vmxnet3.c (78702B)
1/* 2 * QEMU VMWARE VMXNET3 paravirtual NIC 3 * 4 * Copyright (c) 2012 Ravello Systems LTD (http://ravellosystems.com) 5 * 6 * Developed by Daynix Computing LTD (http://www.daynix.com) 7 * 8 * Authors: 9 * Dmitry Fleytman <dmitry@daynix.com> 10 * Tamir Shomer <tamirs@daynix.com> 11 * Yan Vugenfirer <yan@daynix.com> 12 * 13 * This work is licensed under the terms of the GNU GPL, version 2. 14 * See the COPYING file in the top-level directory. 15 * 16 */ 17 18#include "qemu/osdep.h" 19#include "hw/hw.h" 20#include "hw/pci/pci.h" 21#include "hw/qdev-properties.h" 22#include "net/tap.h" 23#include "net/checksum.h" 24#include "sysemu/sysemu.h" 25#include "qemu/bswap.h" 26#include "qemu/log.h" 27#include "qemu/module.h" 28#include "hw/pci/msix.h" 29#include "hw/pci/msi.h" 30#include "migration/register.h" 31#include "migration/vmstate.h" 32 33#include "vmxnet3.h" 34#include "vmxnet3_defs.h" 35#include "vmxnet_debug.h" 36#include "vmware_utils.h" 37#include "net_tx_pkt.h" 38#include "net_rx_pkt.h" 39#include "qom/object.h" 40 41#define PCI_DEVICE_ID_VMWARE_VMXNET3_REVISION 0x1 42#define VMXNET3_MSIX_BAR_SIZE 0x2000 43#define MIN_BUF_SIZE 60 44 45/* Compatibility flags for migration */ 46#define VMXNET3_COMPAT_FLAG_OLD_MSI_OFFSETS_BIT 0 47#define VMXNET3_COMPAT_FLAG_OLD_MSI_OFFSETS \ 48 (1 << VMXNET3_COMPAT_FLAG_OLD_MSI_OFFSETS_BIT) 49#define VMXNET3_COMPAT_FLAG_DISABLE_PCIE_BIT 1 50#define VMXNET3_COMPAT_FLAG_DISABLE_PCIE \ 51 (1 << VMXNET3_COMPAT_FLAG_DISABLE_PCIE_BIT) 52 53#define VMXNET3_EXP_EP_OFFSET (0x48) 54#define VMXNET3_MSI_OFFSET(s) \ 55 ((s)->compat_flags & VMXNET3_COMPAT_FLAG_OLD_MSI_OFFSETS ? 0x50 : 0x84) 56#define VMXNET3_MSIX_OFFSET(s) \ 57 ((s)->compat_flags & VMXNET3_COMPAT_FLAG_OLD_MSI_OFFSETS ? 0 : 0x9c) 58#define VMXNET3_DSN_OFFSET (0x100) 59 60#define VMXNET3_BAR0_IDX (0) 61#define VMXNET3_BAR1_IDX (1) 62#define VMXNET3_MSIX_BAR_IDX (2) 63 64#define VMXNET3_OFF_MSIX_TABLE (0x000) 65#define VMXNET3_OFF_MSIX_PBA(s) \ 66 ((s)->compat_flags & VMXNET3_COMPAT_FLAG_OLD_MSI_OFFSETS ? 0x800 : 0x1000) 67 68/* Link speed in Mbps should be shifted by 16 */ 69#define VMXNET3_LINK_SPEED (1000 << 16) 70 71/* Link status: 1 - up, 0 - down. */ 72#define VMXNET3_LINK_STATUS_UP 0x1 73 74/* Least significant bit should be set for revision and version */ 75#define VMXNET3_UPT_REVISION 0x1 76#define VMXNET3_DEVICE_REVISION 0x1 77 78/* Number of interrupt vectors for non-MSIx modes */ 79#define VMXNET3_MAX_NMSIX_INTRS (1) 80 81/* Macros for rings descriptors access */ 82#define VMXNET3_READ_TX_QUEUE_DESCR8(_d, dpa, field) \ 83 (vmw_shmem_ld8(_d, dpa + offsetof(struct Vmxnet3_TxQueueDesc, field))) 84 85#define VMXNET3_WRITE_TX_QUEUE_DESCR8(_d, dpa, field, value) \ 86 (vmw_shmem_st8(_d, dpa + offsetof(struct Vmxnet3_TxQueueDesc, field, value))) 87 88#define VMXNET3_READ_TX_QUEUE_DESCR32(_d, dpa, field) \ 89 (vmw_shmem_ld32(_d, dpa + offsetof(struct Vmxnet3_TxQueueDesc, field))) 90 91#define VMXNET3_WRITE_TX_QUEUE_DESCR32(_d, dpa, field, value) \ 92 (vmw_shmem_st32(_d, dpa + offsetof(struct Vmxnet3_TxQueueDesc, field), value)) 93 94#define VMXNET3_READ_TX_QUEUE_DESCR64(_d, dpa, field) \ 95 (vmw_shmem_ld64(_d, dpa + offsetof(struct Vmxnet3_TxQueueDesc, field))) 96 97#define VMXNET3_WRITE_TX_QUEUE_DESCR64(_d, dpa, field, value) \ 98 (vmw_shmem_st64(_d, dpa + offsetof(struct Vmxnet3_TxQueueDesc, field), value)) 99 100#define VMXNET3_READ_RX_QUEUE_DESCR64(_d, dpa, field) \ 101 (vmw_shmem_ld64(_d, dpa + offsetof(struct Vmxnet3_RxQueueDesc, field))) 102 103#define VMXNET3_READ_RX_QUEUE_DESCR32(_d, dpa, field) \ 104 (vmw_shmem_ld32(_d, dpa + offsetof(struct Vmxnet3_RxQueueDesc, field))) 105 106#define VMXNET3_WRITE_RX_QUEUE_DESCR64(_d, dpa, field, value) \ 107 (vmw_shmem_st64(_d, dpa + offsetof(struct Vmxnet3_RxQueueDesc, field), value)) 108 109#define VMXNET3_WRITE_RX_QUEUE_DESCR8(_d, dpa, field, value) \ 110 (vmw_shmem_st8(_d, dpa + offsetof(struct Vmxnet3_RxQueueDesc, field), value)) 111 112/* Macros for guest driver shared area access */ 113#define VMXNET3_READ_DRV_SHARED64(_d, shpa, field) \ 114 (vmw_shmem_ld64(_d, shpa + offsetof(struct Vmxnet3_DriverShared, field))) 115 116#define VMXNET3_READ_DRV_SHARED32(_d, shpa, field) \ 117 (vmw_shmem_ld32(_d, shpa + offsetof(struct Vmxnet3_DriverShared, field))) 118 119#define VMXNET3_WRITE_DRV_SHARED32(_d, shpa, field, val) \ 120 (vmw_shmem_st32(_d, shpa + offsetof(struct Vmxnet3_DriverShared, field), val)) 121 122#define VMXNET3_READ_DRV_SHARED16(_d, shpa, field) \ 123 (vmw_shmem_ld16(_d, shpa + offsetof(struct Vmxnet3_DriverShared, field))) 124 125#define VMXNET3_READ_DRV_SHARED8(_d, shpa, field) \ 126 (vmw_shmem_ld8(_d, shpa + offsetof(struct Vmxnet3_DriverShared, field))) 127 128#define VMXNET3_READ_DRV_SHARED(_d, shpa, field, b, l) \ 129 (vmw_shmem_read(_d, shpa + offsetof(struct Vmxnet3_DriverShared, field), b, l)) 130 131#define VMXNET_FLAG_IS_SET(field, flag) (((field) & (flag)) == (flag)) 132 133struct VMXNET3Class { 134 PCIDeviceClass parent_class; 135 DeviceRealize parent_dc_realize; 136}; 137typedef struct VMXNET3Class VMXNET3Class; 138 139DECLARE_CLASS_CHECKERS(VMXNET3Class, VMXNET3_DEVICE, 140 TYPE_VMXNET3) 141 142static inline void vmxnet3_ring_init(PCIDevice *d, 143 Vmxnet3Ring *ring, 144 hwaddr pa, 145 uint32_t size, 146 uint32_t cell_size, 147 bool zero_region) 148{ 149 ring->pa = pa; 150 ring->size = size; 151 ring->cell_size = cell_size; 152 ring->gen = VMXNET3_INIT_GEN; 153 ring->next = 0; 154 155 if (zero_region) { 156 vmw_shmem_set(d, pa, 0, size * cell_size); 157 } 158} 159 160#define VMXNET3_RING_DUMP(macro, ring_name, ridx, r) \ 161 macro("%s#%d: base %" PRIx64 " size %u cell_size %u gen %d next %u", \ 162 (ring_name), (ridx), \ 163 (r)->pa, (r)->size, (r)->cell_size, (r)->gen, (r)->next) 164 165static inline void vmxnet3_ring_inc(Vmxnet3Ring *ring) 166{ 167 if (++ring->next >= ring->size) { 168 ring->next = 0; 169 ring->gen ^= 1; 170 } 171} 172 173static inline void vmxnet3_ring_dec(Vmxnet3Ring *ring) 174{ 175 if (ring->next-- == 0) { 176 ring->next = ring->size - 1; 177 ring->gen ^= 1; 178 } 179} 180 181static inline hwaddr vmxnet3_ring_curr_cell_pa(Vmxnet3Ring *ring) 182{ 183 return ring->pa + ring->next * ring->cell_size; 184} 185 186static inline void vmxnet3_ring_read_curr_cell(PCIDevice *d, Vmxnet3Ring *ring, 187 void *buff) 188{ 189 vmw_shmem_read(d, vmxnet3_ring_curr_cell_pa(ring), buff, ring->cell_size); 190} 191 192static inline void vmxnet3_ring_write_curr_cell(PCIDevice *d, Vmxnet3Ring *ring, 193 void *buff) 194{ 195 vmw_shmem_write(d, vmxnet3_ring_curr_cell_pa(ring), buff, ring->cell_size); 196} 197 198static inline size_t vmxnet3_ring_curr_cell_idx(Vmxnet3Ring *ring) 199{ 200 return ring->next; 201} 202 203static inline uint8_t vmxnet3_ring_curr_gen(Vmxnet3Ring *ring) 204{ 205 return ring->gen; 206} 207 208/* Debug trace-related functions */ 209static inline void 210vmxnet3_dump_tx_descr(struct Vmxnet3_TxDesc *descr) 211{ 212 VMW_PKPRN("TX DESCR: " 213 "addr %" PRIx64 ", len: %d, gen: %d, rsvd: %d, " 214 "dtype: %d, ext1: %d, msscof: %d, hlen: %d, om: %d, " 215 "eop: %d, cq: %d, ext2: %d, ti: %d, tci: %d", 216 descr->addr, descr->len, descr->gen, descr->rsvd, 217 descr->dtype, descr->ext1, descr->msscof, descr->hlen, descr->om, 218 descr->eop, descr->cq, descr->ext2, descr->ti, descr->tci); 219} 220 221static inline void 222vmxnet3_dump_virt_hdr(struct virtio_net_hdr *vhdr) 223{ 224 VMW_PKPRN("VHDR: flags 0x%x, gso_type: 0x%x, hdr_len: %d, gso_size: %d, " 225 "csum_start: %d, csum_offset: %d", 226 vhdr->flags, vhdr->gso_type, vhdr->hdr_len, vhdr->gso_size, 227 vhdr->csum_start, vhdr->csum_offset); 228} 229 230static inline void 231vmxnet3_dump_rx_descr(struct Vmxnet3_RxDesc *descr) 232{ 233 VMW_PKPRN("RX DESCR: addr %" PRIx64 ", len: %d, gen: %d, rsvd: %d, " 234 "dtype: %d, ext1: %d, btype: %d", 235 descr->addr, descr->len, descr->gen, 236 descr->rsvd, descr->dtype, descr->ext1, descr->btype); 237} 238 239/* Interrupt management */ 240 241/* 242 * This function returns sign whether interrupt line is in asserted state 243 * This depends on the type of interrupt used. For INTX interrupt line will 244 * be asserted until explicit deassertion, for MSI(X) interrupt line will 245 * be deasserted automatically due to notification semantics of the MSI(X) 246 * interrupts 247 */ 248static bool _vmxnet3_assert_interrupt_line(VMXNET3State *s, uint32_t int_idx) 249{ 250 PCIDevice *d = PCI_DEVICE(s); 251 252 if (s->msix_used && msix_enabled(d)) { 253 VMW_IRPRN("Sending MSI-X notification for vector %u", int_idx); 254 msix_notify(d, int_idx); 255 return false; 256 } 257 if (msi_enabled(d)) { 258 VMW_IRPRN("Sending MSI notification for vector %u", int_idx); 259 msi_notify(d, int_idx); 260 return false; 261 } 262 263 VMW_IRPRN("Asserting line for interrupt %u", int_idx); 264 pci_irq_assert(d); 265 return true; 266} 267 268static void _vmxnet3_deassert_interrupt_line(VMXNET3State *s, int lidx) 269{ 270 PCIDevice *d = PCI_DEVICE(s); 271 272 /* 273 * This function should never be called for MSI(X) interrupts 274 * because deassertion never required for message interrupts 275 */ 276 assert(!s->msix_used || !msix_enabled(d)); 277 /* 278 * This function should never be called for MSI(X) interrupts 279 * because deassertion never required for message interrupts 280 */ 281 assert(!msi_enabled(d)); 282 283 VMW_IRPRN("Deasserting line for interrupt %u", lidx); 284 pci_irq_deassert(d); 285} 286 287static void vmxnet3_update_interrupt_line_state(VMXNET3State *s, int lidx) 288{ 289 if (!s->interrupt_states[lidx].is_pending && 290 s->interrupt_states[lidx].is_asserted) { 291 VMW_IRPRN("New interrupt line state for index %d is DOWN", lidx); 292 _vmxnet3_deassert_interrupt_line(s, lidx); 293 s->interrupt_states[lidx].is_asserted = false; 294 return; 295 } 296 297 if (s->interrupt_states[lidx].is_pending && 298 !s->interrupt_states[lidx].is_masked && 299 !s->interrupt_states[lidx].is_asserted) { 300 VMW_IRPRN("New interrupt line state for index %d is UP", lidx); 301 s->interrupt_states[lidx].is_asserted = 302 _vmxnet3_assert_interrupt_line(s, lidx); 303 s->interrupt_states[lidx].is_pending = false; 304 return; 305 } 306} 307 308static void vmxnet3_trigger_interrupt(VMXNET3State *s, int lidx) 309{ 310 PCIDevice *d = PCI_DEVICE(s); 311 s->interrupt_states[lidx].is_pending = true; 312 vmxnet3_update_interrupt_line_state(s, lidx); 313 314 if (s->msix_used && msix_enabled(d) && s->auto_int_masking) { 315 goto do_automask; 316 } 317 318 if (msi_enabled(d) && s->auto_int_masking) { 319 goto do_automask; 320 } 321 322 return; 323 324do_automask: 325 s->interrupt_states[lidx].is_masked = true; 326 vmxnet3_update_interrupt_line_state(s, lidx); 327} 328 329static bool vmxnet3_interrupt_asserted(VMXNET3State *s, int lidx) 330{ 331 return s->interrupt_states[lidx].is_asserted; 332} 333 334static void vmxnet3_clear_interrupt(VMXNET3State *s, int int_idx) 335{ 336 s->interrupt_states[int_idx].is_pending = false; 337 if (s->auto_int_masking) { 338 s->interrupt_states[int_idx].is_masked = true; 339 } 340 vmxnet3_update_interrupt_line_state(s, int_idx); 341} 342 343static void 344vmxnet3_on_interrupt_mask_changed(VMXNET3State *s, int lidx, bool is_masked) 345{ 346 s->interrupt_states[lidx].is_masked = is_masked; 347 vmxnet3_update_interrupt_line_state(s, lidx); 348} 349 350static bool vmxnet3_verify_driver_magic(PCIDevice *d, hwaddr dshmem) 351{ 352 return (VMXNET3_READ_DRV_SHARED32(d, dshmem, magic) == VMXNET3_REV1_MAGIC); 353} 354 355#define VMXNET3_GET_BYTE(x, byte_num) (((x) >> (byte_num)*8) & 0xFF) 356#define VMXNET3_MAKE_BYTE(byte_num, val) \ 357 (((uint32_t)((val) & 0xFF)) << (byte_num)*8) 358 359static void vmxnet3_set_variable_mac(VMXNET3State *s, uint32_t h, uint32_t l) 360{ 361 s->conf.macaddr.a[0] = VMXNET3_GET_BYTE(l, 0); 362 s->conf.macaddr.a[1] = VMXNET3_GET_BYTE(l, 1); 363 s->conf.macaddr.a[2] = VMXNET3_GET_BYTE(l, 2); 364 s->conf.macaddr.a[3] = VMXNET3_GET_BYTE(l, 3); 365 s->conf.macaddr.a[4] = VMXNET3_GET_BYTE(h, 0); 366 s->conf.macaddr.a[5] = VMXNET3_GET_BYTE(h, 1); 367 368 VMW_CFPRN("Variable MAC: " MAC_FMT, MAC_ARG(s->conf.macaddr.a)); 369 370 qemu_format_nic_info_str(qemu_get_queue(s->nic), s->conf.macaddr.a); 371} 372 373static uint64_t vmxnet3_get_mac_low(MACAddr *addr) 374{ 375 return VMXNET3_MAKE_BYTE(0, addr->a[0]) | 376 VMXNET3_MAKE_BYTE(1, addr->a[1]) | 377 VMXNET3_MAKE_BYTE(2, addr->a[2]) | 378 VMXNET3_MAKE_BYTE(3, addr->a[3]); 379} 380 381static uint64_t vmxnet3_get_mac_high(MACAddr *addr) 382{ 383 return VMXNET3_MAKE_BYTE(0, addr->a[4]) | 384 VMXNET3_MAKE_BYTE(1, addr->a[5]); 385} 386 387static void 388vmxnet3_inc_tx_consumption_counter(VMXNET3State *s, int qidx) 389{ 390 vmxnet3_ring_inc(&s->txq_descr[qidx].tx_ring); 391} 392 393static inline void 394vmxnet3_inc_rx_consumption_counter(VMXNET3State *s, int qidx, int ridx) 395{ 396 vmxnet3_ring_inc(&s->rxq_descr[qidx].rx_ring[ridx]); 397} 398 399static inline void 400vmxnet3_inc_tx_completion_counter(VMXNET3State *s, int qidx) 401{ 402 vmxnet3_ring_inc(&s->txq_descr[qidx].comp_ring); 403} 404 405static void 406vmxnet3_inc_rx_completion_counter(VMXNET3State *s, int qidx) 407{ 408 vmxnet3_ring_inc(&s->rxq_descr[qidx].comp_ring); 409} 410 411static void 412vmxnet3_dec_rx_completion_counter(VMXNET3State *s, int qidx) 413{ 414 vmxnet3_ring_dec(&s->rxq_descr[qidx].comp_ring); 415} 416 417static void vmxnet3_complete_packet(VMXNET3State *s, int qidx, uint32_t tx_ridx) 418{ 419 struct Vmxnet3_TxCompDesc txcq_descr; 420 PCIDevice *d = PCI_DEVICE(s); 421 422 VMXNET3_RING_DUMP(VMW_RIPRN, "TXC", qidx, &s->txq_descr[qidx].comp_ring); 423 424 memset(&txcq_descr, 0, sizeof(txcq_descr)); 425 txcq_descr.txdIdx = tx_ridx; 426 txcq_descr.gen = vmxnet3_ring_curr_gen(&s->txq_descr[qidx].comp_ring); 427 txcq_descr.val1 = cpu_to_le32(txcq_descr.val1); 428 txcq_descr.val2 = cpu_to_le32(txcq_descr.val2); 429 vmxnet3_ring_write_curr_cell(d, &s->txq_descr[qidx].comp_ring, &txcq_descr); 430 431 /* Flush changes in TX descriptor before changing the counter value */ 432 smp_wmb(); 433 434 vmxnet3_inc_tx_completion_counter(s, qidx); 435 vmxnet3_trigger_interrupt(s, s->txq_descr[qidx].intr_idx); 436} 437 438static bool 439vmxnet3_setup_tx_offloads(VMXNET3State *s) 440{ 441 switch (s->offload_mode) { 442 case VMXNET3_OM_NONE: 443 net_tx_pkt_build_vheader(s->tx_pkt, false, false, 0); 444 break; 445 446 case VMXNET3_OM_CSUM: 447 net_tx_pkt_build_vheader(s->tx_pkt, false, true, 0); 448 VMW_PKPRN("L4 CSO requested\n"); 449 break; 450 451 case VMXNET3_OM_TSO: 452 net_tx_pkt_build_vheader(s->tx_pkt, true, true, 453 s->cso_or_gso_size); 454 net_tx_pkt_update_ip_checksums(s->tx_pkt); 455 VMW_PKPRN("GSO offload requested."); 456 break; 457 458 default: 459 g_assert_not_reached(); 460 return false; 461 } 462 463 return true; 464} 465 466static void 467vmxnet3_tx_retrieve_metadata(VMXNET3State *s, 468 const struct Vmxnet3_TxDesc *txd) 469{ 470 s->offload_mode = txd->om; 471 s->cso_or_gso_size = txd->msscof; 472 s->tci = txd->tci; 473 s->needs_vlan = txd->ti; 474} 475 476typedef enum { 477 VMXNET3_PKT_STATUS_OK, 478 VMXNET3_PKT_STATUS_ERROR, 479 VMXNET3_PKT_STATUS_DISCARD,/* only for tx */ 480 VMXNET3_PKT_STATUS_OUT_OF_BUF /* only for rx */ 481} Vmxnet3PktStatus; 482 483static void 484vmxnet3_on_tx_done_update_stats(VMXNET3State *s, int qidx, 485 Vmxnet3PktStatus status) 486{ 487 size_t tot_len = net_tx_pkt_get_total_len(s->tx_pkt); 488 struct UPT1_TxStats *stats = &s->txq_descr[qidx].txq_stats; 489 490 switch (status) { 491 case VMXNET3_PKT_STATUS_OK: 492 switch (net_tx_pkt_get_packet_type(s->tx_pkt)) { 493 case ETH_PKT_BCAST: 494 stats->bcastPktsTxOK++; 495 stats->bcastBytesTxOK += tot_len; 496 break; 497 case ETH_PKT_MCAST: 498 stats->mcastPktsTxOK++; 499 stats->mcastBytesTxOK += tot_len; 500 break; 501 case ETH_PKT_UCAST: 502 stats->ucastPktsTxOK++; 503 stats->ucastBytesTxOK += tot_len; 504 break; 505 default: 506 g_assert_not_reached(); 507 } 508 509 if (s->offload_mode == VMXNET3_OM_TSO) { 510 /* 511 * According to VMWARE headers this statistic is a number 512 * of packets after segmentation but since we don't have 513 * this information in QEMU model, the best we can do is to 514 * provide number of non-segmented packets 515 */ 516 stats->TSOPktsTxOK++; 517 stats->TSOBytesTxOK += tot_len; 518 } 519 break; 520 521 case VMXNET3_PKT_STATUS_DISCARD: 522 stats->pktsTxDiscard++; 523 break; 524 525 case VMXNET3_PKT_STATUS_ERROR: 526 stats->pktsTxError++; 527 break; 528 529 default: 530 g_assert_not_reached(); 531 } 532} 533 534static void 535vmxnet3_on_rx_done_update_stats(VMXNET3State *s, 536 int qidx, 537 Vmxnet3PktStatus status) 538{ 539 struct UPT1_RxStats *stats = &s->rxq_descr[qidx].rxq_stats; 540 size_t tot_len = net_rx_pkt_get_total_len(s->rx_pkt); 541 542 switch (status) { 543 case VMXNET3_PKT_STATUS_OUT_OF_BUF: 544 stats->pktsRxOutOfBuf++; 545 break; 546 547 case VMXNET3_PKT_STATUS_ERROR: 548 stats->pktsRxError++; 549 break; 550 case VMXNET3_PKT_STATUS_OK: 551 switch (net_rx_pkt_get_packet_type(s->rx_pkt)) { 552 case ETH_PKT_BCAST: 553 stats->bcastPktsRxOK++; 554 stats->bcastBytesRxOK += tot_len; 555 break; 556 case ETH_PKT_MCAST: 557 stats->mcastPktsRxOK++; 558 stats->mcastBytesRxOK += tot_len; 559 break; 560 case ETH_PKT_UCAST: 561 stats->ucastPktsRxOK++; 562 stats->ucastBytesRxOK += tot_len; 563 break; 564 default: 565 g_assert_not_reached(); 566 } 567 568 if (tot_len > s->mtu) { 569 stats->LROPktsRxOK++; 570 stats->LROBytesRxOK += tot_len; 571 } 572 break; 573 default: 574 g_assert_not_reached(); 575 } 576} 577 578static inline void 579vmxnet3_ring_read_curr_txdesc(PCIDevice *pcidev, Vmxnet3Ring *ring, 580 struct Vmxnet3_TxDesc *txd) 581{ 582 vmxnet3_ring_read_curr_cell(pcidev, ring, txd); 583 txd->addr = le64_to_cpu(txd->addr); 584 txd->val1 = le32_to_cpu(txd->val1); 585 txd->val2 = le32_to_cpu(txd->val2); 586} 587 588static inline bool 589vmxnet3_pop_next_tx_descr(VMXNET3State *s, 590 int qidx, 591 struct Vmxnet3_TxDesc *txd, 592 uint32_t *descr_idx) 593{ 594 Vmxnet3Ring *ring = &s->txq_descr[qidx].tx_ring; 595 PCIDevice *d = PCI_DEVICE(s); 596 597 vmxnet3_ring_read_curr_txdesc(d, ring, txd); 598 if (txd->gen == vmxnet3_ring_curr_gen(ring)) { 599 /* Only read after generation field verification */ 600 smp_rmb(); 601 /* Re-read to be sure we got the latest version */ 602 vmxnet3_ring_read_curr_txdesc(d, ring, txd); 603 VMXNET3_RING_DUMP(VMW_RIPRN, "TX", qidx, ring); 604 *descr_idx = vmxnet3_ring_curr_cell_idx(ring); 605 vmxnet3_inc_tx_consumption_counter(s, qidx); 606 return true; 607 } 608 609 return false; 610} 611 612static bool 613vmxnet3_send_packet(VMXNET3State *s, uint32_t qidx) 614{ 615 Vmxnet3PktStatus status = VMXNET3_PKT_STATUS_OK; 616 617 if (!vmxnet3_setup_tx_offloads(s)) { 618 status = VMXNET3_PKT_STATUS_ERROR; 619 goto func_exit; 620 } 621 622 /* debug prints */ 623 vmxnet3_dump_virt_hdr(net_tx_pkt_get_vhdr(s->tx_pkt)); 624 net_tx_pkt_dump(s->tx_pkt); 625 626 if (!net_tx_pkt_send(s->tx_pkt, qemu_get_queue(s->nic))) { 627 status = VMXNET3_PKT_STATUS_DISCARD; 628 goto func_exit; 629 } 630 631func_exit: 632 vmxnet3_on_tx_done_update_stats(s, qidx, status); 633 return (status == VMXNET3_PKT_STATUS_OK); 634} 635 636static void vmxnet3_process_tx_queue(VMXNET3State *s, int qidx) 637{ 638 struct Vmxnet3_TxDesc txd; 639 uint32_t txd_idx; 640 uint32_t data_len; 641 hwaddr data_pa; 642 643 for (;;) { 644 if (!vmxnet3_pop_next_tx_descr(s, qidx, &txd, &txd_idx)) { 645 break; 646 } 647 648 vmxnet3_dump_tx_descr(&txd); 649 650 if (!s->skip_current_tx_pkt) { 651 data_len = (txd.len > 0) ? txd.len : VMXNET3_MAX_TX_BUF_SIZE; 652 data_pa = txd.addr; 653 654 if (!net_tx_pkt_add_raw_fragment(s->tx_pkt, 655 data_pa, 656 data_len)) { 657 s->skip_current_tx_pkt = true; 658 } 659 } 660 661 if (s->tx_sop) { 662 vmxnet3_tx_retrieve_metadata(s, &txd); 663 s->tx_sop = false; 664 } 665 666 if (txd.eop) { 667 if (!s->skip_current_tx_pkt && net_tx_pkt_parse(s->tx_pkt)) { 668 if (s->needs_vlan) { 669 net_tx_pkt_setup_vlan_header(s->tx_pkt, s->tci); 670 } 671 672 vmxnet3_send_packet(s, qidx); 673 } else { 674 vmxnet3_on_tx_done_update_stats(s, qidx, 675 VMXNET3_PKT_STATUS_ERROR); 676 } 677 678 vmxnet3_complete_packet(s, qidx, txd_idx); 679 s->tx_sop = true; 680 s->skip_current_tx_pkt = false; 681 net_tx_pkt_reset(s->tx_pkt); 682 } 683 } 684} 685 686static inline void 687vmxnet3_read_next_rx_descr(VMXNET3State *s, int qidx, int ridx, 688 struct Vmxnet3_RxDesc *dbuf, uint32_t *didx) 689{ 690 PCIDevice *d = PCI_DEVICE(s); 691 692 Vmxnet3Ring *ring = &s->rxq_descr[qidx].rx_ring[ridx]; 693 *didx = vmxnet3_ring_curr_cell_idx(ring); 694 vmxnet3_ring_read_curr_cell(d, ring, dbuf); 695 dbuf->addr = le64_to_cpu(dbuf->addr); 696 dbuf->val1 = le32_to_cpu(dbuf->val1); 697 dbuf->ext1 = le32_to_cpu(dbuf->ext1); 698} 699 700static inline uint8_t 701vmxnet3_get_rx_ring_gen(VMXNET3State *s, int qidx, int ridx) 702{ 703 return s->rxq_descr[qidx].rx_ring[ridx].gen; 704} 705 706static inline hwaddr 707vmxnet3_pop_rxc_descr(VMXNET3State *s, int qidx, uint32_t *descr_gen) 708{ 709 uint8_t ring_gen; 710 struct Vmxnet3_RxCompDesc rxcd; 711 712 hwaddr daddr = 713 vmxnet3_ring_curr_cell_pa(&s->rxq_descr[qidx].comp_ring); 714 715 pci_dma_read(PCI_DEVICE(s), 716 daddr, &rxcd, sizeof(struct Vmxnet3_RxCompDesc)); 717 rxcd.val1 = le32_to_cpu(rxcd.val1); 718 rxcd.val2 = le32_to_cpu(rxcd.val2); 719 rxcd.val3 = le32_to_cpu(rxcd.val3); 720 ring_gen = vmxnet3_ring_curr_gen(&s->rxq_descr[qidx].comp_ring); 721 722 if (rxcd.gen != ring_gen) { 723 *descr_gen = ring_gen; 724 vmxnet3_inc_rx_completion_counter(s, qidx); 725 return daddr; 726 } 727 728 return 0; 729} 730 731static inline void 732vmxnet3_revert_rxc_descr(VMXNET3State *s, int qidx) 733{ 734 vmxnet3_dec_rx_completion_counter(s, qidx); 735} 736 737#define RXQ_IDX (0) 738#define RX_HEAD_BODY_RING (0) 739#define RX_BODY_ONLY_RING (1) 740 741static bool 742vmxnet3_get_next_head_rx_descr(VMXNET3State *s, 743 struct Vmxnet3_RxDesc *descr_buf, 744 uint32_t *descr_idx, 745 uint32_t *ridx) 746{ 747 for (;;) { 748 uint32_t ring_gen; 749 vmxnet3_read_next_rx_descr(s, RXQ_IDX, RX_HEAD_BODY_RING, 750 descr_buf, descr_idx); 751 752 /* If no more free descriptors - return */ 753 ring_gen = vmxnet3_get_rx_ring_gen(s, RXQ_IDX, RX_HEAD_BODY_RING); 754 if (descr_buf->gen != ring_gen) { 755 return false; 756 } 757 758 /* Only read after generation field verification */ 759 smp_rmb(); 760 /* Re-read to be sure we got the latest version */ 761 vmxnet3_read_next_rx_descr(s, RXQ_IDX, RX_HEAD_BODY_RING, 762 descr_buf, descr_idx); 763 764 /* Mark current descriptor as used/skipped */ 765 vmxnet3_inc_rx_consumption_counter(s, RXQ_IDX, RX_HEAD_BODY_RING); 766 767 /* If this is what we are looking for - return */ 768 if (descr_buf->btype == VMXNET3_RXD_BTYPE_HEAD) { 769 *ridx = RX_HEAD_BODY_RING; 770 return true; 771 } 772 } 773} 774 775static bool 776vmxnet3_get_next_body_rx_descr(VMXNET3State *s, 777 struct Vmxnet3_RxDesc *d, 778 uint32_t *didx, 779 uint32_t *ridx) 780{ 781 vmxnet3_read_next_rx_descr(s, RXQ_IDX, RX_HEAD_BODY_RING, d, didx); 782 783 /* Try to find corresponding descriptor in head/body ring */ 784 if (d->gen == vmxnet3_get_rx_ring_gen(s, RXQ_IDX, RX_HEAD_BODY_RING)) { 785 /* Only read after generation field verification */ 786 smp_rmb(); 787 /* Re-read to be sure we got the latest version */ 788 vmxnet3_read_next_rx_descr(s, RXQ_IDX, RX_HEAD_BODY_RING, d, didx); 789 if (d->btype == VMXNET3_RXD_BTYPE_BODY) { 790 vmxnet3_inc_rx_consumption_counter(s, RXQ_IDX, RX_HEAD_BODY_RING); 791 *ridx = RX_HEAD_BODY_RING; 792 return true; 793 } 794 } 795 796 /* 797 * If there is no free descriptors on head/body ring or next free 798 * descriptor is a head descriptor switch to body only ring 799 */ 800 vmxnet3_read_next_rx_descr(s, RXQ_IDX, RX_BODY_ONLY_RING, d, didx); 801 802 /* If no more free descriptors - return */ 803 if (d->gen == vmxnet3_get_rx_ring_gen(s, RXQ_IDX, RX_BODY_ONLY_RING)) { 804 /* Only read after generation field verification */ 805 smp_rmb(); 806 /* Re-read to be sure we got the latest version */ 807 vmxnet3_read_next_rx_descr(s, RXQ_IDX, RX_BODY_ONLY_RING, d, didx); 808 assert(d->btype == VMXNET3_RXD_BTYPE_BODY); 809 *ridx = RX_BODY_ONLY_RING; 810 vmxnet3_inc_rx_consumption_counter(s, RXQ_IDX, RX_BODY_ONLY_RING); 811 return true; 812 } 813 814 return false; 815} 816 817static inline bool 818vmxnet3_get_next_rx_descr(VMXNET3State *s, bool is_head, 819 struct Vmxnet3_RxDesc *descr_buf, 820 uint32_t *descr_idx, 821 uint32_t *ridx) 822{ 823 if (is_head || !s->rx_packets_compound) { 824 return vmxnet3_get_next_head_rx_descr(s, descr_buf, descr_idx, ridx); 825 } else { 826 return vmxnet3_get_next_body_rx_descr(s, descr_buf, descr_idx, ridx); 827 } 828} 829 830/* In case packet was csum offloaded (either NEEDS_CSUM or DATA_VALID), 831 * the implementation always passes an RxCompDesc with a "Checksum 832 * calculated and found correct" to the OS (cnc=0 and tuc=1, see 833 * vmxnet3_rx_update_descr). This emulates the observed ESXi behavior. 834 * 835 * Therefore, if packet has the NEEDS_CSUM set, we must calculate 836 * and place a fully computed checksum into the tcp/udp header. 837 * Otherwise, the OS driver will receive a checksum-correct indication 838 * (CHECKSUM_UNNECESSARY), but with the actual tcp/udp checksum field 839 * having just the pseudo header csum value. 840 * 841 * While this is not a problem if packet is destined for local delivery, 842 * in the case the host OS performs forwarding, it will forward an 843 * incorrectly checksummed packet. 844 */ 845static void vmxnet3_rx_need_csum_calculate(struct NetRxPkt *pkt, 846 const void *pkt_data, 847 size_t pkt_len) 848{ 849 struct virtio_net_hdr *vhdr; 850 bool isip4, isip6, istcp, isudp; 851 uint8_t *data; 852 int len; 853 854 if (!net_rx_pkt_has_virt_hdr(pkt)) { 855 return; 856 } 857 858 vhdr = net_rx_pkt_get_vhdr(pkt); 859 if (!VMXNET_FLAG_IS_SET(vhdr->flags, VIRTIO_NET_HDR_F_NEEDS_CSUM)) { 860 return; 861 } 862 863 net_rx_pkt_get_protocols(pkt, &isip4, &isip6, &isudp, &istcp); 864 if (!(isip4 || isip6) || !(istcp || isudp)) { 865 return; 866 } 867 868 vmxnet3_dump_virt_hdr(vhdr); 869 870 /* Validate packet len: csum_start + scum_offset + length of csum field */ 871 if (pkt_len < (vhdr->csum_start + vhdr->csum_offset + 2)) { 872 VMW_PKPRN("packet len:%zu < csum_start(%d) + csum_offset(%d) + 2, " 873 "cannot calculate checksum", 874 pkt_len, vhdr->csum_start, vhdr->csum_offset); 875 return; 876 } 877 878 data = (uint8_t *)pkt_data + vhdr->csum_start; 879 len = pkt_len - vhdr->csum_start; 880 /* Put the checksum obtained into the packet */ 881 stw_be_p(data + vhdr->csum_offset, 882 net_checksum_finish_nozero(net_checksum_add(len, data))); 883 884 vhdr->flags &= ~VIRTIO_NET_HDR_F_NEEDS_CSUM; 885 vhdr->flags |= VIRTIO_NET_HDR_F_DATA_VALID; 886} 887 888static void vmxnet3_rx_update_descr(struct NetRxPkt *pkt, 889 struct Vmxnet3_RxCompDesc *rxcd) 890{ 891 int csum_ok, is_gso; 892 bool isip4, isip6, istcp, isudp; 893 struct virtio_net_hdr *vhdr; 894 uint8_t offload_type; 895 896 if (net_rx_pkt_is_vlan_stripped(pkt)) { 897 rxcd->ts = 1; 898 rxcd->tci = net_rx_pkt_get_vlan_tag(pkt); 899 } 900 901 if (!net_rx_pkt_has_virt_hdr(pkt)) { 902 goto nocsum; 903 } 904 905 vhdr = net_rx_pkt_get_vhdr(pkt); 906 /* 907 * Checksum is valid when lower level tell so or when lower level 908 * requires checksum offload telling that packet produced/bridged 909 * locally and did travel over network after last checksum calculation 910 * or production 911 */ 912 csum_ok = VMXNET_FLAG_IS_SET(vhdr->flags, VIRTIO_NET_HDR_F_DATA_VALID) || 913 VMXNET_FLAG_IS_SET(vhdr->flags, VIRTIO_NET_HDR_F_NEEDS_CSUM); 914 915 offload_type = vhdr->gso_type & ~VIRTIO_NET_HDR_GSO_ECN; 916 is_gso = (offload_type != VIRTIO_NET_HDR_GSO_NONE) ? 1 : 0; 917 918 if (!csum_ok && !is_gso) { 919 goto nocsum; 920 } 921 922 net_rx_pkt_get_protocols(pkt, &isip4, &isip6, &isudp, &istcp); 923 if ((!istcp && !isudp) || (!isip4 && !isip6)) { 924 goto nocsum; 925 } 926 927 rxcd->cnc = 0; 928 rxcd->v4 = isip4 ? 1 : 0; 929 rxcd->v6 = isip6 ? 1 : 0; 930 rxcd->tcp = istcp ? 1 : 0; 931 rxcd->udp = isudp ? 1 : 0; 932 rxcd->fcs = rxcd->tuc = rxcd->ipc = 1; 933 return; 934 935nocsum: 936 rxcd->cnc = 1; 937 return; 938} 939 940static void 941vmxnet3_pci_dma_writev(PCIDevice *pci_dev, 942 const struct iovec *iov, 943 size_t start_iov_off, 944 hwaddr target_addr, 945 size_t bytes_to_copy) 946{ 947 size_t curr_off = 0; 948 size_t copied = 0; 949 950 while (bytes_to_copy) { 951 if (start_iov_off < (curr_off + iov->iov_len)) { 952 size_t chunk_len = 953 MIN((curr_off + iov->iov_len) - start_iov_off, bytes_to_copy); 954 955 pci_dma_write(pci_dev, target_addr + copied, 956 iov->iov_base + start_iov_off - curr_off, 957 chunk_len); 958 959 copied += chunk_len; 960 start_iov_off += chunk_len; 961 curr_off = start_iov_off; 962 bytes_to_copy -= chunk_len; 963 } else { 964 curr_off += iov->iov_len; 965 } 966 iov++; 967 } 968} 969 970static void 971vmxnet3_pci_dma_write_rxcd(PCIDevice *pcidev, dma_addr_t pa, 972 struct Vmxnet3_RxCompDesc *rxcd) 973{ 974 rxcd->val1 = cpu_to_le32(rxcd->val1); 975 rxcd->val2 = cpu_to_le32(rxcd->val2); 976 rxcd->val3 = cpu_to_le32(rxcd->val3); 977 pci_dma_write(pcidev, pa, rxcd, sizeof(*rxcd)); 978} 979 980static bool 981vmxnet3_indicate_packet(VMXNET3State *s) 982{ 983 struct Vmxnet3_RxDesc rxd; 984 PCIDevice *d = PCI_DEVICE(s); 985 bool is_head = true; 986 uint32_t rxd_idx; 987 uint32_t rx_ridx = 0; 988 989 struct Vmxnet3_RxCompDesc rxcd; 990 uint32_t new_rxcd_gen = VMXNET3_INIT_GEN; 991 hwaddr new_rxcd_pa = 0; 992 hwaddr ready_rxcd_pa = 0; 993 struct iovec *data = net_rx_pkt_get_iovec(s->rx_pkt); 994 size_t bytes_copied = 0; 995 size_t bytes_left = net_rx_pkt_get_total_len(s->rx_pkt); 996 uint16_t num_frags = 0; 997 size_t chunk_size; 998 999 net_rx_pkt_dump(s->rx_pkt); 1000 1001 while (bytes_left > 0) { 1002 1003 /* cannot add more frags to packet */ 1004 if (num_frags == s->max_rx_frags) { 1005 break; 1006 } 1007 1008 new_rxcd_pa = vmxnet3_pop_rxc_descr(s, RXQ_IDX, &new_rxcd_gen); 1009 if (!new_rxcd_pa) { 1010 break; 1011 } 1012 1013 if (!vmxnet3_get_next_rx_descr(s, is_head, &rxd, &rxd_idx, &rx_ridx)) { 1014 break; 1015 } 1016 1017 chunk_size = MIN(bytes_left, rxd.len); 1018 vmxnet3_pci_dma_writev(d, data, bytes_copied, rxd.addr, chunk_size); 1019 bytes_copied += chunk_size; 1020 bytes_left -= chunk_size; 1021 1022 vmxnet3_dump_rx_descr(&rxd); 1023 1024 if (ready_rxcd_pa != 0) { 1025 vmxnet3_pci_dma_write_rxcd(d, ready_rxcd_pa, &rxcd); 1026 } 1027 1028 memset(&rxcd, 0, sizeof(struct Vmxnet3_RxCompDesc)); 1029 rxcd.rxdIdx = rxd_idx; 1030 rxcd.len = chunk_size; 1031 rxcd.sop = is_head; 1032 rxcd.gen = new_rxcd_gen; 1033 rxcd.rqID = RXQ_IDX + rx_ridx * s->rxq_num; 1034 1035 if (bytes_left == 0) { 1036 vmxnet3_rx_update_descr(s->rx_pkt, &rxcd); 1037 } 1038 1039 VMW_RIPRN("RX Completion descriptor: rxRing: %lu rxIdx %lu len %lu " 1040 "sop %d csum_correct %lu", 1041 (unsigned long) rx_ridx, 1042 (unsigned long) rxcd.rxdIdx, 1043 (unsigned long) rxcd.len, 1044 (int) rxcd.sop, 1045 (unsigned long) rxcd.tuc); 1046 1047 is_head = false; 1048 ready_rxcd_pa = new_rxcd_pa; 1049 new_rxcd_pa = 0; 1050 num_frags++; 1051 } 1052 1053 if (ready_rxcd_pa != 0) { 1054 rxcd.eop = 1; 1055 rxcd.err = (bytes_left != 0); 1056 1057 vmxnet3_pci_dma_write_rxcd(d, ready_rxcd_pa, &rxcd); 1058 1059 /* Flush RX descriptor changes */ 1060 smp_wmb(); 1061 } 1062 1063 if (new_rxcd_pa != 0) { 1064 vmxnet3_revert_rxc_descr(s, RXQ_IDX); 1065 } 1066 1067 vmxnet3_trigger_interrupt(s, s->rxq_descr[RXQ_IDX].intr_idx); 1068 1069 if (bytes_left == 0) { 1070 vmxnet3_on_rx_done_update_stats(s, RXQ_IDX, VMXNET3_PKT_STATUS_OK); 1071 return true; 1072 } else if (num_frags == s->max_rx_frags) { 1073 vmxnet3_on_rx_done_update_stats(s, RXQ_IDX, VMXNET3_PKT_STATUS_ERROR); 1074 return false; 1075 } else { 1076 vmxnet3_on_rx_done_update_stats(s, RXQ_IDX, 1077 VMXNET3_PKT_STATUS_OUT_OF_BUF); 1078 return false; 1079 } 1080} 1081 1082static void 1083vmxnet3_io_bar0_write(void *opaque, hwaddr addr, 1084 uint64_t val, unsigned size) 1085{ 1086 VMXNET3State *s = opaque; 1087 1088 if (!s->device_active) { 1089 return; 1090 } 1091 1092 if (VMW_IS_MULTIREG_ADDR(addr, VMXNET3_REG_TXPROD, 1093 VMXNET3_DEVICE_MAX_TX_QUEUES, VMXNET3_REG_ALIGN)) { 1094 int tx_queue_idx = 1095 VMW_MULTIREG_IDX_BY_ADDR(addr, VMXNET3_REG_TXPROD, 1096 VMXNET3_REG_ALIGN); 1097 if (tx_queue_idx <= s->txq_num) { 1098 vmxnet3_process_tx_queue(s, tx_queue_idx); 1099 } else { 1100 qemu_log_mask(LOG_GUEST_ERROR, "vmxnet3: Illegal TX queue %d/%d\n", 1101 tx_queue_idx, s->txq_num); 1102 } 1103 return; 1104 } 1105 1106 if (VMW_IS_MULTIREG_ADDR(addr, VMXNET3_REG_IMR, 1107 VMXNET3_MAX_INTRS, VMXNET3_REG_ALIGN)) { 1108 int l = VMW_MULTIREG_IDX_BY_ADDR(addr, VMXNET3_REG_IMR, 1109 VMXNET3_REG_ALIGN); 1110 1111 VMW_CBPRN("Interrupt mask for line %d written: 0x%" PRIx64, l, val); 1112 1113 vmxnet3_on_interrupt_mask_changed(s, l, val); 1114 return; 1115 } 1116 1117 if (VMW_IS_MULTIREG_ADDR(addr, VMXNET3_REG_RXPROD, 1118 VMXNET3_DEVICE_MAX_RX_QUEUES, VMXNET3_REG_ALIGN) || 1119 VMW_IS_MULTIREG_ADDR(addr, VMXNET3_REG_RXPROD2, 1120 VMXNET3_DEVICE_MAX_RX_QUEUES, VMXNET3_REG_ALIGN)) { 1121 return; 1122 } 1123 1124 VMW_WRPRN("BAR0 unknown write [%" PRIx64 "] = %" PRIx64 ", size %d", 1125 (uint64_t) addr, val, size); 1126} 1127 1128static uint64_t 1129vmxnet3_io_bar0_read(void *opaque, hwaddr addr, unsigned size) 1130{ 1131 VMXNET3State *s = opaque; 1132 1133 if (VMW_IS_MULTIREG_ADDR(addr, VMXNET3_REG_IMR, 1134 VMXNET3_MAX_INTRS, VMXNET3_REG_ALIGN)) { 1135 int l = VMW_MULTIREG_IDX_BY_ADDR(addr, VMXNET3_REG_IMR, 1136 VMXNET3_REG_ALIGN); 1137 return s->interrupt_states[l].is_masked; 1138 } 1139 1140 VMW_CBPRN("BAR0 unknown read [%" PRIx64 "], size %d", addr, size); 1141 return 0; 1142} 1143 1144static void vmxnet3_reset_interrupt_states(VMXNET3State *s) 1145{ 1146 int i; 1147 for (i = 0; i < ARRAY_SIZE(s->interrupt_states); i++) { 1148 s->interrupt_states[i].is_asserted = false; 1149 s->interrupt_states[i].is_pending = false; 1150 s->interrupt_states[i].is_masked = true; 1151 } 1152} 1153 1154static void vmxnet3_reset_mac(VMXNET3State *s) 1155{ 1156 memcpy(&s->conf.macaddr.a, &s->perm_mac.a, sizeof(s->perm_mac.a)); 1157 VMW_CFPRN("MAC address set to: " MAC_FMT, MAC_ARG(s->conf.macaddr.a)); 1158} 1159 1160static void vmxnet3_deactivate_device(VMXNET3State *s) 1161{ 1162 if (s->device_active) { 1163 VMW_CBPRN("Deactivating vmxnet3..."); 1164 net_tx_pkt_reset(s->tx_pkt); 1165 net_tx_pkt_uninit(s->tx_pkt); 1166 net_rx_pkt_uninit(s->rx_pkt); 1167 s->device_active = false; 1168 } 1169} 1170 1171static void vmxnet3_reset(VMXNET3State *s) 1172{ 1173 VMW_CBPRN("Resetting vmxnet3..."); 1174 1175 vmxnet3_deactivate_device(s); 1176 vmxnet3_reset_interrupt_states(s); 1177 s->drv_shmem = 0; 1178 s->tx_sop = true; 1179 s->skip_current_tx_pkt = false; 1180} 1181 1182static void vmxnet3_update_rx_mode(VMXNET3State *s) 1183{ 1184 PCIDevice *d = PCI_DEVICE(s); 1185 1186 s->rx_mode = VMXNET3_READ_DRV_SHARED32(d, s->drv_shmem, 1187 devRead.rxFilterConf.rxMode); 1188 VMW_CFPRN("RX mode: 0x%08X", s->rx_mode); 1189} 1190 1191static void vmxnet3_update_vlan_filters(VMXNET3State *s) 1192{ 1193 int i; 1194 PCIDevice *d = PCI_DEVICE(s); 1195 1196 /* Copy configuration from shared memory */ 1197 VMXNET3_READ_DRV_SHARED(d, s->drv_shmem, 1198 devRead.rxFilterConf.vfTable, 1199 s->vlan_table, 1200 sizeof(s->vlan_table)); 1201 1202 /* Invert byte order when needed */ 1203 for (i = 0; i < ARRAY_SIZE(s->vlan_table); i++) { 1204 s->vlan_table[i] = le32_to_cpu(s->vlan_table[i]); 1205 } 1206 1207 /* Dump configuration for debugging purposes */ 1208 VMW_CFPRN("Configured VLANs:"); 1209 for (i = 0; i < sizeof(s->vlan_table) * 8; i++) { 1210 if (VMXNET3_VFTABLE_ENTRY_IS_SET(s->vlan_table, i)) { 1211 VMW_CFPRN("\tVLAN %d is present", i); 1212 } 1213 } 1214} 1215 1216static void vmxnet3_update_mcast_filters(VMXNET3State *s) 1217{ 1218 PCIDevice *d = PCI_DEVICE(s); 1219 1220 uint16_t list_bytes = 1221 VMXNET3_READ_DRV_SHARED16(d, s->drv_shmem, 1222 devRead.rxFilterConf.mfTableLen); 1223 1224 s->mcast_list_len = list_bytes / sizeof(s->mcast_list[0]); 1225 1226 s->mcast_list = g_realloc(s->mcast_list, list_bytes); 1227 if (!s->mcast_list) { 1228 if (s->mcast_list_len == 0) { 1229 VMW_CFPRN("Current multicast list is empty"); 1230 } else { 1231 VMW_ERPRN("Failed to allocate multicast list of %d elements", 1232 s->mcast_list_len); 1233 } 1234 s->mcast_list_len = 0; 1235 } else { 1236 int i; 1237 hwaddr mcast_list_pa = 1238 VMXNET3_READ_DRV_SHARED64(d, s->drv_shmem, 1239 devRead.rxFilterConf.mfTablePA); 1240 1241 pci_dma_read(d, mcast_list_pa, s->mcast_list, list_bytes); 1242 1243 VMW_CFPRN("Current multicast list len is %d:", s->mcast_list_len); 1244 for (i = 0; i < s->mcast_list_len; i++) { 1245 VMW_CFPRN("\t" MAC_FMT, MAC_ARG(s->mcast_list[i].a)); 1246 } 1247 } 1248} 1249 1250static void vmxnet3_setup_rx_filtering(VMXNET3State *s) 1251{ 1252 vmxnet3_update_rx_mode(s); 1253 vmxnet3_update_vlan_filters(s); 1254 vmxnet3_update_mcast_filters(s); 1255} 1256 1257static uint32_t vmxnet3_get_interrupt_config(VMXNET3State *s) 1258{ 1259 uint32_t interrupt_mode = VMXNET3_IT_AUTO | (VMXNET3_IMM_AUTO << 2); 1260 VMW_CFPRN("Interrupt config is 0x%X", interrupt_mode); 1261 return interrupt_mode; 1262} 1263 1264static void vmxnet3_fill_stats(VMXNET3State *s) 1265{ 1266 int i; 1267 PCIDevice *d = PCI_DEVICE(s); 1268 1269 if (!s->device_active) 1270 return; 1271 1272 for (i = 0; i < s->txq_num; i++) { 1273 pci_dma_write(d, 1274 s->txq_descr[i].tx_stats_pa, 1275 &s->txq_descr[i].txq_stats, 1276 sizeof(s->txq_descr[i].txq_stats)); 1277 } 1278 1279 for (i = 0; i < s->rxq_num; i++) { 1280 pci_dma_write(d, 1281 s->rxq_descr[i].rx_stats_pa, 1282 &s->rxq_descr[i].rxq_stats, 1283 sizeof(s->rxq_descr[i].rxq_stats)); 1284 } 1285} 1286 1287static void vmxnet3_adjust_by_guest_type(VMXNET3State *s) 1288{ 1289 struct Vmxnet3_GOSInfo gos; 1290 PCIDevice *d = PCI_DEVICE(s); 1291 1292 VMXNET3_READ_DRV_SHARED(d, s->drv_shmem, devRead.misc.driverInfo.gos, 1293 &gos, sizeof(gos)); 1294 s->rx_packets_compound = 1295 (gos.gosType == VMXNET3_GOS_TYPE_WIN) ? false : true; 1296 1297 VMW_CFPRN("Guest type specifics: RXCOMPOUND: %d", s->rx_packets_compound); 1298} 1299 1300static void 1301vmxnet3_dump_conf_descr(const char *name, 1302 struct Vmxnet3_VariableLenConfDesc *pm_descr) 1303{ 1304 VMW_CFPRN("%s descriptor dump: Version %u, Length %u", 1305 name, pm_descr->confVer, pm_descr->confLen); 1306 1307}; 1308 1309static void vmxnet3_update_pm_state(VMXNET3State *s) 1310{ 1311 struct Vmxnet3_VariableLenConfDesc pm_descr; 1312 PCIDevice *d = PCI_DEVICE(s); 1313 1314 pm_descr.confLen = 1315 VMXNET3_READ_DRV_SHARED32(d, s->drv_shmem, devRead.pmConfDesc.confLen); 1316 pm_descr.confVer = 1317 VMXNET3_READ_DRV_SHARED32(d, s->drv_shmem, devRead.pmConfDesc.confVer); 1318 pm_descr.confPA = 1319 VMXNET3_READ_DRV_SHARED64(d, s->drv_shmem, devRead.pmConfDesc.confPA); 1320 1321 vmxnet3_dump_conf_descr("PM State", &pm_descr); 1322} 1323 1324static void vmxnet3_update_features(VMXNET3State *s) 1325{ 1326 uint32_t guest_features; 1327 int rxcso_supported; 1328 PCIDevice *d = PCI_DEVICE(s); 1329 1330 guest_features = VMXNET3_READ_DRV_SHARED32(d, s->drv_shmem, 1331 devRead.misc.uptFeatures); 1332 1333 rxcso_supported = VMXNET_FLAG_IS_SET(guest_features, UPT1_F_RXCSUM); 1334 s->rx_vlan_stripping = VMXNET_FLAG_IS_SET(guest_features, UPT1_F_RXVLAN); 1335 s->lro_supported = VMXNET_FLAG_IS_SET(guest_features, UPT1_F_LRO); 1336 1337 VMW_CFPRN("Features configuration: LRO: %d, RXCSUM: %d, VLANSTRIP: %d", 1338 s->lro_supported, rxcso_supported, 1339 s->rx_vlan_stripping); 1340 if (s->peer_has_vhdr) { 1341 qemu_set_offload(qemu_get_queue(s->nic)->peer, 1342 rxcso_supported, 1343 s->lro_supported, 1344 s->lro_supported, 1345 0, 1346 0); 1347 } 1348} 1349 1350static bool vmxnet3_verify_intx(VMXNET3State *s, int intx) 1351{ 1352 return s->msix_used || msi_enabled(PCI_DEVICE(s)) 1353 || intx == pci_get_byte(s->parent_obj.config + PCI_INTERRUPT_PIN) - 1; 1354} 1355 1356static void vmxnet3_validate_interrupt_idx(bool is_msix, int idx) 1357{ 1358 int max_ints = is_msix ? VMXNET3_MAX_INTRS : VMXNET3_MAX_NMSIX_INTRS; 1359 if (idx >= max_ints) { 1360 hw_error("Bad interrupt index: %d\n", idx); 1361 } 1362} 1363 1364static void vmxnet3_validate_interrupts(VMXNET3State *s) 1365{ 1366 int i; 1367 1368 VMW_CFPRN("Verifying event interrupt index (%d)", s->event_int_idx); 1369 vmxnet3_validate_interrupt_idx(s->msix_used, s->event_int_idx); 1370 1371 for (i = 0; i < s->txq_num; i++) { 1372 int idx = s->txq_descr[i].intr_idx; 1373 VMW_CFPRN("Verifying TX queue %d interrupt index (%d)", i, idx); 1374 vmxnet3_validate_interrupt_idx(s->msix_used, idx); 1375 } 1376 1377 for (i = 0; i < s->rxq_num; i++) { 1378 int idx = s->rxq_descr[i].intr_idx; 1379 VMW_CFPRN("Verifying RX queue %d interrupt index (%d)", i, idx); 1380 vmxnet3_validate_interrupt_idx(s->msix_used, idx); 1381 } 1382} 1383 1384static bool vmxnet3_validate_queues(VMXNET3State *s) 1385{ 1386 /* 1387 * txq_num and rxq_num are total number of queues 1388 * configured by guest. These numbers must not 1389 * exceed corresponding maximal values. 1390 */ 1391 1392 if (s->txq_num > VMXNET3_DEVICE_MAX_TX_QUEUES) { 1393 qemu_log_mask(LOG_GUEST_ERROR, "vmxnet3: Bad TX queues number: %d\n", 1394 s->txq_num); 1395 return false; 1396 } 1397 1398 if (s->rxq_num > VMXNET3_DEVICE_MAX_RX_QUEUES) { 1399 qemu_log_mask(LOG_GUEST_ERROR, "vmxnet3: Bad RX queues number: %d\n", 1400 s->rxq_num); 1401 return false; 1402 } 1403 1404 return true; 1405} 1406 1407static void vmxnet3_activate_device(VMXNET3State *s) 1408{ 1409 int i; 1410 static const uint32_t VMXNET3_DEF_TX_THRESHOLD = 1; 1411 PCIDevice *d = PCI_DEVICE(s); 1412 hwaddr qdescr_table_pa; 1413 uint64_t pa; 1414 uint32_t size; 1415 1416 /* Verify configuration consistency */ 1417 if (!vmxnet3_verify_driver_magic(d, s->drv_shmem)) { 1418 VMW_ERPRN("Device configuration received from driver is invalid"); 1419 return; 1420 } 1421 1422 /* Verify if device is active */ 1423 if (s->device_active) { 1424 VMW_CFPRN("Vmxnet3 device is active"); 1425 return; 1426 } 1427 1428 s->txq_num = 1429 VMXNET3_READ_DRV_SHARED8(d, s->drv_shmem, devRead.misc.numTxQueues); 1430 s->rxq_num = 1431 VMXNET3_READ_DRV_SHARED8(d, s->drv_shmem, devRead.misc.numRxQueues); 1432 1433 VMW_CFPRN("Number of TX/RX queues %u/%u", s->txq_num, s->rxq_num); 1434 if (!vmxnet3_validate_queues(s)) { 1435 return; 1436 } 1437 1438 vmxnet3_adjust_by_guest_type(s); 1439 vmxnet3_update_features(s); 1440 vmxnet3_update_pm_state(s); 1441 vmxnet3_setup_rx_filtering(s); 1442 /* Cache fields from shared memory */ 1443 s->mtu = VMXNET3_READ_DRV_SHARED32(d, s->drv_shmem, devRead.misc.mtu); 1444 VMW_CFPRN("MTU is %u", s->mtu); 1445 1446 s->max_rx_frags = 1447 VMXNET3_READ_DRV_SHARED16(d, s->drv_shmem, devRead.misc.maxNumRxSG); 1448 1449 if (s->max_rx_frags == 0) { 1450 s->max_rx_frags = 1; 1451 } 1452 1453 VMW_CFPRN("Max RX fragments is %u", s->max_rx_frags); 1454 1455 s->event_int_idx = 1456 VMXNET3_READ_DRV_SHARED8(d, s->drv_shmem, devRead.intrConf.eventIntrIdx); 1457 assert(vmxnet3_verify_intx(s, s->event_int_idx)); 1458 VMW_CFPRN("Events interrupt line is %u", s->event_int_idx); 1459 1460 s->auto_int_masking = 1461 VMXNET3_READ_DRV_SHARED8(d, s->drv_shmem, devRead.intrConf.autoMask); 1462 VMW_CFPRN("Automatic interrupt masking is %d", (int)s->auto_int_masking); 1463 1464 qdescr_table_pa = 1465 VMXNET3_READ_DRV_SHARED64(d, s->drv_shmem, devRead.misc.queueDescPA); 1466 VMW_CFPRN("TX queues descriptors table is at 0x%" PRIx64, qdescr_table_pa); 1467 1468 /* 1469 * Worst-case scenario is a packet that holds all TX rings space so 1470 * we calculate total size of all TX rings for max TX fragments number 1471 */ 1472 s->max_tx_frags = 0; 1473 1474 /* TX queues */ 1475 for (i = 0; i < s->txq_num; i++) { 1476 hwaddr qdescr_pa = 1477 qdescr_table_pa + i * sizeof(struct Vmxnet3_TxQueueDesc); 1478 1479 /* Read interrupt number for this TX queue */ 1480 s->txq_descr[i].intr_idx = 1481 VMXNET3_READ_TX_QUEUE_DESCR8(d, qdescr_pa, conf.intrIdx); 1482 assert(vmxnet3_verify_intx(s, s->txq_descr[i].intr_idx)); 1483 1484 VMW_CFPRN("TX Queue %d interrupt: %d", i, s->txq_descr[i].intr_idx); 1485 1486 /* Read rings memory locations for TX queues */ 1487 pa = VMXNET3_READ_TX_QUEUE_DESCR64(d, qdescr_pa, conf.txRingBasePA); 1488 size = VMXNET3_READ_TX_QUEUE_DESCR32(d, qdescr_pa, conf.txRingSize); 1489 1490 vmxnet3_ring_init(d, &s->txq_descr[i].tx_ring, pa, size, 1491 sizeof(struct Vmxnet3_TxDesc), false); 1492 VMXNET3_RING_DUMP(VMW_CFPRN, "TX", i, &s->txq_descr[i].tx_ring); 1493 1494 s->max_tx_frags += size; 1495 1496 /* TXC ring */ 1497 pa = VMXNET3_READ_TX_QUEUE_DESCR64(d, qdescr_pa, conf.compRingBasePA); 1498 size = VMXNET3_READ_TX_QUEUE_DESCR32(d, qdescr_pa, conf.compRingSize); 1499 vmxnet3_ring_init(d, &s->txq_descr[i].comp_ring, pa, size, 1500 sizeof(struct Vmxnet3_TxCompDesc), true); 1501 VMXNET3_RING_DUMP(VMW_CFPRN, "TXC", i, &s->txq_descr[i].comp_ring); 1502 1503 s->txq_descr[i].tx_stats_pa = 1504 qdescr_pa + offsetof(struct Vmxnet3_TxQueueDesc, stats); 1505 1506 memset(&s->txq_descr[i].txq_stats, 0, 1507 sizeof(s->txq_descr[i].txq_stats)); 1508 1509 /* Fill device-managed parameters for queues */ 1510 VMXNET3_WRITE_TX_QUEUE_DESCR32(d, qdescr_pa, 1511 ctrl.txThreshold, 1512 VMXNET3_DEF_TX_THRESHOLD); 1513 } 1514 1515 /* Preallocate TX packet wrapper */ 1516 VMW_CFPRN("Max TX fragments is %u", s->max_tx_frags); 1517 net_tx_pkt_init(&s->tx_pkt, PCI_DEVICE(s), 1518 s->max_tx_frags, s->peer_has_vhdr); 1519 net_rx_pkt_init(&s->rx_pkt, s->peer_has_vhdr); 1520 1521 /* Read rings memory locations for RX queues */ 1522 for (i = 0; i < s->rxq_num; i++) { 1523 int j; 1524 hwaddr qd_pa = 1525 qdescr_table_pa + s->txq_num * sizeof(struct Vmxnet3_TxQueueDesc) + 1526 i * sizeof(struct Vmxnet3_RxQueueDesc); 1527 1528 /* Read interrupt number for this RX queue */ 1529 s->rxq_descr[i].intr_idx = 1530 VMXNET3_READ_TX_QUEUE_DESCR8(d, qd_pa, conf.intrIdx); 1531 assert(vmxnet3_verify_intx(s, s->rxq_descr[i].intr_idx)); 1532 1533 VMW_CFPRN("RX Queue %d interrupt: %d", i, s->rxq_descr[i].intr_idx); 1534 1535 /* Read rings memory locations */ 1536 for (j = 0; j < VMXNET3_RX_RINGS_PER_QUEUE; j++) { 1537 /* RX rings */ 1538 pa = VMXNET3_READ_RX_QUEUE_DESCR64(d, qd_pa, conf.rxRingBasePA[j]); 1539 size = VMXNET3_READ_RX_QUEUE_DESCR32(d, qd_pa, conf.rxRingSize[j]); 1540 vmxnet3_ring_init(d, &s->rxq_descr[i].rx_ring[j], pa, size, 1541 sizeof(struct Vmxnet3_RxDesc), false); 1542 VMW_CFPRN("RX queue %d:%d: Base: %" PRIx64 ", Size: %d", 1543 i, j, pa, size); 1544 } 1545 1546 /* RXC ring */ 1547 pa = VMXNET3_READ_RX_QUEUE_DESCR64(d, qd_pa, conf.compRingBasePA); 1548 size = VMXNET3_READ_RX_QUEUE_DESCR32(d, qd_pa, conf.compRingSize); 1549 vmxnet3_ring_init(d, &s->rxq_descr[i].comp_ring, pa, size, 1550 sizeof(struct Vmxnet3_RxCompDesc), true); 1551 VMW_CFPRN("RXC queue %d: Base: %" PRIx64 ", Size: %d", i, pa, size); 1552 1553 s->rxq_descr[i].rx_stats_pa = 1554 qd_pa + offsetof(struct Vmxnet3_RxQueueDesc, stats); 1555 memset(&s->rxq_descr[i].rxq_stats, 0, 1556 sizeof(s->rxq_descr[i].rxq_stats)); 1557 } 1558 1559 vmxnet3_validate_interrupts(s); 1560 1561 /* Make sure everything is in place before device activation */ 1562 smp_wmb(); 1563 1564 vmxnet3_reset_mac(s); 1565 1566 s->device_active = true; 1567} 1568 1569static void vmxnet3_handle_command(VMXNET3State *s, uint64_t cmd) 1570{ 1571 s->last_command = cmd; 1572 1573 switch (cmd) { 1574 case VMXNET3_CMD_GET_PERM_MAC_HI: 1575 VMW_CBPRN("Set: Get upper part of permanent MAC"); 1576 break; 1577 1578 case VMXNET3_CMD_GET_PERM_MAC_LO: 1579 VMW_CBPRN("Set: Get lower part of permanent MAC"); 1580 break; 1581 1582 case VMXNET3_CMD_GET_STATS: 1583 VMW_CBPRN("Set: Get device statistics"); 1584 vmxnet3_fill_stats(s); 1585 break; 1586 1587 case VMXNET3_CMD_ACTIVATE_DEV: 1588 VMW_CBPRN("Set: Activating vmxnet3 device"); 1589 vmxnet3_activate_device(s); 1590 break; 1591 1592 case VMXNET3_CMD_UPDATE_RX_MODE: 1593 VMW_CBPRN("Set: Update rx mode"); 1594 vmxnet3_update_rx_mode(s); 1595 break; 1596 1597 case VMXNET3_CMD_UPDATE_VLAN_FILTERS: 1598 VMW_CBPRN("Set: Update VLAN filters"); 1599 vmxnet3_update_vlan_filters(s); 1600 break; 1601 1602 case VMXNET3_CMD_UPDATE_MAC_FILTERS: 1603 VMW_CBPRN("Set: Update MAC filters"); 1604 vmxnet3_update_mcast_filters(s); 1605 break; 1606 1607 case VMXNET3_CMD_UPDATE_FEATURE: 1608 VMW_CBPRN("Set: Update features"); 1609 vmxnet3_update_features(s); 1610 break; 1611 1612 case VMXNET3_CMD_UPDATE_PMCFG: 1613 VMW_CBPRN("Set: Update power management config"); 1614 vmxnet3_update_pm_state(s); 1615 break; 1616 1617 case VMXNET3_CMD_GET_LINK: 1618 VMW_CBPRN("Set: Get link"); 1619 break; 1620 1621 case VMXNET3_CMD_RESET_DEV: 1622 VMW_CBPRN("Set: Reset device"); 1623 vmxnet3_reset(s); 1624 break; 1625 1626 case VMXNET3_CMD_QUIESCE_DEV: 1627 VMW_CBPRN("Set: VMXNET3_CMD_QUIESCE_DEV - deactivate the device"); 1628 vmxnet3_deactivate_device(s); 1629 break; 1630 1631 case VMXNET3_CMD_GET_CONF_INTR: 1632 VMW_CBPRN("Set: VMXNET3_CMD_GET_CONF_INTR - interrupt configuration"); 1633 break; 1634 1635 case VMXNET3_CMD_GET_ADAPTIVE_RING_INFO: 1636 VMW_CBPRN("Set: VMXNET3_CMD_GET_ADAPTIVE_RING_INFO - " 1637 "adaptive ring info flags"); 1638 break; 1639 1640 case VMXNET3_CMD_GET_DID_LO: 1641 VMW_CBPRN("Set: Get lower part of device ID"); 1642 break; 1643 1644 case VMXNET3_CMD_GET_DID_HI: 1645 VMW_CBPRN("Set: Get upper part of device ID"); 1646 break; 1647 1648 case VMXNET3_CMD_GET_DEV_EXTRA_INFO: 1649 VMW_CBPRN("Set: Get device extra info"); 1650 break; 1651 1652 default: 1653 VMW_CBPRN("Received unknown command: %" PRIx64, cmd); 1654 break; 1655 } 1656} 1657 1658static uint64_t vmxnet3_get_command_status(VMXNET3State *s) 1659{ 1660 uint64_t ret; 1661 1662 switch (s->last_command) { 1663 case VMXNET3_CMD_ACTIVATE_DEV: 1664 ret = (s->device_active) ? 0 : 1; 1665 VMW_CFPRN("Device active: %" PRIx64, ret); 1666 break; 1667 1668 case VMXNET3_CMD_RESET_DEV: 1669 case VMXNET3_CMD_QUIESCE_DEV: 1670 case VMXNET3_CMD_GET_QUEUE_STATUS: 1671 case VMXNET3_CMD_GET_DEV_EXTRA_INFO: 1672 ret = 0; 1673 break; 1674 1675 case VMXNET3_CMD_GET_LINK: 1676 ret = s->link_status_and_speed; 1677 VMW_CFPRN("Link and speed: %" PRIx64, ret); 1678 break; 1679 1680 case VMXNET3_CMD_GET_PERM_MAC_LO: 1681 ret = vmxnet3_get_mac_low(&s->perm_mac); 1682 break; 1683 1684 case VMXNET3_CMD_GET_PERM_MAC_HI: 1685 ret = vmxnet3_get_mac_high(&s->perm_mac); 1686 break; 1687 1688 case VMXNET3_CMD_GET_CONF_INTR: 1689 ret = vmxnet3_get_interrupt_config(s); 1690 break; 1691 1692 case VMXNET3_CMD_GET_ADAPTIVE_RING_INFO: 1693 ret = VMXNET3_DISABLE_ADAPTIVE_RING; 1694 break; 1695 1696 case VMXNET3_CMD_GET_DID_LO: 1697 ret = PCI_DEVICE_ID_VMWARE_VMXNET3; 1698 break; 1699 1700 case VMXNET3_CMD_GET_DID_HI: 1701 ret = VMXNET3_DEVICE_REVISION; 1702 break; 1703 1704 default: 1705 VMW_WRPRN("Received request for unknown command: %x", s->last_command); 1706 ret = 0; 1707 break; 1708 } 1709 1710 return ret; 1711} 1712 1713static void vmxnet3_set_events(VMXNET3State *s, uint32_t val) 1714{ 1715 uint32_t events; 1716 PCIDevice *d = PCI_DEVICE(s); 1717 1718 VMW_CBPRN("Setting events: 0x%x", val); 1719 events = VMXNET3_READ_DRV_SHARED32(d, s->drv_shmem, ecr) | val; 1720 VMXNET3_WRITE_DRV_SHARED32(d, s->drv_shmem, ecr, events); 1721} 1722 1723static void vmxnet3_ack_events(VMXNET3State *s, uint32_t val) 1724{ 1725 PCIDevice *d = PCI_DEVICE(s); 1726 uint32_t events; 1727 1728 VMW_CBPRN("Clearing events: 0x%x", val); 1729 events = VMXNET3_READ_DRV_SHARED32(d, s->drv_shmem, ecr) & ~val; 1730 VMXNET3_WRITE_DRV_SHARED32(d, s->drv_shmem, ecr, events); 1731} 1732 1733static void 1734vmxnet3_io_bar1_write(void *opaque, 1735 hwaddr addr, 1736 uint64_t val, 1737 unsigned size) 1738{ 1739 VMXNET3State *s = opaque; 1740 1741 switch (addr) { 1742 /* Vmxnet3 Revision Report Selection */ 1743 case VMXNET3_REG_VRRS: 1744 VMW_CBPRN("Write BAR1 [VMXNET3_REG_VRRS] = %" PRIx64 ", size %d", 1745 val, size); 1746 break; 1747 1748 /* UPT Version Report Selection */ 1749 case VMXNET3_REG_UVRS: 1750 VMW_CBPRN("Write BAR1 [VMXNET3_REG_UVRS] = %" PRIx64 ", size %d", 1751 val, size); 1752 break; 1753 1754 /* Driver Shared Address Low */ 1755 case VMXNET3_REG_DSAL: 1756 VMW_CBPRN("Write BAR1 [VMXNET3_REG_DSAL] = %" PRIx64 ", size %d", 1757 val, size); 1758 /* 1759 * Guest driver will first write the low part of the shared 1760 * memory address. We save it to temp variable and set the 1761 * shared address only after we get the high part 1762 */ 1763 if (val == 0) { 1764 vmxnet3_deactivate_device(s); 1765 } 1766 s->temp_shared_guest_driver_memory = val; 1767 s->drv_shmem = 0; 1768 break; 1769 1770 /* Driver Shared Address High */ 1771 case VMXNET3_REG_DSAH: 1772 VMW_CBPRN("Write BAR1 [VMXNET3_REG_DSAH] = %" PRIx64 ", size %d", 1773 val, size); 1774 /* 1775 * Set the shared memory between guest driver and device. 1776 * We already should have low address part. 1777 */ 1778 s->drv_shmem = s->temp_shared_guest_driver_memory | (val << 32); 1779 break; 1780 1781 /* Command */ 1782 case VMXNET3_REG_CMD: 1783 VMW_CBPRN("Write BAR1 [VMXNET3_REG_CMD] = %" PRIx64 ", size %d", 1784 val, size); 1785 vmxnet3_handle_command(s, val); 1786 break; 1787 1788 /* MAC Address Low */ 1789 case VMXNET3_REG_MACL: 1790 VMW_CBPRN("Write BAR1 [VMXNET3_REG_MACL] = %" PRIx64 ", size %d", 1791 val, size); 1792 s->temp_mac = val; 1793 break; 1794 1795 /* MAC Address High */ 1796 case VMXNET3_REG_MACH: 1797 VMW_CBPRN("Write BAR1 [VMXNET3_REG_MACH] = %" PRIx64 ", size %d", 1798 val, size); 1799 vmxnet3_set_variable_mac(s, val, s->temp_mac); 1800 break; 1801 1802 /* Interrupt Cause Register */ 1803 case VMXNET3_REG_ICR: 1804 VMW_CBPRN("Write BAR1 [VMXNET3_REG_ICR] = %" PRIx64 ", size %d", 1805 val, size); 1806 g_assert_not_reached(); 1807 break; 1808 1809 /* Event Cause Register */ 1810 case VMXNET3_REG_ECR: 1811 VMW_CBPRN("Write BAR1 [VMXNET3_REG_ECR] = %" PRIx64 ", size %d", 1812 val, size); 1813 vmxnet3_ack_events(s, val); 1814 break; 1815 1816 default: 1817 VMW_CBPRN("Unknown Write to BAR1 [%" PRIx64 "] = %" PRIx64 ", size %d", 1818 addr, val, size); 1819 break; 1820 } 1821} 1822 1823static uint64_t 1824vmxnet3_io_bar1_read(void *opaque, hwaddr addr, unsigned size) 1825{ 1826 VMXNET3State *s = opaque; 1827 uint64_t ret = 0; 1828 1829 switch (addr) { 1830 /* Vmxnet3 Revision Report Selection */ 1831 case VMXNET3_REG_VRRS: 1832 VMW_CBPRN("Read BAR1 [VMXNET3_REG_VRRS], size %d", size); 1833 ret = VMXNET3_DEVICE_REVISION; 1834 break; 1835 1836 /* UPT Version Report Selection */ 1837 case VMXNET3_REG_UVRS: 1838 VMW_CBPRN("Read BAR1 [VMXNET3_REG_UVRS], size %d", size); 1839 ret = VMXNET3_UPT_REVISION; 1840 break; 1841 1842 /* Command */ 1843 case VMXNET3_REG_CMD: 1844 VMW_CBPRN("Read BAR1 [VMXNET3_REG_CMD], size %d", size); 1845 ret = vmxnet3_get_command_status(s); 1846 break; 1847 1848 /* MAC Address Low */ 1849 case VMXNET3_REG_MACL: 1850 VMW_CBPRN("Read BAR1 [VMXNET3_REG_MACL], size %d", size); 1851 ret = vmxnet3_get_mac_low(&s->conf.macaddr); 1852 break; 1853 1854 /* MAC Address High */ 1855 case VMXNET3_REG_MACH: 1856 VMW_CBPRN("Read BAR1 [VMXNET3_REG_MACH], size %d", size); 1857 ret = vmxnet3_get_mac_high(&s->conf.macaddr); 1858 break; 1859 1860 /* 1861 * Interrupt Cause Register 1862 * Used for legacy interrupts only so interrupt index always 0 1863 */ 1864 case VMXNET3_REG_ICR: 1865 VMW_CBPRN("Read BAR1 [VMXNET3_REG_ICR], size %d", size); 1866 if (vmxnet3_interrupt_asserted(s, 0)) { 1867 vmxnet3_clear_interrupt(s, 0); 1868 ret = true; 1869 } else { 1870 ret = false; 1871 } 1872 break; 1873 1874 default: 1875 VMW_CBPRN("Unknow read BAR1[%" PRIx64 "], %d bytes", addr, size); 1876 break; 1877 } 1878 1879 return ret; 1880} 1881 1882static int 1883vmxnet3_can_receive(NetClientState *nc) 1884{ 1885 VMXNET3State *s = qemu_get_nic_opaque(nc); 1886 return s->device_active && 1887 VMXNET_FLAG_IS_SET(s->link_status_and_speed, VMXNET3_LINK_STATUS_UP); 1888} 1889 1890static inline bool 1891vmxnet3_is_registered_vlan(VMXNET3State *s, const void *data) 1892{ 1893 uint16_t vlan_tag = eth_get_pkt_tci(data) & VLAN_VID_MASK; 1894 if (IS_SPECIAL_VLAN_ID(vlan_tag)) { 1895 return true; 1896 } 1897 1898 return VMXNET3_VFTABLE_ENTRY_IS_SET(s->vlan_table, vlan_tag); 1899} 1900 1901static bool 1902vmxnet3_is_allowed_mcast_group(VMXNET3State *s, const uint8_t *group_mac) 1903{ 1904 int i; 1905 for (i = 0; i < s->mcast_list_len; i++) { 1906 if (!memcmp(group_mac, s->mcast_list[i].a, sizeof(s->mcast_list[i]))) { 1907 return true; 1908 } 1909 } 1910 return false; 1911} 1912 1913static bool 1914vmxnet3_rx_filter_may_indicate(VMXNET3State *s, const void *data, 1915 size_t size) 1916{ 1917 struct eth_header *ehdr = PKT_GET_ETH_HDR(data); 1918 1919 if (VMXNET_FLAG_IS_SET(s->rx_mode, VMXNET3_RXM_PROMISC)) { 1920 return true; 1921 } 1922 1923 if (!vmxnet3_is_registered_vlan(s, data)) { 1924 return false; 1925 } 1926 1927 switch (net_rx_pkt_get_packet_type(s->rx_pkt)) { 1928 case ETH_PKT_UCAST: 1929 if (!VMXNET_FLAG_IS_SET(s->rx_mode, VMXNET3_RXM_UCAST)) { 1930 return false; 1931 } 1932 if (memcmp(s->conf.macaddr.a, ehdr->h_dest, ETH_ALEN)) { 1933 return false; 1934 } 1935 break; 1936 1937 case ETH_PKT_BCAST: 1938 if (!VMXNET_FLAG_IS_SET(s->rx_mode, VMXNET3_RXM_BCAST)) { 1939 return false; 1940 } 1941 break; 1942 1943 case ETH_PKT_MCAST: 1944 if (VMXNET_FLAG_IS_SET(s->rx_mode, VMXNET3_RXM_ALL_MULTI)) { 1945 return true; 1946 } 1947 if (!VMXNET_FLAG_IS_SET(s->rx_mode, VMXNET3_RXM_MCAST)) { 1948 return false; 1949 } 1950 if (!vmxnet3_is_allowed_mcast_group(s, ehdr->h_dest)) { 1951 return false; 1952 } 1953 break; 1954 1955 default: 1956 g_assert_not_reached(); 1957 } 1958 1959 return true; 1960} 1961 1962static ssize_t 1963vmxnet3_receive(NetClientState *nc, const uint8_t *buf, size_t size) 1964{ 1965 VMXNET3State *s = qemu_get_nic_opaque(nc); 1966 size_t bytes_indicated; 1967 uint8_t min_buf[MIN_BUF_SIZE]; 1968 1969 if (!vmxnet3_can_receive(nc)) { 1970 VMW_PKPRN("Cannot receive now"); 1971 return -1; 1972 } 1973 1974 if (s->peer_has_vhdr) { 1975 net_rx_pkt_set_vhdr(s->rx_pkt, (struct virtio_net_hdr *)buf); 1976 buf += sizeof(struct virtio_net_hdr); 1977 size -= sizeof(struct virtio_net_hdr); 1978 } 1979 1980 /* Pad to minimum Ethernet frame length */ 1981 if (size < sizeof(min_buf)) { 1982 memcpy(min_buf, buf, size); 1983 memset(&min_buf[size], 0, sizeof(min_buf) - size); 1984 buf = min_buf; 1985 size = sizeof(min_buf); 1986 } 1987 1988 net_rx_pkt_set_packet_type(s->rx_pkt, 1989 get_eth_packet_type(PKT_GET_ETH_HDR(buf))); 1990 1991 if (vmxnet3_rx_filter_may_indicate(s, buf, size)) { 1992 net_rx_pkt_set_protocols(s->rx_pkt, buf, size); 1993 vmxnet3_rx_need_csum_calculate(s->rx_pkt, buf, size); 1994 net_rx_pkt_attach_data(s->rx_pkt, buf, size, s->rx_vlan_stripping); 1995 bytes_indicated = vmxnet3_indicate_packet(s) ? size : -1; 1996 if (bytes_indicated < size) { 1997 VMW_PKPRN("RX: %zu of %zu bytes indicated", bytes_indicated, size); 1998 } 1999 } else { 2000 VMW_PKPRN("Packet dropped by RX filter"); 2001 bytes_indicated = size; 2002 } 2003 2004 assert(size > 0); 2005 assert(bytes_indicated != 0); 2006 return bytes_indicated; 2007} 2008 2009static void vmxnet3_set_link_status(NetClientState *nc) 2010{ 2011 VMXNET3State *s = qemu_get_nic_opaque(nc); 2012 2013 if (nc->link_down) { 2014 s->link_status_and_speed &= ~VMXNET3_LINK_STATUS_UP; 2015 } else { 2016 s->link_status_and_speed |= VMXNET3_LINK_STATUS_UP; 2017 } 2018 2019 vmxnet3_set_events(s, VMXNET3_ECR_LINK); 2020 vmxnet3_trigger_interrupt(s, s->event_int_idx); 2021} 2022 2023static NetClientInfo net_vmxnet3_info = { 2024 .type = NET_CLIENT_DRIVER_NIC, 2025 .size = sizeof(NICState), 2026 .receive = vmxnet3_receive, 2027 .link_status_changed = vmxnet3_set_link_status, 2028}; 2029 2030static bool vmxnet3_peer_has_vnet_hdr(VMXNET3State *s) 2031{ 2032 NetClientState *nc = qemu_get_queue(s->nic); 2033 2034 if (qemu_has_vnet_hdr(nc->peer)) { 2035 return true; 2036 } 2037 2038 return false; 2039} 2040 2041static void vmxnet3_net_uninit(VMXNET3State *s) 2042{ 2043 g_free(s->mcast_list); 2044 vmxnet3_deactivate_device(s); 2045 qemu_del_nic(s->nic); 2046} 2047 2048static void vmxnet3_net_init(VMXNET3State *s) 2049{ 2050 DeviceState *d = DEVICE(s); 2051 2052 VMW_CBPRN("vmxnet3_net_init called..."); 2053 2054 qemu_macaddr_default_if_unset(&s->conf.macaddr); 2055 2056 /* Windows guest will query the address that was set on init */ 2057 memcpy(&s->perm_mac.a, &s->conf.macaddr.a, sizeof(s->perm_mac.a)); 2058 2059 s->mcast_list = NULL; 2060 s->mcast_list_len = 0; 2061 2062 s->link_status_and_speed = VMXNET3_LINK_SPEED | VMXNET3_LINK_STATUS_UP; 2063 2064 VMW_CFPRN("Permanent MAC: " MAC_FMT, MAC_ARG(s->perm_mac.a)); 2065 2066 s->nic = qemu_new_nic(&net_vmxnet3_info, &s->conf, 2067 object_get_typename(OBJECT(s)), 2068 d->id, s); 2069 2070 s->peer_has_vhdr = vmxnet3_peer_has_vnet_hdr(s); 2071 s->tx_sop = true; 2072 s->skip_current_tx_pkt = false; 2073 s->tx_pkt = NULL; 2074 s->rx_pkt = NULL; 2075 s->rx_vlan_stripping = false; 2076 s->lro_supported = false; 2077 2078 if (s->peer_has_vhdr) { 2079 qemu_set_vnet_hdr_len(qemu_get_queue(s->nic)->peer, 2080 sizeof(struct virtio_net_hdr)); 2081 2082 qemu_using_vnet_hdr(qemu_get_queue(s->nic)->peer, 1); 2083 } 2084 2085 qemu_format_nic_info_str(qemu_get_queue(s->nic), s->conf.macaddr.a); 2086} 2087 2088static void 2089vmxnet3_unuse_msix_vectors(VMXNET3State *s, int num_vectors) 2090{ 2091 PCIDevice *d = PCI_DEVICE(s); 2092 int i; 2093 for (i = 0; i < num_vectors; i++) { 2094 msix_vector_unuse(d, i); 2095 } 2096} 2097 2098static bool 2099vmxnet3_use_msix_vectors(VMXNET3State *s, int num_vectors) 2100{ 2101 PCIDevice *d = PCI_DEVICE(s); 2102 int i; 2103 for (i = 0; i < num_vectors; i++) { 2104 int res = msix_vector_use(d, i); 2105 if (0 > res) { 2106 VMW_WRPRN("Failed to use MSI-X vector %d, error %d", i, res); 2107 vmxnet3_unuse_msix_vectors(s, i); 2108 return false; 2109 } 2110 } 2111 return true; 2112} 2113 2114static bool 2115vmxnet3_init_msix(VMXNET3State *s) 2116{ 2117 PCIDevice *d = PCI_DEVICE(s); 2118 int res = msix_init(d, VMXNET3_MAX_INTRS, 2119 &s->msix_bar, 2120 VMXNET3_MSIX_BAR_IDX, VMXNET3_OFF_MSIX_TABLE, 2121 &s->msix_bar, 2122 VMXNET3_MSIX_BAR_IDX, VMXNET3_OFF_MSIX_PBA(s), 2123 VMXNET3_MSIX_OFFSET(s), NULL); 2124 2125 if (0 > res) { 2126 VMW_WRPRN("Failed to initialize MSI-X, error %d", res); 2127 s->msix_used = false; 2128 } else { 2129 if (!vmxnet3_use_msix_vectors(s, VMXNET3_MAX_INTRS)) { 2130 VMW_WRPRN("Failed to use MSI-X vectors, error %d", res); 2131 msix_uninit(d, &s->msix_bar, &s->msix_bar); 2132 s->msix_used = false; 2133 } else { 2134 s->msix_used = true; 2135 } 2136 } 2137 return s->msix_used; 2138} 2139 2140static void 2141vmxnet3_cleanup_msix(VMXNET3State *s) 2142{ 2143 PCIDevice *d = PCI_DEVICE(s); 2144 2145 if (s->msix_used) { 2146 vmxnet3_unuse_msix_vectors(s, VMXNET3_MAX_INTRS); 2147 msix_uninit(d, &s->msix_bar, &s->msix_bar); 2148 } 2149} 2150 2151static void 2152vmxnet3_cleanup_msi(VMXNET3State *s) 2153{ 2154 PCIDevice *d = PCI_DEVICE(s); 2155 2156 msi_uninit(d); 2157} 2158 2159static const MemoryRegionOps b0_ops = { 2160 .read = vmxnet3_io_bar0_read, 2161 .write = vmxnet3_io_bar0_write, 2162 .endianness = DEVICE_LITTLE_ENDIAN, 2163 .impl = { 2164 .min_access_size = 4, 2165 .max_access_size = 4, 2166 }, 2167}; 2168 2169static const MemoryRegionOps b1_ops = { 2170 .read = vmxnet3_io_bar1_read, 2171 .write = vmxnet3_io_bar1_write, 2172 .endianness = DEVICE_LITTLE_ENDIAN, 2173 .impl = { 2174 .min_access_size = 4, 2175 .max_access_size = 4, 2176 }, 2177}; 2178 2179static uint64_t vmxnet3_device_serial_num(VMXNET3State *s) 2180{ 2181 uint64_t dsn_payload; 2182 uint8_t *dsnp = (uint8_t *)&dsn_payload; 2183 2184 dsnp[0] = 0xfe; 2185 dsnp[1] = s->conf.macaddr.a[3]; 2186 dsnp[2] = s->conf.macaddr.a[4]; 2187 dsnp[3] = s->conf.macaddr.a[5]; 2188 dsnp[4] = s->conf.macaddr.a[0]; 2189 dsnp[5] = s->conf.macaddr.a[1]; 2190 dsnp[6] = s->conf.macaddr.a[2]; 2191 dsnp[7] = 0xff; 2192 return dsn_payload; 2193} 2194 2195 2196#define VMXNET3_USE_64BIT (true) 2197#define VMXNET3_PER_VECTOR_MASK (false) 2198 2199static void vmxnet3_pci_realize(PCIDevice *pci_dev, Error **errp) 2200{ 2201 VMXNET3State *s = VMXNET3(pci_dev); 2202 int ret; 2203 2204 VMW_CBPRN("Starting init..."); 2205 2206 memory_region_init_io(&s->bar0, OBJECT(s), &b0_ops, s, 2207 "vmxnet3-b0", VMXNET3_PT_REG_SIZE); 2208 pci_register_bar(pci_dev, VMXNET3_BAR0_IDX, 2209 PCI_BASE_ADDRESS_SPACE_MEMORY, &s->bar0); 2210 2211 memory_region_init_io(&s->bar1, OBJECT(s), &b1_ops, s, 2212 "vmxnet3-b1", VMXNET3_VD_REG_SIZE); 2213 pci_register_bar(pci_dev, VMXNET3_BAR1_IDX, 2214 PCI_BASE_ADDRESS_SPACE_MEMORY, &s->bar1); 2215 2216 memory_region_init(&s->msix_bar, OBJECT(s), "vmxnet3-msix-bar", 2217 VMXNET3_MSIX_BAR_SIZE); 2218 pci_register_bar(pci_dev, VMXNET3_MSIX_BAR_IDX, 2219 PCI_BASE_ADDRESS_SPACE_MEMORY, &s->msix_bar); 2220 2221 vmxnet3_reset_interrupt_states(s); 2222 2223 /* Interrupt pin A */ 2224 pci_dev->config[PCI_INTERRUPT_PIN] = 0x01; 2225 2226 ret = msi_init(pci_dev, VMXNET3_MSI_OFFSET(s), VMXNET3_MAX_NMSIX_INTRS, 2227 VMXNET3_USE_64BIT, VMXNET3_PER_VECTOR_MASK, NULL); 2228 /* Any error other than -ENOTSUP(board's MSI support is broken) 2229 * is a programming error. Fall back to INTx silently on -ENOTSUP */ 2230 assert(!ret || ret == -ENOTSUP); 2231 2232 if (!vmxnet3_init_msix(s)) { 2233 VMW_WRPRN("Failed to initialize MSI-X, configuration is inconsistent."); 2234 } 2235 2236 vmxnet3_net_init(s); 2237 2238 if (pci_is_express(pci_dev)) { 2239 if (pci_bus_is_express(pci_get_bus(pci_dev))) { 2240 pcie_endpoint_cap_init(pci_dev, VMXNET3_EXP_EP_OFFSET); 2241 } 2242 2243 pcie_dev_ser_num_init(pci_dev, VMXNET3_DSN_OFFSET, 2244 vmxnet3_device_serial_num(s)); 2245 } 2246} 2247 2248static void vmxnet3_instance_init(Object *obj) 2249{ 2250 VMXNET3State *s = VMXNET3(obj); 2251 device_add_bootindex_property(obj, &s->conf.bootindex, 2252 "bootindex", "/ethernet-phy@0", 2253 DEVICE(obj)); 2254} 2255 2256static void vmxnet3_pci_uninit(PCIDevice *pci_dev) 2257{ 2258 VMXNET3State *s = VMXNET3(pci_dev); 2259 2260 VMW_CBPRN("Starting uninit..."); 2261 2262 vmxnet3_net_uninit(s); 2263 2264 vmxnet3_cleanup_msix(s); 2265 2266 vmxnet3_cleanup_msi(s); 2267} 2268 2269static void vmxnet3_qdev_reset(DeviceState *dev) 2270{ 2271 PCIDevice *d = PCI_DEVICE(dev); 2272 VMXNET3State *s = VMXNET3(d); 2273 2274 VMW_CBPRN("Starting QDEV reset..."); 2275 vmxnet3_reset(s); 2276} 2277 2278static bool vmxnet3_mc_list_needed(void *opaque) 2279{ 2280 return true; 2281} 2282 2283static int vmxnet3_mcast_list_pre_load(void *opaque) 2284{ 2285 VMXNET3State *s = opaque; 2286 2287 s->mcast_list = g_malloc(s->mcast_list_buff_size); 2288 2289 return 0; 2290} 2291 2292 2293static int vmxnet3_pre_save(void *opaque) 2294{ 2295 VMXNET3State *s = opaque; 2296 2297 s->mcast_list_buff_size = s->mcast_list_len * sizeof(MACAddr); 2298 2299 return 0; 2300} 2301 2302static const VMStateDescription vmxstate_vmxnet3_mcast_list = { 2303 .name = "vmxnet3/mcast_list", 2304 .version_id = 1, 2305 .minimum_version_id = 1, 2306 .pre_load = vmxnet3_mcast_list_pre_load, 2307 .needed = vmxnet3_mc_list_needed, 2308 .fields = (VMStateField[]) { 2309 VMSTATE_VBUFFER_UINT32(mcast_list, VMXNET3State, 0, NULL, 2310 mcast_list_buff_size), 2311 VMSTATE_END_OF_LIST() 2312 } 2313}; 2314 2315static const VMStateDescription vmstate_vmxnet3_ring = { 2316 .name = "vmxnet3-ring", 2317 .version_id = 0, 2318 .fields = (VMStateField[]) { 2319 VMSTATE_UINT64(pa, Vmxnet3Ring), 2320 VMSTATE_UINT32(size, Vmxnet3Ring), 2321 VMSTATE_UINT32(cell_size, Vmxnet3Ring), 2322 VMSTATE_UINT32(next, Vmxnet3Ring), 2323 VMSTATE_UINT8(gen, Vmxnet3Ring), 2324 VMSTATE_END_OF_LIST() 2325 } 2326}; 2327 2328static const VMStateDescription vmstate_vmxnet3_tx_stats = { 2329 .name = "vmxnet3-tx-stats", 2330 .version_id = 0, 2331 .fields = (VMStateField[]) { 2332 VMSTATE_UINT64(TSOPktsTxOK, struct UPT1_TxStats), 2333 VMSTATE_UINT64(TSOBytesTxOK, struct UPT1_TxStats), 2334 VMSTATE_UINT64(ucastPktsTxOK, struct UPT1_TxStats), 2335 VMSTATE_UINT64(ucastBytesTxOK, struct UPT1_TxStats), 2336 VMSTATE_UINT64(mcastPktsTxOK, struct UPT1_TxStats), 2337 VMSTATE_UINT64(mcastBytesTxOK, struct UPT1_TxStats), 2338 VMSTATE_UINT64(bcastPktsTxOK, struct UPT1_TxStats), 2339 VMSTATE_UINT64(bcastBytesTxOK, struct UPT1_TxStats), 2340 VMSTATE_UINT64(pktsTxError, struct UPT1_TxStats), 2341 VMSTATE_UINT64(pktsTxDiscard, struct UPT1_TxStats), 2342 VMSTATE_END_OF_LIST() 2343 } 2344}; 2345 2346static const VMStateDescription vmstate_vmxnet3_txq_descr = { 2347 .name = "vmxnet3-txq-descr", 2348 .version_id = 0, 2349 .fields = (VMStateField[]) { 2350 VMSTATE_STRUCT(tx_ring, Vmxnet3TxqDescr, 0, vmstate_vmxnet3_ring, 2351 Vmxnet3Ring), 2352 VMSTATE_STRUCT(comp_ring, Vmxnet3TxqDescr, 0, vmstate_vmxnet3_ring, 2353 Vmxnet3Ring), 2354 VMSTATE_UINT8(intr_idx, Vmxnet3TxqDescr), 2355 VMSTATE_UINT64(tx_stats_pa, Vmxnet3TxqDescr), 2356 VMSTATE_STRUCT(txq_stats, Vmxnet3TxqDescr, 0, vmstate_vmxnet3_tx_stats, 2357 struct UPT1_TxStats), 2358 VMSTATE_END_OF_LIST() 2359 } 2360}; 2361 2362static const VMStateDescription vmstate_vmxnet3_rx_stats = { 2363 .name = "vmxnet3-rx-stats", 2364 .version_id = 0, 2365 .fields = (VMStateField[]) { 2366 VMSTATE_UINT64(LROPktsRxOK, struct UPT1_RxStats), 2367 VMSTATE_UINT64(LROBytesRxOK, struct UPT1_RxStats), 2368 VMSTATE_UINT64(ucastPktsRxOK, struct UPT1_RxStats), 2369 VMSTATE_UINT64(ucastBytesRxOK, struct UPT1_RxStats), 2370 VMSTATE_UINT64(mcastPktsRxOK, struct UPT1_RxStats), 2371 VMSTATE_UINT64(mcastBytesRxOK, struct UPT1_RxStats), 2372 VMSTATE_UINT64(bcastPktsRxOK, struct UPT1_RxStats), 2373 VMSTATE_UINT64(bcastBytesRxOK, struct UPT1_RxStats), 2374 VMSTATE_UINT64(pktsRxOutOfBuf, struct UPT1_RxStats), 2375 VMSTATE_UINT64(pktsRxError, struct UPT1_RxStats), 2376 VMSTATE_END_OF_LIST() 2377 } 2378}; 2379 2380static const VMStateDescription vmstate_vmxnet3_rxq_descr = { 2381 .name = "vmxnet3-rxq-descr", 2382 .version_id = 0, 2383 .fields = (VMStateField[]) { 2384 VMSTATE_STRUCT_ARRAY(rx_ring, Vmxnet3RxqDescr, 2385 VMXNET3_RX_RINGS_PER_QUEUE, 0, 2386 vmstate_vmxnet3_ring, Vmxnet3Ring), 2387 VMSTATE_STRUCT(comp_ring, Vmxnet3RxqDescr, 0, vmstate_vmxnet3_ring, 2388 Vmxnet3Ring), 2389 VMSTATE_UINT8(intr_idx, Vmxnet3RxqDescr), 2390 VMSTATE_UINT64(rx_stats_pa, Vmxnet3RxqDescr), 2391 VMSTATE_STRUCT(rxq_stats, Vmxnet3RxqDescr, 0, vmstate_vmxnet3_rx_stats, 2392 struct UPT1_RxStats), 2393 VMSTATE_END_OF_LIST() 2394 } 2395}; 2396 2397static int vmxnet3_post_load(void *opaque, int version_id) 2398{ 2399 VMXNET3State *s = opaque; 2400 PCIDevice *d = PCI_DEVICE(s); 2401 2402 net_tx_pkt_init(&s->tx_pkt, PCI_DEVICE(s), 2403 s->max_tx_frags, s->peer_has_vhdr); 2404 net_rx_pkt_init(&s->rx_pkt, s->peer_has_vhdr); 2405 2406 if (s->msix_used) { 2407 if (!vmxnet3_use_msix_vectors(s, VMXNET3_MAX_INTRS)) { 2408 VMW_WRPRN("Failed to re-use MSI-X vectors"); 2409 msix_uninit(d, &s->msix_bar, &s->msix_bar); 2410 s->msix_used = false; 2411 return -1; 2412 } 2413 } 2414 2415 if (!vmxnet3_validate_queues(s)) { 2416 return -1; 2417 } 2418 vmxnet3_validate_interrupts(s); 2419 2420 return 0; 2421} 2422 2423static const VMStateDescription vmstate_vmxnet3_int_state = { 2424 .name = "vmxnet3-int-state", 2425 .version_id = 0, 2426 .fields = (VMStateField[]) { 2427 VMSTATE_BOOL(is_masked, Vmxnet3IntState), 2428 VMSTATE_BOOL(is_pending, Vmxnet3IntState), 2429 VMSTATE_BOOL(is_asserted, Vmxnet3IntState), 2430 VMSTATE_END_OF_LIST() 2431 } 2432}; 2433 2434static const VMStateDescription vmstate_vmxnet3 = { 2435 .name = "vmxnet3", 2436 .version_id = 1, 2437 .minimum_version_id = 1, 2438 .pre_save = vmxnet3_pre_save, 2439 .post_load = vmxnet3_post_load, 2440 .fields = (VMStateField[]) { 2441 VMSTATE_PCI_DEVICE(parent_obj, VMXNET3State), 2442 VMSTATE_MSIX(parent_obj, VMXNET3State), 2443 VMSTATE_BOOL(rx_packets_compound, VMXNET3State), 2444 VMSTATE_BOOL(rx_vlan_stripping, VMXNET3State), 2445 VMSTATE_BOOL(lro_supported, VMXNET3State), 2446 VMSTATE_UINT32(rx_mode, VMXNET3State), 2447 VMSTATE_UINT32(mcast_list_len, VMXNET3State), 2448 VMSTATE_UINT32(mcast_list_buff_size, VMXNET3State), 2449 VMSTATE_UINT32_ARRAY(vlan_table, VMXNET3State, VMXNET3_VFT_SIZE), 2450 VMSTATE_UINT32(mtu, VMXNET3State), 2451 VMSTATE_UINT16(max_rx_frags, VMXNET3State), 2452 VMSTATE_UINT32(max_tx_frags, VMXNET3State), 2453 VMSTATE_UINT8(event_int_idx, VMXNET3State), 2454 VMSTATE_BOOL(auto_int_masking, VMXNET3State), 2455 VMSTATE_UINT8(txq_num, VMXNET3State), 2456 VMSTATE_UINT8(rxq_num, VMXNET3State), 2457 VMSTATE_UINT32(device_active, VMXNET3State), 2458 VMSTATE_UINT32(last_command, VMXNET3State), 2459 VMSTATE_UINT32(link_status_and_speed, VMXNET3State), 2460 VMSTATE_UINT32(temp_mac, VMXNET3State), 2461 VMSTATE_UINT64(drv_shmem, VMXNET3State), 2462 VMSTATE_UINT64(temp_shared_guest_driver_memory, VMXNET3State), 2463 2464 VMSTATE_STRUCT_ARRAY(txq_descr, VMXNET3State, 2465 VMXNET3_DEVICE_MAX_TX_QUEUES, 0, vmstate_vmxnet3_txq_descr, 2466 Vmxnet3TxqDescr), 2467 VMSTATE_STRUCT_ARRAY(rxq_descr, VMXNET3State, 2468 VMXNET3_DEVICE_MAX_RX_QUEUES, 0, vmstate_vmxnet3_rxq_descr, 2469 Vmxnet3RxqDescr), 2470 VMSTATE_STRUCT_ARRAY(interrupt_states, VMXNET3State, 2471 VMXNET3_MAX_INTRS, 0, vmstate_vmxnet3_int_state, 2472 Vmxnet3IntState), 2473 2474 VMSTATE_END_OF_LIST() 2475 }, 2476 .subsections = (const VMStateDescription*[]) { 2477 &vmxstate_vmxnet3_mcast_list, 2478 NULL 2479 } 2480}; 2481 2482static Property vmxnet3_properties[] = { 2483 DEFINE_NIC_PROPERTIES(VMXNET3State, conf), 2484 DEFINE_PROP_BIT("x-old-msi-offsets", VMXNET3State, compat_flags, 2485 VMXNET3_COMPAT_FLAG_OLD_MSI_OFFSETS_BIT, false), 2486 DEFINE_PROP_BIT("x-disable-pcie", VMXNET3State, compat_flags, 2487 VMXNET3_COMPAT_FLAG_DISABLE_PCIE_BIT, false), 2488 DEFINE_PROP_END_OF_LIST(), 2489}; 2490 2491static void vmxnet3_realize(DeviceState *qdev, Error **errp) 2492{ 2493 VMXNET3Class *vc = VMXNET3_DEVICE_GET_CLASS(qdev); 2494 PCIDevice *pci_dev = PCI_DEVICE(qdev); 2495 VMXNET3State *s = VMXNET3(qdev); 2496 2497 if (!(s->compat_flags & VMXNET3_COMPAT_FLAG_DISABLE_PCIE)) { 2498 pci_dev->cap_present |= QEMU_PCI_CAP_EXPRESS; 2499 } 2500 2501 vc->parent_dc_realize(qdev, errp); 2502} 2503 2504static void vmxnet3_class_init(ObjectClass *class, void *data) 2505{ 2506 DeviceClass *dc = DEVICE_CLASS(class); 2507 PCIDeviceClass *c = PCI_DEVICE_CLASS(class); 2508 VMXNET3Class *vc = VMXNET3_DEVICE_CLASS(class); 2509 2510 c->realize = vmxnet3_pci_realize; 2511 c->exit = vmxnet3_pci_uninit; 2512 c->vendor_id = PCI_VENDOR_ID_VMWARE; 2513 c->device_id = PCI_DEVICE_ID_VMWARE_VMXNET3; 2514 c->revision = PCI_DEVICE_ID_VMWARE_VMXNET3_REVISION; 2515 c->romfile = "efi-vmxnet3.rom"; 2516 c->class_id = PCI_CLASS_NETWORK_ETHERNET; 2517 c->subsystem_vendor_id = PCI_VENDOR_ID_VMWARE; 2518 c->subsystem_id = PCI_DEVICE_ID_VMWARE_VMXNET3; 2519 device_class_set_parent_realize(dc, vmxnet3_realize, 2520 &vc->parent_dc_realize); 2521 dc->desc = "VMWare Paravirtualized Ethernet v3"; 2522 dc->reset = vmxnet3_qdev_reset; 2523 dc->vmsd = &vmstate_vmxnet3; 2524 device_class_set_props(dc, vmxnet3_properties); 2525 set_bit(DEVICE_CATEGORY_NETWORK, dc->categories); 2526} 2527 2528static const TypeInfo vmxnet3_info = { 2529 .name = TYPE_VMXNET3, 2530 .parent = TYPE_PCI_DEVICE, 2531 .class_size = sizeof(VMXNET3Class), 2532 .instance_size = sizeof(VMXNET3State), 2533 .class_init = vmxnet3_class_init, 2534 .instance_init = vmxnet3_instance_init, 2535 .interfaces = (InterfaceInfo[]) { 2536 { INTERFACE_PCIE_DEVICE }, 2537 { INTERFACE_CONVENTIONAL_PCI_DEVICE }, 2538 { } 2539 }, 2540}; 2541 2542static void vmxnet3_register_types(void) 2543{ 2544 VMW_CBPRN("vmxnet3_register_types called..."); 2545 type_register_static(&vmxnet3_info); 2546} 2547 2548type_init(vmxnet3_register_types)