1/****************************************************************************** 2 3 Copyright (c) 2001-2017, Intel Corporation 4 All rights reserved. 5 6 Redistribution and use in source and binary forms, with or without 7 modification, are permitted provided that the following conditions are met: 8 9 1. Redistributions of source code must retain the above copyright notice, 10 this list of conditions and the following disclaimer. 11 12 2. Redistributions in binary form must reproduce the above copyright 13 notice, this list of conditions and the following disclaimer in the 14 documentation and/or other materials provided with the distribution. 15 16 3. Neither the name of the Intel Corporation nor the names of its 17 contributors may be used to endorse or promote products derived from 18 this software without specific prior written permission. 19 20 THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" 21 AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 22 IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 23 ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE 24 LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR 25 CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF 26 SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS 27 INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN 28 CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) 29 ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE 30 POSSIBILITY OF SUCH DAMAGE. 31 32******************************************************************************/ 33/*$FreeBSD$*/ 34 35 36#ifndef IXGBE_STANDALONE_BUILD 37#include "opt_inet.h" 38#include "opt_inet6.h" 39#include "opt_rss.h" 40#endif 41 42#include "ixgbe.h" 43 44 45/************************************************************************ 46 * Local Function prototypes 47 ************************************************************************/ 48static int ixgbe_isc_txd_encap(void *arg, if_pkt_info_t pi); 49static void ixgbe_isc_txd_flush(void *arg, uint16_t txqid, qidx_t pidx); 50static int ixgbe_isc_txd_credits_update(void *arg, uint16_t txqid, bool clear); 51 52static void ixgbe_isc_rxd_refill(void *arg, if_rxd_update_t iru); 53static void ixgbe_isc_rxd_flush(void *arg, uint16_t qsidx, uint8_t flidx __unused, qidx_t pidx); 54static int ixgbe_isc_rxd_available(void *arg, uint16_t qsidx, qidx_t pidx, 55 qidx_t budget); 56static int ixgbe_isc_rxd_pkt_get(void *arg, if_rxd_info_t ri); 57 58static void ixgbe_rx_checksum(u32 staterr, if_rxd_info_t ri, u32 ptype); 59static int ixgbe_tx_ctx_setup(struct ixgbe_adv_tx_context_desc *, if_pkt_info_t); 60 61extern void ixgbe_if_enable_intr(if_ctx_t ctx); 62static int ixgbe_determine_rsstype(u16 pkt_info); 63 64struct if_txrx ixgbe_txrx = { 65 .ift_txd_encap = ixgbe_isc_txd_encap, 66 .ift_txd_flush = ixgbe_isc_txd_flush, 67 .ift_txd_credits_update = ixgbe_isc_txd_credits_update, 68 .ift_rxd_available = ixgbe_isc_rxd_available, 69 .ift_rxd_pkt_get = ixgbe_isc_rxd_pkt_get, 70 .ift_rxd_refill = ixgbe_isc_rxd_refill, 71 .ift_rxd_flush = ixgbe_isc_rxd_flush, 72 .ift_legacy_intr = NULL 73}; 74 75extern if_shared_ctx_t ixgbe_sctx; 76 77/************************************************************************ 78 * ixgbe_tx_ctx_setup 79 * 80 * Advanced Context Descriptor setup for VLAN, CSUM or TSO 81 * 82 ************************************************************************/ 83static int 84ixgbe_tx_ctx_setup(struct ixgbe_adv_tx_context_desc *TXD, if_pkt_info_t pi) 85{ 86 u32 vlan_macip_lens, type_tucmd_mlhl; 87 u32 olinfo_status, mss_l4len_idx, pktlen, offload; 88 u8 ehdrlen; 89 90 offload = TRUE; 91 olinfo_status = mss_l4len_idx = vlan_macip_lens = type_tucmd_mlhl = 0; 92 /* VLAN MACLEN IPLEN */ 93 vlan_macip_lens |= (htole16(pi->ipi_vtag) << IXGBE_ADVTXD_VLAN_SHIFT); 94 95 /* 96 * Some of our VF devices need a context descriptor for every 97 * packet. That means the ehdrlen needs to be non-zero in order 98 * for the host driver not to flag a malicious event. The stack 99 * will most likely populate this for all other reasons of why 100 * this function was called. 101 */ 102 if (pi->ipi_ehdrlen == 0) { 103 ehdrlen = ETHER_HDR_LEN; 104 ehdrlen += (pi->ipi_vtag != 0) ? ETHER_VLAN_ENCAP_LEN : 0; 105 } else 106 ehdrlen = pi->ipi_ehdrlen; 107 vlan_macip_lens |= ehdrlen << IXGBE_ADVTXD_MACLEN_SHIFT; 108 109 pktlen = pi->ipi_len; 110 /* First check if TSO is to be used */ 111 if (pi->ipi_csum_flags & CSUM_TSO) { 112 /* This is used in the transmit desc in encap */ 113 pktlen = pi->ipi_len - ehdrlen - pi->ipi_ip_hlen - pi->ipi_tcp_hlen; 114 mss_l4len_idx |= (pi->ipi_tso_segsz << IXGBE_ADVTXD_MSS_SHIFT); 115 mss_l4len_idx |= (pi->ipi_tcp_hlen << IXGBE_ADVTXD_L4LEN_SHIFT); 116 } 117 118 olinfo_status |= pktlen << IXGBE_ADVTXD_PAYLEN_SHIFT; 119 120 if (pi->ipi_flags & IPI_TX_IPV4) { 121 type_tucmd_mlhl |= IXGBE_ADVTXD_TUCMD_IPV4; 122 /* Tell transmit desc to also do IPv4 checksum. */ 123 if (pi->ipi_csum_flags & (CSUM_IP|CSUM_TSO)) 124 olinfo_status |= IXGBE_TXD_POPTS_IXSM << 8; 125 } else if (pi->ipi_flags & IPI_TX_IPV6) 126 type_tucmd_mlhl |= IXGBE_ADVTXD_TUCMD_IPV6; 127 else 128 offload = FALSE; 129 130 vlan_macip_lens |= pi->ipi_ip_hlen; 131 132 switch (pi->ipi_ipproto) { 133 case IPPROTO_TCP: 134 if (pi->ipi_csum_flags & (CSUM_IP_TCP | CSUM_IP6_TCP | CSUM_TSO)) 135 type_tucmd_mlhl |= IXGBE_ADVTXD_TUCMD_L4T_TCP; 136 else 137 offload = FALSE; 138 break; 139 case IPPROTO_UDP: 140 if (pi->ipi_csum_flags & (CSUM_IP_UDP | CSUM_IP6_UDP)) 141 type_tucmd_mlhl |= IXGBE_ADVTXD_TUCMD_L4T_UDP; 142 else 143 offload = FALSE; 144 break; 145 case IPPROTO_SCTP: 146 if (pi->ipi_csum_flags & (CSUM_IP_SCTP | CSUM_IP6_SCTP)) 147 type_tucmd_mlhl |= IXGBE_ADVTXD_TUCMD_L4T_SCTP; 148 else 149 offload = FALSE; 150 break; 151 default: 152 offload = FALSE; 153 break; 154 } 155/* Insert L4 checksum into data descriptors */ 156 if (offload) 157 olinfo_status |= IXGBE_TXD_POPTS_TXSM << 8; 158 159 type_tucmd_mlhl |= IXGBE_ADVTXD_DCMD_DEXT | IXGBE_ADVTXD_DTYP_CTXT; 160 161 /* Now copy bits into descriptor */ 162 TXD->vlan_macip_lens = htole32(vlan_macip_lens); 163 TXD->type_tucmd_mlhl = htole32(type_tucmd_mlhl); 164 TXD->seqnum_seed = htole32(0); 165 TXD->mss_l4len_idx = htole32(mss_l4len_idx); 166 167 return (olinfo_status); 168} /* ixgbe_tx_ctx_setup */ 169 170/************************************************************************ 171 * ixgbe_isc_txd_encap 172 ************************************************************************/ 173static int 174ixgbe_isc_txd_encap(void *arg, if_pkt_info_t pi) 175{ 176 struct adapter *sc = arg; 177 if_softc_ctx_t scctx = sc->shared; 178 struct ix_tx_queue *que = &sc->tx_queues[pi->ipi_qsidx]; 179 struct tx_ring *txr = &que->txr; 180 int nsegs = pi->ipi_nsegs; 181 bus_dma_segment_t *segs = pi->ipi_segs; 182 union ixgbe_adv_tx_desc *txd = NULL; 183 struct ixgbe_adv_tx_context_desc *TXD; 184 int i, j, first, pidx_last; 185 u32 olinfo_status, cmd, flags; 186 qidx_t ntxd; 187 188 cmd = (IXGBE_ADVTXD_DTYP_DATA | 189 IXGBE_ADVTXD_DCMD_IFCS | IXGBE_ADVTXD_DCMD_DEXT); 190 191 if (pi->ipi_mflags & M_VLANTAG) 192 cmd |= IXGBE_ADVTXD_DCMD_VLE; 193 194 i = first = pi->ipi_pidx; 195 flags = (pi->ipi_flags & IPI_TX_INTR) ? IXGBE_TXD_CMD_RS : 0; 196 ntxd = scctx->isc_ntxd[0]; 197 198 TXD = (struct ixgbe_adv_tx_context_desc *) &txr->tx_base[first]; 199 if ((pi->ipi_csum_flags & CSUM_OFFLOAD) || 200 (sc->feat_en & IXGBE_FEATURE_NEEDS_CTXD) || 201 pi->ipi_vtag) { 202 /********************************************* 203 * Set up the appropriate offload context 204 * this will consume the first descriptor 205 *********************************************/ 206 olinfo_status = ixgbe_tx_ctx_setup(TXD, pi); 207 if (pi->ipi_csum_flags & CSUM_TSO) { 208 cmd |= IXGBE_ADVTXD_DCMD_TSE; 209 ++txr->tso_tx; 210 } 211 212 if (++i == scctx->isc_ntxd[0]) 213 i = 0; 214 } else { 215 /* Indicate the whole packet as payload when not doing TSO */ 216 olinfo_status = pi->ipi_len << IXGBE_ADVTXD_PAYLEN_SHIFT; 217 } 218 219 olinfo_status |= IXGBE_ADVTXD_CC; 220 pidx_last = 0; 221 for (j = 0; j < nsegs; j++) { 222 bus_size_t seglen; 223 224 txd = &txr->tx_base[i]; 225 seglen = segs[j].ds_len; 226 227 txd->read.buffer_addr = htole64(segs[j].ds_addr); 228 txd->read.cmd_type_len = htole32(cmd | seglen); 229 txd->read.olinfo_status = htole32(olinfo_status); 230 231 pidx_last = i; 232 if (++i == scctx->isc_ntxd[0]) { 233 i = 0; 234 } 235 } 236 237 if (flags) { 238 txr->tx_rsq[txr->tx_rs_pidx] = pidx_last; 239 txr->tx_rs_pidx = (txr->tx_rs_pidx + 1) & (ntxd - 1); 240 } 241 txd->read.cmd_type_len |= htole32(IXGBE_TXD_CMD_EOP | flags); 242 243 txr->bytes += pi->ipi_len; 244 pi->ipi_new_pidx = i; 245 246 ++txr->total_packets; 247 248 return (0); 249} /* ixgbe_isc_txd_encap */ 250 251/************************************************************************ 252 * ixgbe_isc_txd_flush 253 ************************************************************************/ 254static void 255ixgbe_isc_txd_flush(void *arg, uint16_t txqid, qidx_t pidx) 256{ 257 struct adapter *sc = arg; 258 struct ix_tx_queue *que = &sc->tx_queues[txqid]; 259 struct tx_ring *txr = &que->txr; 260 261 IXGBE_WRITE_REG(&sc->hw, txr->tail, pidx); 262} /* ixgbe_isc_txd_flush */ 263 264/************************************************************************ 265 * ixgbe_isc_txd_credits_update 266 ************************************************************************/ 267static int 268ixgbe_isc_txd_credits_update(void *arg, uint16_t txqid, bool clear) 269{ 270 struct adapter *sc = arg; 271 if_softc_ctx_t scctx = sc->shared; 272 struct ix_tx_queue *que = &sc->tx_queues[txqid]; 273 struct tx_ring *txr = &que->txr; 274 qidx_t processed = 0; 275 int updated; 276 qidx_t cur, prev, ntxd, rs_cidx; 277 int32_t delta; 278 uint8_t status; 279 280 rs_cidx = txr->tx_rs_cidx; 281 if (rs_cidx == txr->tx_rs_pidx) 282 return (0); 283 284 cur = txr->tx_rsq[rs_cidx]; 285 status = txr->tx_base[cur].wb.status; 286 updated = !!(status & IXGBE_TXD_STAT_DD); 287 288 if (!updated) 289 return (0); 290 291 /* If clear is false just let caller know that there 292 * are descriptors to reclaim */ 293 if (!clear) 294 return (1); 295 296 prev = txr->tx_cidx_processed; 297 ntxd = scctx->isc_ntxd[0]; 298 do { 299 MPASS(prev != cur); 300 delta = (int32_t)cur - (int32_t)prev; 301 if (delta < 0) 302 delta += ntxd; 303 MPASS(delta > 0); 304 305 processed += delta; 306 prev = cur; 307 rs_cidx = (rs_cidx + 1) & (ntxd - 1); 308 if (rs_cidx == txr->tx_rs_pidx) 309 break; 310 311 cur = txr->tx_rsq[rs_cidx]; 312 status = txr->tx_base[cur].wb.status; 313 } while ((status & IXGBE_TXD_STAT_DD)); 314 315 txr->tx_rs_cidx = rs_cidx; 316 txr->tx_cidx_processed = prev; 317 318 return (processed); 319} /* ixgbe_isc_txd_credits_update */ 320 321/************************************************************************ 322 * ixgbe_isc_rxd_refill 323 ************************************************************************/ 324static void 325ixgbe_isc_rxd_refill(void *arg, if_rxd_update_t iru) 326{ 327 struct adapter *sc = arg; 328 struct ix_rx_queue *que = &sc->rx_queues[iru->iru_qsidx]; 329 struct rx_ring *rxr = &que->rxr; 330 uint64_t *paddrs; 331 int i; 332 uint32_t next_pidx, pidx; 333 uint16_t count; 334 335 paddrs = iru->iru_paddrs; 336 pidx = iru->iru_pidx; 337 count = iru->iru_count; 338 339 for (i = 0, next_pidx = pidx; i < count; i++) { 340 rxr->rx_base[next_pidx].read.pkt_addr = htole64(paddrs[i]); 341 if (++next_pidx == sc->shared->isc_nrxd[0]) 342 next_pidx = 0; 343 } 344} /* ixgbe_isc_rxd_refill */ 345 346/************************************************************************ 347 * ixgbe_isc_rxd_flush 348 ************************************************************************/ 349static void 350ixgbe_isc_rxd_flush(void *arg, uint16_t qsidx, uint8_t flidx __unused, qidx_t pidx) 351{ 352 struct adapter *sc = arg; 353 struct ix_rx_queue *que = &sc->rx_queues[qsidx]; 354 struct rx_ring *rxr = &que->rxr; 355 356 IXGBE_WRITE_REG(&sc->hw, rxr->tail, pidx); 357} /* ixgbe_isc_rxd_flush */ 358 359/************************************************************************ 360 * ixgbe_isc_rxd_available 361 ************************************************************************/ 362static int 363ixgbe_isc_rxd_available(void *arg, uint16_t qsidx, qidx_t pidx, qidx_t budget) 364{ 365 struct adapter *sc = arg; 366 struct ix_rx_queue *que = &sc->rx_queues[qsidx]; 367 struct rx_ring *rxr = &que->rxr; 368 union ixgbe_adv_rx_desc *rxd; 369 u32 staterr; 370 int cnt, i, nrxd; 371 372 nrxd = sc->shared->isc_nrxd[0]; 373 for (cnt = 0, i = pidx; cnt < nrxd && cnt <= budget;) { 374 rxd = &rxr->rx_base[i]; 375 staterr = le32toh(rxd->wb.upper.status_error); 376 377 if ((staterr & IXGBE_RXD_STAT_DD) == 0) 378 break; 379 if (++i == nrxd) 380 i = 0; 381 if (staterr & IXGBE_RXD_STAT_EOP) 382 cnt++; 383 } 384 return (cnt); 385} /* ixgbe_isc_rxd_available */ 386 387/************************************************************************ 388 * ixgbe_isc_rxd_pkt_get 389 * 390 * Routine sends data which has been dma'ed into host memory 391 * to upper layer. Initialize ri structure. 392 * 393 * Returns 0 upon success, errno on failure 394 ************************************************************************/ 395 396static int 397ixgbe_isc_rxd_pkt_get(void *arg, if_rxd_info_t ri) 398{ 399 struct adapter *adapter = arg; 400 struct ix_rx_queue *que = &adapter->rx_queues[ri->iri_qsidx]; 401 struct rx_ring *rxr = &que->rxr; 402 struct ifnet *ifp = iflib_get_ifp(adapter->ctx); 403 union ixgbe_adv_rx_desc *rxd; 404 405 u16 pkt_info, len, cidx, i; 406 u16 vtag = 0; 407 u32 ptype; 408 u32 staterr = 0; 409 bool eop; 410 411 i = 0; 412 cidx = ri->iri_cidx; 413 do { 414 rxd = &rxr->rx_base[cidx]; 415 staterr = le32toh(rxd->wb.upper.status_error); 416 pkt_info = le16toh(rxd->wb.lower.lo_dword.hs_rss.pkt_info); 417 418 /* Error Checking then decrement count */ 419 MPASS ((staterr & IXGBE_RXD_STAT_DD) != 0); 420 421 len = le16toh(rxd->wb.upper.length); 422 ptype = le32toh(rxd->wb.lower.lo_dword.data) & 423 IXGBE_RXDADV_PKTTYPE_MASK; 424 425 ri->iri_len += len; 426 rxr->bytes += len; 427 428 rxd->wb.upper.status_error = 0; 429 eop = ((staterr & IXGBE_RXD_STAT_EOP) != 0); 430 431 if ( (rxr->vtag_strip) && (staterr & IXGBE_RXD_STAT_VP) ) { 432 vtag = le16toh(rxd->wb.upper.vlan); 433 } else { 434 vtag = 0; 435 } 436 437 /* Make sure bad packets are discarded */ 438 if (eop && (staterr & IXGBE_RXDADV_ERR_FRAME_ERR_MASK) != 0) { 439 440#if __FreeBSD_version >= 1100036 441 if (adapter->feat_en & IXGBE_FEATURE_VF) 442 if_inc_counter(ifp, IFCOUNTER_IERRORS, 1); 443#endif 444 445 rxr->rx_discarded++; 446 return (EBADMSG); 447 } 448 ri->iri_frags[i].irf_flid = 0; 449 ri->iri_frags[i].irf_idx = cidx; 450 ri->iri_frags[i].irf_len = len; 451 if (++cidx == adapter->shared->isc_nrxd[0]) 452 cidx = 0; 453 i++; 454 /* even a 16K packet shouldn't consume more than 8 clusters */ 455 MPASS(i < 9); 456 } while (!eop); 457 458 rxr->rx_packets++; 459 rxr->packets++; 460 rxr->rx_bytes += ri->iri_len; 461 462 if ((ifp->if_capenable & IFCAP_RXCSUM) != 0) 463 ixgbe_rx_checksum(staterr, ri, ptype); 464 465 ri->iri_flowid = le32toh(rxd->wb.lower.hi_dword.rss); 466 ri->iri_rsstype = ixgbe_determine_rsstype(pkt_info); 467 if ((adapter->feat_en & IXGBE_FEATURE_RSS) == 0) { 468 if (ri->iri_rsstype == M_HASHTYPE_OPAQUE) 469 ri->iri_rsstype = M_HASHTYPE_NONE; 470 else 471 ri->iri_rsstype = M_HASHTYPE_OPAQUE_HASH; 472 } 473 ri->iri_vtag = vtag; 474 ri->iri_nfrags = i; 475 if (vtag) 476 ri->iri_flags |= M_VLANTAG; 477 return (0); 478} /* ixgbe_isc_rxd_pkt_get */ 479 480/************************************************************************ 481 * ixgbe_rx_checksum 482 * 483 * Verify that the hardware indicated that the checksum is valid. 484 * Inform the stack about the status of checksum so that stack 485 * doesn't spend time verifying the checksum. 486 ************************************************************************/ 487static void 488ixgbe_rx_checksum(u32 staterr, if_rxd_info_t ri, u32 ptype) 489{ 490 u16 status = (u16)staterr; 491 u8 errors = (u8)(staterr >> 24); 492 bool sctp = false; 493 494 if ((ptype & IXGBE_RXDADV_PKTTYPE_ETQF) == 0 && 495 (ptype & IXGBE_RXDADV_PKTTYPE_SCTP) != 0) 496 sctp = TRUE; 497 498 /* IPv4 checksum */ 499 if (status & IXGBE_RXD_STAT_IPCS) { 500 if (!(errors & IXGBE_RXD_ERR_IPE)) { 501 /* IP Checksum Good */ 502 ri->iri_csum_flags = CSUM_IP_CHECKED | CSUM_IP_VALID; 503 } else 504 ri->iri_csum_flags = 0; 505 } 506 /* TCP/UDP/SCTP checksum */ 507 if (status & IXGBE_RXD_STAT_L4CS) { 508 u64 type = (CSUM_DATA_VALID | CSUM_PSEUDO_HDR); 509#if __FreeBSD_version >= 800000 510 if (sctp) 511 type = CSUM_SCTP_VALID; 512#endif 513 if (!(errors & IXGBE_RXD_ERR_TCPE)) { 514 ri->iri_csum_flags |= type; 515 if (!sctp) 516 ri->iri_csum_data = htons(0xffff); 517 } 518 } 519} /* ixgbe_rx_checksum */ 520 521/************************************************************************ 522 * ixgbe_determine_rsstype 523 * 524 * Parse the packet type to determine the appropriate hash 525 ************************************************************************/ 526static int 527ixgbe_determine_rsstype(u16 pkt_info) 528{ 529 switch (pkt_info & IXGBE_RXDADV_RSSTYPE_MASK) { 530 case IXGBE_RXDADV_RSSTYPE_IPV4_TCP: 531 return M_HASHTYPE_RSS_TCP_IPV4; 532 case IXGBE_RXDADV_RSSTYPE_IPV4: 533 return M_HASHTYPE_RSS_IPV4; 534 case IXGBE_RXDADV_RSSTYPE_IPV6_TCP: 535 return M_HASHTYPE_RSS_TCP_IPV6; 536 case IXGBE_RXDADV_RSSTYPE_IPV6_EX: 537 return M_HASHTYPE_RSS_IPV6_EX; 538 case IXGBE_RXDADV_RSSTYPE_IPV6: 539 return M_HASHTYPE_RSS_IPV6; 540 case IXGBE_RXDADV_RSSTYPE_IPV6_TCP_EX: 541 return M_HASHTYPE_RSS_TCP_IPV6_EX; 542 case IXGBE_RXDADV_RSSTYPE_IPV4_UDP: 543 return M_HASHTYPE_RSS_UDP_IPV4; 544 case IXGBE_RXDADV_RSSTYPE_IPV6_UDP: 545 return M_HASHTYPE_RSS_UDP_IPV6; 546 case IXGBE_RXDADV_RSSTYPE_IPV6_UDP_EX: 547 return M_HASHTYPE_RSS_UDP_IPV6_EX; 548 default: 549 return M_HASHTYPE_OPAQUE; 550 } 551} /* ixgbe_determine_rsstype */ 552