dev, 5438 SYS_RES_MEMORY, &bar, RF_ACTIVE); 5439 if (ctx->ifc_msix_mem == NULL) { 5440 /* May not be enabled */ 5441 device_printf(dev, "Unable to map MSIX table \n"); 5442 goto msi; 5443 } 5444 } 5445 /* First try MSI/X */ 5446 if ((msgs = pci_msix_count(dev)) == 0) { /* system has msix disabled */ 5447 device_printf(dev, "System has MSIX disabled \n"); 5448 bus_release_resource(dev, SYS_RES_MEMORY, 5449 bar, ctx->ifc_msix_mem); 5450 ctx->ifc_msix_mem = NULL; 5451 goto msi; 5452 } 5453 #if IFLIB_DEBUG 5454 /* use only 1 qset in debug mode */ 5455 queuemsgs = min(msgs - admincnt, 1); 5456 #else 5457 queuemsgs = msgs - admincnt; 5458 #endif 5459 #ifdef RSS 5460 queues = imin(queuemsgs, rss_getnumbuckets()); 5461 #else 5462 queues = queuemsgs; 5463 #endif 5464 queues = imin(CPU_COUNT(&ctx->ifc_cpus), queues); 5465 device_printf(dev, "pxm cpus: %d queue msgs: %d admincnt: %d\n", 5466 CPU_COUNT(&ctx->ifc_cpus), queuemsgs, admincnt); 5467 #ifdef RSS 5468 /* If we're doing RSS, clamp at the number of RSS buckets */ 5469 if (queues > rss_getnumbuckets()) 5470 queues = rss_getnumbuckets(); 5471 #endif 5472 if (iflib_num_rx_queues > 0 && iflib_num_rx_queues < queuemsgs - admincnt) 5473 rx_queues = iflib_num_rx_queues; 5474 else 5475 rx_queues = queues; 5476 5477 if (rx_queues > scctx->isc_nrxqsets) 5478 rx_queues = scctx->isc_nrxqsets; 5479 5480 /* 5481 * We want this to be all logical CPUs by default 5482 */ 5483 if (iflib_num_tx_queues > 0 && iflib_num_tx_queues < queues) 5484 tx_queues = iflib_num_tx_queues; 5485 else 5486 tx_queues = mp_ncpus; 5487 5488 if (tx_queues > scctx->isc_ntxqsets) 5489 tx_queues = scctx->isc_ntxqsets; 5490 5491 if (ctx->ifc_sysctl_qs_eq_override == 0) { 5492 #ifdef INVARIANTS 5493 if (tx_queues != rx_queues) 5494 device_printf(dev, "queue equality override not set, capping rx_queues at %d and tx_queues at %d\n", 5495 min(rx_queues, tx_queues), min(rx_queues, tx_queues)); 5496 #endif 5497 tx_queues = min(rx_queues, tx_queues); 5498 rx_queues = min(rx_queues, tx_queues); 5499 } 5500 5501 device_printf(dev, "using %d rx queues %d tx queues \n", rx_queues, tx_queues); 5502 5503 vectors = rx_queues + admincnt; 5504 if ((err = pci_alloc_msix(dev, &vectors)) == 0) { 5505 device_printf(dev, 5506 "Using MSIX interrupts with %d vectors\n", vectors); 5507 scctx->isc_vectors = vectors; 5508 scctx->isc_nrxqsets = rx_queues; 5509 scctx->isc_ntxqsets = tx_queues; 5510 scctx->isc_intr = IFLIB_INTR_MSIX; 5511 5512 return (vectors); 5513 } else { 5514 device_printf(dev, "failed to allocate %d msix vectors, err: %d - using MSI\n", vectors, err); 5515 } 5516 msi: 5517 vectors = pci_msi_count(dev); 5518 scctx->isc_nrxqsets = 1; 5519 scctx->isc_ntxqsets = 1; 5520 scctx->isc_vectors = vectors; 5521 if (vectors == 1 && pci_alloc_msi(dev, &vectors) == 0) { 5522 device_printf(dev,"Using an MSI interrupt\n"); 5523 scctx->isc_intr = IFLIB_INTR_MSI; 5524 } else { 5525 device_printf(dev,"Using a Legacy interrupt\n"); 5526 scctx->isc_intr = IFLIB_INTR_LEGACY; 5527 } 5528 5529 return (vectors); 5530 } 5531 5532 char * ring_states[] = { "IDLE", "BUSY", "STALLED", "ABDICATED" }; 5533 5534 static int 5535 mp_ring_state_handler(SYSCTL_HANDLER_ARGS) 5536 { 5537 int rc; 5538 uint16_t *state = ((uint16_t *)oidp->oid_arg1); 5539 struct sbuf *sb; 5540 char *ring_state = "UNKNOWN"; 5541 5542 /* XXX needed ? */ 5543 rc = sysctl_wire_old_buffer(req, 0); 5544 MPASS(rc == 0); 5545 if (rc != 0) 5546 return (rc); 5547 sb = sbuf_new_for_sysctl(NULL, NULL, 80, req); 5548 MPASS(sb != NULL); 5549 if (sb == NULL) 5550 return (ENOMEM); 5551 if (state[3] <= 3) 5552 ring_state = ring_states[state[3]]; 5553 5554 sbuf_printf(sb, "pidx_head: %04hd pidx_tail: %04hd cidx: %04hd state: %s", 5555 state[0], state[1], state[2], ring_state); 5556 rc = sbuf_finish(sb); 5557 sbuf_delete(sb); 5558 return(rc); 5559 } 5560 5561 enum iflib_ndesc_handler { 5562 IFLIB_NTXD_HANDLER, 5563 IFLIB_NRXD_HANDLER, 5564 }; 5565 5566 static int 5567 mp_ndesc_handler(SYSCTL_HANDLER_ARGS) 5568 { 5569 if_ctx_t ctx = (void *)arg1; 5570 enum iflib_ndesc_handler type = arg2; 5571 char buf[256] = {0}; 5572 qidx_t *ndesc; 5573 char *p, *next; 5574 int nqs, rc, i; 5575 5576 MPASS(type == IFLIB_NTXD_HANDLER || type == IFLIB_NRXD_HANDLER); 5577 5578 nqs = 8; 5579 switch(type) { 5580 case IFLIB_NTXD_HANDLER: 5581 ndesc = ctx->ifc_sysctl_ntxds; 5582 if (ctx->ifc_sctx) 5583 nqs = ctx->ifc_sctx->isc_ntxqs; 5584 break; 5585 case IFLIB_NRXD_HANDLER: 5586 ndesc = ctx->ifc_sysctl_nrxds; 5587 if (ctx->ifc_sctx) 5588 nqs = ctx->ifc_sctx->isc_nrxqs; 5589 break; 5590 } 5591 if (nqs == 0) 5592 nqs = 8; 5593 5594 for (i=0; i<8; i++) { 5595 if (i >= nqs) 5596 break; 5597 if (i) 5598 strcat(buf, ","); 5599 sprintf(strchr(buf, 0), "%d", ndesc[i]); 5600 } 5601 5602 rc = sysctl_handle_string(oidp, buf, sizeof(buf), req); 5603 if (rc || req->newptr == NULL) 5604 return rc; 5605 5606 for (i = 0, next = buf, p = strsep(&next, " ,"); i < 8 && p; 5607 i++, p = strsep(&next, " ,")) { 5608 ndesc[i] = strtoul(p, NULL, 10); 5609 } 5610 5611 return(rc); 5612 } 5613 5614 #define NAME_BUFLEN 32 5615 static void 5616 iflib_add_device_sysctl_pre(if_ctx_t ctx) 5617 { 5618 device_t dev = iflib_get_dev(ctx); 5619 struct sysctl_oid_list *child, *oid_list; 5620 struct sysctl_ctx_list *ctx_list; 5621 struct sysctl_oid *node; 5622 5623 ctx_list = device_get_sysctl_ctx(dev); 5624 child = SYSCTL_CHILDREN(device_get_sysctl_tree(dev)); 5625 ctx->ifc_sysctl_node = node = SYSCTL_ADD_NODE(ctx_list, child, OID_AUTO, "iflib", 5626 CTLFLAG_RD, NULL, "IFLIB fields"); 5627 oid_list = SYSCTL_CHILDREN(node); 5628 5629 SYSCTL_ADD_STRING(ctx_list, oid_list, OID_AUTO, "driver_version", 5630 CTLFLAG_RD, ctx->ifc_sctx->isc_driver_version, 0, 5631 "driver version"); 5632 5633 SYSCTL_ADD_U16(ctx_list, oid_list, OID_AUTO, "override_ntxqs", 5634 CTLFLAG_RWTUN, &ctx->ifc_sysctl_ntxqs, 0, 5635 "# of txqs to use, 0 => use default #"); 5636 SYSCTL_ADD_U16(ctx_list, oid_list, OID_AUTO, "override_nrxqs", 5637 CTLFLAG_RWTUN, &ctx->ifc_sysctl_nrxqs, 0, 5638 "# of rxqs to use, 0 => use default #"); 5639 SYSCTL_ADD_U16(ctx_list, oid_list, OID_AUTO, "override_qs_enable", 5640 CTLFLAG_RWTUN, &ctx->ifc_sysctl_qs_eq_override, 0, 5641 "permit #txq != #rxq"); 5642 SYSCTL_ADD_INT(ctx_list, oid_list, OID_AUTO, "disable_msix", 5643 CTLFLAG_RWTUN, &ctx->ifc_softc_ctx.isc_disable_msix, 0, 5644 "disable MSIX (default 0)"); 5645 SYSCTL_ADD_U16(ctx_list, oid_list, OID_AUTO, "rx_budget", 5646 CTLFLAG_RWTUN, &ctx->ifc_sysctl_rx_budget, 0, 5647 "set the rx budget"); 5648 5649 /* XXX change for per-queue sizes */ 5650 SYSCTL_ADD_PROC(ctx_list, oid_list, OID_AUTO, "override_ntxds", 5651 CTLTYPE_STRING|CTLFLAG_RWTUN, ctx, IFLIB_NTXD_HANDLER, 5652 mp_ndesc_handler, "A", 5653 "list of # of tx descriptors to use, 0 = use default #"); 5654 SYSCTL_ADD_PROC(ctx_list, oid_list, OID_AUTO, "override_nrxds", 5655 CTLTYPE_STRING|CTLFLAG_RWTUN, ctx, IFLIB_NRXD_HANDLER, 5656 mp_ndesc_handler, "A", 5657 "list of # of rx descriptors to use, 0 = use default #"); 5658 } 5659 5660 static void 5661 iflib_add_device_sysctl_post(if_ctx_t ctx) 5662 { 5663 if_shared_ctx_t sctx = ctx->ifc_sctx; 5664 if_softc_ctx_t scctx = &ctx->ifc_softc_ctx; 5665 device_t dev = iflib_get_dev(ctx); 5666 struct sysctl_oid_list *child; 5667 struct sysctl_ctx_list *ctx_list; 5668 iflib_fl_t fl; 5669 iflib_txq_t txq; 5670 iflib_rxq_t rxq; 5671 int i, j; 5672 char namebuf[NAME_BUFLEN]; 5673 char *qfmt; 5674 struct sysctl_oid *queue_node, *fl_node, *node; 5675 struct sysctl_oid_list *queue_list, *fl_list; 5676 ctx_list = device_get_sysctl_ctx(dev); 5677 5678 node = ctx->ifc_sysctl_node; 5679 child = SYSCTL_CHILDREN(node); 5680 5681 if (scctx->isc_ntxqsets > 100) 5682 qfmt = "txq%03d"; 5683 else if (scctx->isc_ntxqsets > 10) 5684 qfmt = "txq%02d"; 5685 else 5686 qfmt = "txq%d"; 5687 for (i = 0, txq = ctx->ifc_txqs; i < scctx->isc_ntxqsets; i++, txq++) { 5688 snprintf(namebuf, NAME_BUFLEN, qfmt, i); 5689 queue_node = SYSCTL_ADD_NODE(ctx_list, child, OID_AUTO, namebuf, 5690 CTLFLAG_RD, NULL, "Queue Name"); 5691 queue_list = SYSCTL_CHILDREN(queue_node); 5692 #if MEMORY_LOGGING 5693 SYSCTL_ADD_QUAD(ctx_list, queue_list, OID_AUTO, "txq_dequeued", 5694 CTLFLAG_RD, 5695 &txq->ift_dequeued, "total mbufs freed"); 5696 SYSCTL_ADD_QUAD(ctx_list, queue_list, OID_AUTO, "txq_enqueued", 5697 CTLFLAG_RD, 5698 &txq->ift_enqueued, "total mbufs enqueued"); 5699 #endif 5700 SYSCTL_ADD_QUAD(ctx_list, queue_list, OID_AUTO, "mbuf_defrag", 5701 CTLFLAG_RD, 5702 &txq->ift_mbuf_defrag, "# of times m_defrag was called"); 5703 SYSCTL_ADD_QUAD(ctx_list, queue_list, OID_AUTO, "m_pullups", 5704 CTLFLAG_RD, 5705 &txq->ift_pullups, "# of times m_pullup was called"); 5706 SYSCTL_ADD_QUAD(ctx_list, queue_list, OID_AUTO, "mbuf_defrag_failed", 5707 CTLFLAG_RD, 5708 &txq->ift_mbuf_defrag_failed, "# of times m_defrag failed"); 5709 SYSCTL_ADD_QUAD(ctx_list, queue_list, OID_AUTO, "no_desc_avail", 5710 CTLFLAG_RD, 5711 &txq->ift_no_desc_avail, "# of times no descriptors were available"); 5712 SYSCTL_ADD_QUAD(ctx_list, queue_list, OID_AUTO, "tx_map_failed", 5713 CTLFLAG_RD, 5714 &txq->ift_map_failed, "# of times dma map failed"); 5715 SYSCTL_ADD_QUAD(ctx_list, queue_list, OID_AUTO, "txd_encap_efbig", 5716 CTLFLAG_RD, 5717 &txq->ift_txd_encap_efbig, "# of times txd_encap returned EFBIG"); 5718 SYSCTL_ADD_QUAD(ctx_list, queue_list, OID_AUTO, "no_tx_dma_setup", 5719 CTLFLAG_RD, 5720 &txq->ift_no_tx_dma_setup, "# of times map failed for other than EFBIG"); 5721 SYSCTL_ADD_U16(ctx_list, queue_list, OID_AUTO, "txq_pidx", 5722 CTLFLAG_RD, 5723 &txq->ift_pidx, 1, "Producer Index"); 5724 SYSCTL_ADD_U16(ctx_list, queue_list, OID_AUTO, "txq_cidx", 5725 CTLFLAG_RD, 5726 &txq->ift_cidx, 1, "Consumer Index"); 5727 SYSCTL_ADD_U16(ctx_list, queue_list, OID_AUTO, "txq_cidx_processed", 5728 CTLFLAG_RD, 5729 &txq->ift_cidx_processed, 1, "Consumer Index seen by credit update"); 5730 SYSCTL_ADD_U16(ctx_list, queue_list, OID_AUTO, "txq_in_use", 5731 CTLFLAG_RD, 5732 &txq->ift_in_use, 1, "descriptors in use"); 5733 SYSCTL_ADD_QUAD(ctx_list, queue_list, OID_AUTO, "txq_processed", 5734 CTLFLAG_RD, 5735 &txq->ift_processed, "descriptors procesed for clean"); 5736 SYSCTL_ADD_QUAD(ctx_list, queue_list, OID_AUTO, "txq_cleaned", 5737 CTLFLAG_RD, 5738 &txq->ift_cleaned, "total cleaned"); 5739 SYSCTL_ADD_PROC(ctx_list, queue_list, OID_AUTO, "ring_state", 5740 CTLTYPE_STRING | CTLFLAG_RD, __DEVOLATILE(uint64_t *, &txq->ift_br->state), 5741 0, mp_ring_state_handler, "A", "soft ring state"); 5742 SYSCTL_ADD_COUNTER_U64(ctx_list, queue_list, OID_AUTO, "r_enqueues", 5743 CTLFLAG_RD, &txq->ift_br->enqueues, 5744 "# of enqueues to the mp_ring for this queue"); 5745 SYSCTL_ADD_COUNTER_U64(ctx_list, queue_list, OID_AUTO, "r_drops", 5746 CTLFLAG_RD, &txq->ift_br->drops, 5747 "# of drops in the mp_ring for this queue"); 5748 SYSCTL_ADD_COUNTER_U64(ctx_list, queue_list, OID_AUTO, "r_starts", 5749 CTLFLAG_RD, &txq->ift_br->starts, 5750 "# of normal consumer starts in the mp_ring for this queue"); 5751 SYSCTL_ADD_COUNTER_U64(ctx_list, queue_list, OID_AUTO, "r_stalls", 5752 CTLFLAG_RD, &txq->ift_br->stalls, 5753 "# of consumer stalls in the mp_ring for this queue"); 5754 SYSCTL_ADD_COUNTER_U64(ctx_list, queue_list, OID_AUTO, "r_restarts", 5755 CTLFLAG_RD, &txq->ift_br->restarts, 5756 "# of consumer restarts in the mp_ring for this queue"); 5757 SYSCTL_ADD_COUNTER_U64(ctx_list, queue_list, OID_AUTO, "r_abdications", 5758 CTLFLAG_RD, &txq->ift_br->abdications, 5759 "# of consumer abdications in the mp_ring for this queue"); 5760 } 5761 5762 if (scctx->isc_nrxqsets > 100) 5763 qfmt = "rxq%03d"; 5764 else if (scctx->isc_nrxqsets > 10) 5765 qfmt = "rxq%02d"; 5766 else 5767 qfmt = "rxq%d"; 5768 for (i = 0, rxq = ctx->ifc_rxqs; i < scctx->isc_nrxqsets; i++, rxq++) { 5769 snprintf(namebuf, NAME_BUFLEN, qfmt, i); 5770 queue_node = SYSCTL_ADD_NODE(ctx_list, child, OID_AUTO, namebuf, 5771 CTLFLAG_RD, NULL, "Queue Name"); 5772 queue_list = SYSCTL_CHILDREN(queue_node); 5773 if (sctx->isc_flags & IFLIB_HAS_RXCQ) { 5774 SYSCTL_ADD_U16(ctx_list, queue_list, OID_AUTO, "rxq_cq_pidx", 5775 CTLFLAG_RD, 5776 &rxq->ifr_cq_pidx, 1, "Producer Index"); 5777 SYSCTL_ADD_U16(ctx_list, queue_list, OID_AUTO, "rxq_cq_cidx", 5778 CTLFLAG_RD, 5779 &rxq->ifr_cq_cidx, 1, "Consumer Index"); 5780 } 5781 5782 for (j = 0, fl = rxq->ifr_fl; j < rxq->ifr_nfl; j++, fl++) { 5783 snprintf(namebuf, NAME_BUFLEN, "rxq_fl%d", j); 5784 fl_node = SYSCTL_ADD_NODE(ctx_list, queue_list, OID_AUTO, namebuf, 5785 CTLFLAG_RD, NULL, "freelist Name"); 5786 fl_list = SYSCTL_CHILDREN(fl_node); 5787 SYSCTL_ADD_U16(ctx_list, fl_list, OID_AUTO, "pidx", 5788 CTLFLAG_RD, 5789 &fl->ifl_pidx, 1, "Producer Index"); 5790 SYSCTL_ADD_U16(ctx_list, fl_list, OID_AUTO, "cidx", 5791 CTLFLAG_RD, 5792 &fl->ifl_cidx, 1, "Consumer Index"); 5793 SYSCTL_ADD_U16(ctx_list, fl_list, OID_AUTO, "credits", 5794 CTLFLAG_RD, 5795 &fl->ifl_credits, 1, "credits available"); 5796 #if MEMORY_LOGGING 5797 SYSCTL_ADD_QUAD(ctx_list, fl_list, OID_AUTO, "fl_m_enqueued", 5798 CTLFLAG_RD, 5799 &fl->ifl_m_enqueued, "mbufs allocated"); 5800 SYSCTL_ADD_QUAD(ctx_list, fl_list, OID_AUTO, "fl_m_dequeued", 5801 CTLFLAG_RD, 5802 &fl->ifl_m_dequeued, "mbufs freed"); 5803 SYSCTL_ADD_QUAD(ctx_list, fl_list, OID_AUTO, "fl_cl_enqueued", 5804 CTLFLAG_RD, 5805 &fl->ifl_cl_enqueued, "clusters allocated"); 5806 SYSCTL_ADD_QUAD(ctx_list, fl_list, OID_AUTO, "fl_cl_dequeued", 5807 CTLFLAG_RD, 5808 &fl->ifl_cl_dequeued, "clusters freed"); 5809 #endif 5810 5811 } 5812 } 5813 5814 } 5815 5816 #ifndef __NO_STRICT_ALIGNMENT 5817 static struct mbuf * 5818 iflib_fixup_rx(struct mbuf *m) 5819 { 5820 struct mbuf *n; 5821 5822 if (m->m_len <= (MCLBYTES - ETHER_HDR_LEN)) { 5823 bcopy(m->m_data, m->m_data + ETHER_HDR_LEN, m->m_len); 5824 m->m_data += ETHER_HDR_LEN; 5825 n = m; 5826 } else { 5827 MGETHDR(n, M_NOWAIT, MT_DATA); 5828 if (n == NULL) { 5829 m_freem(m); 5830 return (NULL); 5831 } 5832 bcopy(m->m_data, n->m_data, ETHER_HDR_LEN); 5833 m->m_data += ETHER_HDR_LEN; 5834 m->m_len -= ETHER_HDR_LEN; 5835 n->m_len = ETHER_HDR_LEN; 5836 M_MOVE_PKTHDR(n, m); 5837 n->m_next = m; 5838 } 5839 return (n); 5840 } 5841 #endif 5842