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 "hw/hw.h" 19 #include "hw/pci/pci.h" 20 #include "net/net.h" 21 #include "net/tap.h" 22 #include "net/checksum.h" 23 #include "sysemu/sysemu.h" 24 #include "qemu-common.h" 25 #include "qemu/bswap.h" 26 #include "hw/pci/msix.h" 27 #include "hw/pci/msi.h" 28 29 #include "vmxnet3.h" 30 #include "vmxnet_debug.h" 31 #include "vmware_utils.h" 32 #include "vmxnet_tx_pkt.h" 33 #include "vmxnet_rx_pkt.h" 34 35 #define PCI_DEVICE_ID_VMWARE_VMXNET3_REVISION 0x1 36 #define VMXNET3_MSIX_BAR_SIZE 0x2000 37 #define MIN_BUF_SIZE 60 38 39 #define VMXNET3_BAR0_IDX (0) 40 #define VMXNET3_BAR1_IDX (1) 41 #define VMXNET3_MSIX_BAR_IDX (2) 42 43 #define VMXNET3_OFF_MSIX_TABLE (0x000) 44 #define VMXNET3_OFF_MSIX_PBA (0x800) 45 46 /* Link speed in Mbps should be shifted by 16 */ 47 #define VMXNET3_LINK_SPEED (1000 << 16) 48 49 /* Link status: 1 - up, 0 - down. */ 50 #define VMXNET3_LINK_STATUS_UP 0x1 51 52 /* Least significant bit should be set for revision and version */ 53 #define VMXNET3_DEVICE_VERSION 0x1 54 #define VMXNET3_DEVICE_REVISION 0x1 55 56 /* Number of interrupt vectors for non-MSIx modes */ 57 #define VMXNET3_MAX_NMSIX_INTRS (1) 58 59 /* Macros for rings descriptors access */ 60 #define VMXNET3_READ_TX_QUEUE_DESCR8(dpa, field) \ 61 (vmw_shmem_ld8(dpa + offsetof(struct Vmxnet3_TxQueueDesc, field))) 62 63 #define VMXNET3_WRITE_TX_QUEUE_DESCR8(dpa, field, value) \ 64 (vmw_shmem_st8(dpa + offsetof(struct Vmxnet3_TxQueueDesc, field, value))) 65 66 #define VMXNET3_READ_TX_QUEUE_DESCR32(dpa, field) \ 67 (vmw_shmem_ld32(dpa + offsetof(struct Vmxnet3_TxQueueDesc, field))) 68 69 #define VMXNET3_WRITE_TX_QUEUE_DESCR32(dpa, field, value) \ 70 (vmw_shmem_st32(dpa + offsetof(struct Vmxnet3_TxQueueDesc, field), value)) 71 72 #define VMXNET3_READ_TX_QUEUE_DESCR64(dpa, field) \ 73 (vmw_shmem_ld64(dpa + offsetof(struct Vmxnet3_TxQueueDesc, field))) 74 75 #define VMXNET3_WRITE_TX_QUEUE_DESCR64(dpa, field, value) \ 76 (vmw_shmem_st64(dpa + offsetof(struct Vmxnet3_TxQueueDesc, field), value)) 77 78 #define VMXNET3_READ_RX_QUEUE_DESCR64(dpa, field) \ 79 (vmw_shmem_ld64(dpa + offsetof(struct Vmxnet3_RxQueueDesc, field))) 80 81 #define VMXNET3_READ_RX_QUEUE_DESCR32(dpa, field) \ 82 (vmw_shmem_ld32(dpa + offsetof(struct Vmxnet3_RxQueueDesc, field))) 83 84 #define VMXNET3_WRITE_RX_QUEUE_DESCR64(dpa, field, value) \ 85 (vmw_shmem_st64(dpa + offsetof(struct Vmxnet3_RxQueueDesc, field), value)) 86 87 #define VMXNET3_WRITE_RX_QUEUE_DESCR8(dpa, field, value) \ 88 (vmw_shmem_st8(dpa + offsetof(struct Vmxnet3_RxQueueDesc, field), value)) 89 90 /* Macros for guest driver shared area access */ 91 #define VMXNET3_READ_DRV_SHARED64(shpa, field) \ 92 (vmw_shmem_ld64(shpa + offsetof(struct Vmxnet3_DriverShared, field))) 93 94 #define VMXNET3_READ_DRV_SHARED32(shpa, field) \ 95 (vmw_shmem_ld32(shpa + offsetof(struct Vmxnet3_DriverShared, field))) 96 97 #define VMXNET3_WRITE_DRV_SHARED32(shpa, field, val) \ 98 (vmw_shmem_st32(shpa + offsetof(struct Vmxnet3_DriverShared, field), val)) 99 100 #define VMXNET3_READ_DRV_SHARED16(shpa, field) \ 101 (vmw_shmem_ld16(shpa + offsetof(struct Vmxnet3_DriverShared, field))) 102 103 #define VMXNET3_READ_DRV_SHARED8(shpa, field) \ 104 (vmw_shmem_ld8(shpa + offsetof(struct Vmxnet3_DriverShared, field))) 105 106 #define VMXNET3_READ_DRV_SHARED(shpa, field, b, l) \ 107 (vmw_shmem_read(shpa + offsetof(struct Vmxnet3_DriverShared, field), b, l)) 108 109 #define VMXNET_FLAG_IS_SET(field, flag) (((field) & (flag)) == (flag)) 110 111 #define TYPE_VMXNET3 "vmxnet3" 112 #define VMXNET3(obj) OBJECT_CHECK(VMXNET3State, (obj), TYPE_VMXNET3) 113 114 /* Cyclic ring abstraction */ 115 typedef struct { 116 hwaddr pa; 117 size_t size; 118 size_t cell_size; 119 size_t next; 120 uint8_t gen; 121 } Vmxnet3Ring; 122 123 static inline void vmxnet3_ring_init(Vmxnet3Ring *ring, 124 hwaddr pa, 125 size_t size, 126 size_t cell_size, 127 bool zero_region) 128 { 129 ring->pa = pa; 130 ring->size = size; 131 ring->cell_size = cell_size; 132 ring->gen = VMXNET3_INIT_GEN; 133 ring->next = 0; 134 135 if (zero_region) { 136 vmw_shmem_set(pa, 0, size * cell_size); 137 } 138 } 139 140 #define VMXNET3_RING_DUMP(macro, ring_name, ridx, r) \ 141 macro("%s#%d: base %" PRIx64 " size %lu cell_size %lu gen %d next %lu", \ 142 (ring_name), (ridx), \ 143 (r)->pa, (r)->size, (r)->cell_size, (r)->gen, (r)->next) 144 145 static inline void vmxnet3_ring_inc(Vmxnet3Ring *ring) 146 { 147 if (++ring->next >= ring->size) { 148 ring->next = 0; 149 ring->gen ^= 1; 150 } 151 } 152 153 static inline void vmxnet3_ring_dec(Vmxnet3Ring *ring) 154 { 155 if (ring->next-- == 0) { 156 ring->next = ring->size - 1; 157 ring->gen ^= 1; 158 } 159 } 160 161 static inline hwaddr vmxnet3_ring_curr_cell_pa(Vmxnet3Ring *ring) 162 { 163 return ring->pa + ring->next * ring->cell_size; 164 } 165 166 static inline void vmxnet3_ring_read_curr_cell(Vmxnet3Ring *ring, void *buff) 167 { 168 vmw_shmem_read(vmxnet3_ring_curr_cell_pa(ring), buff, ring->cell_size); 169 } 170 171 static inline void vmxnet3_ring_write_curr_cell(Vmxnet3Ring *ring, void *buff) 172 { 173 vmw_shmem_write(vmxnet3_ring_curr_cell_pa(ring), buff, ring->cell_size); 174 } 175 176 static inline size_t vmxnet3_ring_curr_cell_idx(Vmxnet3Ring *ring) 177 { 178 return ring->next; 179 } 180 181 static inline uint8_t vmxnet3_ring_curr_gen(Vmxnet3Ring *ring) 182 { 183 return ring->gen; 184 } 185 186 /* Debug trace-related functions */ 187 static inline void 188 vmxnet3_dump_tx_descr(struct Vmxnet3_TxDesc *descr) 189 { 190 VMW_PKPRN("TX DESCR: " 191 "addr %" PRIx64 ", len: %d, gen: %d, rsvd: %d, " 192 "dtype: %d, ext1: %d, msscof: %d, hlen: %d, om: %d, " 193 "eop: %d, cq: %d, ext2: %d, ti: %d, tci: %d", 194 le64_to_cpu(descr->addr), descr->len, descr->gen, descr->rsvd, 195 descr->dtype, descr->ext1, descr->msscof, descr->hlen, descr->om, 196 descr->eop, descr->cq, descr->ext2, descr->ti, descr->tci); 197 } 198 199 static inline void 200 vmxnet3_dump_virt_hdr(struct virtio_net_hdr *vhdr) 201 { 202 VMW_PKPRN("VHDR: flags 0x%x, gso_type: 0x%x, hdr_len: %d, gso_size: %d, " 203 "csum_start: %d, csum_offset: %d", 204 vhdr->flags, vhdr->gso_type, vhdr->hdr_len, vhdr->gso_size, 205 vhdr->csum_start, vhdr->csum_offset); 206 } 207 208 static inline void 209 vmxnet3_dump_rx_descr(struct Vmxnet3_RxDesc *descr) 210 { 211 VMW_PKPRN("RX DESCR: addr %" PRIx64 ", len: %d, gen: %d, rsvd: %d, " 212 "dtype: %d, ext1: %d, btype: %d", 213 le64_to_cpu(descr->addr), descr->len, descr->gen, 214 descr->rsvd, descr->dtype, descr->ext1, descr->btype); 215 } 216 217 /* Device state and helper functions */ 218 #define VMXNET3_RX_RINGS_PER_QUEUE (2) 219 220 typedef struct { 221 Vmxnet3Ring tx_ring; 222 Vmxnet3Ring comp_ring; 223 224 uint8_t intr_idx; 225 hwaddr tx_stats_pa; 226 struct UPT1_TxStats txq_stats; 227 } Vmxnet3TxqDescr; 228 229 typedef struct { 230 Vmxnet3Ring rx_ring[VMXNET3_RX_RINGS_PER_QUEUE]; 231 Vmxnet3Ring comp_ring; 232 uint8_t intr_idx; 233 hwaddr rx_stats_pa; 234 struct UPT1_RxStats rxq_stats; 235 } Vmxnet3RxqDescr; 236 237 typedef struct { 238 bool is_masked; 239 bool is_pending; 240 bool is_asserted; 241 } Vmxnet3IntState; 242 243 typedef struct { 244 PCIDevice parent_obj; 245 NICState *nic; 246 NICConf conf; 247 MemoryRegion bar0; 248 MemoryRegion bar1; 249 MemoryRegion msix_bar; 250 251 Vmxnet3RxqDescr rxq_descr[VMXNET3_DEVICE_MAX_RX_QUEUES]; 252 Vmxnet3TxqDescr txq_descr[VMXNET3_DEVICE_MAX_TX_QUEUES]; 253 254 /* Whether MSI-X support was installed successfully */ 255 bool msix_used; 256 /* Whether MSI support was installed successfully */ 257 bool msi_used; 258 hwaddr drv_shmem; 259 hwaddr temp_shared_guest_driver_memory; 260 261 uint8_t txq_num; 262 263 /* This boolean tells whether RX packet being indicated has to */ 264 /* be split into head and body chunks from different RX rings */ 265 bool rx_packets_compound; 266 267 bool rx_vlan_stripping; 268 bool lro_supported; 269 270 uint8_t rxq_num; 271 272 /* Network MTU */ 273 uint32_t mtu; 274 275 /* Maximum number of fragments for indicated TX packets */ 276 uint32_t max_tx_frags; 277 278 /* Maximum number of fragments for indicated RX packets */ 279 uint16_t max_rx_frags; 280 281 /* Index for events interrupt */ 282 uint8_t event_int_idx; 283 284 /* Whether automatic interrupts masking enabled */ 285 bool auto_int_masking; 286 287 bool peer_has_vhdr; 288 289 /* TX packets to QEMU interface */ 290 struct VmxnetTxPkt *tx_pkt; 291 uint32_t offload_mode; 292 uint32_t cso_or_gso_size; 293 uint16_t tci; 294 bool needs_vlan; 295 296 struct VmxnetRxPkt *rx_pkt; 297 298 bool tx_sop; 299 bool skip_current_tx_pkt; 300 301 uint32_t device_active; 302 uint32_t last_command; 303 304 uint32_t link_status_and_speed; 305 306 Vmxnet3IntState interrupt_states[VMXNET3_MAX_INTRS]; 307 308 uint32_t temp_mac; /* To store the low part first */ 309 310 MACAddr perm_mac; 311 uint32_t vlan_table[VMXNET3_VFT_SIZE]; 312 uint32_t rx_mode; 313 MACAddr *mcast_list; 314 uint32_t mcast_list_len; 315 uint32_t mcast_list_buff_size; /* needed for live migration. */ 316 } VMXNET3State; 317 318 /* Interrupt management */ 319 320 /* 321 *This function returns sign whether interrupt line is in asserted state 322 * This depends on the type of interrupt used. For INTX interrupt line will 323 * be asserted until explicit deassertion, for MSI(X) interrupt line will 324 * be deasserted automatically due to notification semantics of the MSI(X) 325 * interrupts 326 */ 327 static bool _vmxnet3_assert_interrupt_line(VMXNET3State *s, uint32_t int_idx) 328 { 329 PCIDevice *d = PCI_DEVICE(s); 330 331 if (s->msix_used && msix_enabled(d)) { 332 VMW_IRPRN("Sending MSI-X notification for vector %u", int_idx); 333 msix_notify(d, int_idx); 334 return false; 335 } 336 if (s->msi_used && msi_enabled(d)) { 337 VMW_IRPRN("Sending MSI notification for vector %u", int_idx); 338 msi_notify(d, int_idx); 339 return false; 340 } 341 342 VMW_IRPRN("Asserting line for interrupt %u", int_idx); 343 pci_irq_assert(d); 344 return true; 345 } 346 347 static void _vmxnet3_deassert_interrupt_line(VMXNET3State *s, int lidx) 348 { 349 PCIDevice *d = PCI_DEVICE(s); 350 351 /* 352 * This function should never be called for MSI(X) interrupts 353 * because deassertion never required for message interrupts 354 */ 355 assert(!s->msix_used || !msix_enabled(d)); 356 /* 357 * This function should never be called for MSI(X) interrupts 358 * because deassertion never required for message interrupts 359 */ 360 assert(!s->msi_used || !msi_enabled(d)); 361 362 VMW_IRPRN("Deasserting line for interrupt %u", lidx); 363 pci_irq_deassert(d); 364 } 365 366 static void vmxnet3_update_interrupt_line_state(VMXNET3State *s, int lidx) 367 { 368 if (!s->interrupt_states[lidx].is_pending && 369 s->interrupt_states[lidx].is_asserted) { 370 VMW_IRPRN("New interrupt line state for index %d is DOWN", lidx); 371 _vmxnet3_deassert_interrupt_line(s, lidx); 372 s->interrupt_states[lidx].is_asserted = false; 373 return; 374 } 375 376 if (s->interrupt_states[lidx].is_pending && 377 !s->interrupt_states[lidx].is_masked && 378 !s->interrupt_states[lidx].is_asserted) { 379 VMW_IRPRN("New interrupt line state for index %d is UP", lidx); 380 s->interrupt_states[lidx].is_asserted = 381 _vmxnet3_assert_interrupt_line(s, lidx); 382 s->interrupt_states[lidx].is_pending = false; 383 return; 384 } 385 } 386 387 static void vmxnet3_trigger_interrupt(VMXNET3State *s, int lidx) 388 { 389 PCIDevice *d = PCI_DEVICE(s); 390 s->interrupt_states[lidx].is_pending = true; 391 vmxnet3_update_interrupt_line_state(s, lidx); 392 393 if (s->msix_used && msix_enabled(d) && s->auto_int_masking) { 394 goto do_automask; 395 } 396 397 if (s->msi_used && msi_enabled(d) && s->auto_int_masking) { 398 goto do_automask; 399 } 400 401 return; 402 403 do_automask: 404 s->interrupt_states[lidx].is_masked = true; 405 vmxnet3_update_interrupt_line_state(s, lidx); 406 } 407 408 static bool vmxnet3_interrupt_asserted(VMXNET3State *s, int lidx) 409 { 410 return s->interrupt_states[lidx].is_asserted; 411 } 412 413 static void vmxnet3_clear_interrupt(VMXNET3State *s, int int_idx) 414 { 415 s->interrupt_states[int_idx].is_pending = false; 416 if (s->auto_int_masking) { 417 s->interrupt_states[int_idx].is_masked = true; 418 } 419 vmxnet3_update_interrupt_line_state(s, int_idx); 420 } 421 422 static void 423 vmxnet3_on_interrupt_mask_changed(VMXNET3State *s, int lidx, bool is_masked) 424 { 425 s->interrupt_states[lidx].is_masked = is_masked; 426 vmxnet3_update_interrupt_line_state(s, lidx); 427 } 428 429 static bool vmxnet3_verify_driver_magic(hwaddr dshmem) 430 { 431 return (VMXNET3_READ_DRV_SHARED32(dshmem, magic) == VMXNET3_REV1_MAGIC); 432 } 433 434 #define VMXNET3_GET_BYTE(x, byte_num) (((x) >> (byte_num)*8) & 0xFF) 435 #define VMXNET3_MAKE_BYTE(byte_num, val) \ 436 (((uint32_t)((val) & 0xFF)) << (byte_num)*8) 437 438 static void vmxnet3_set_variable_mac(VMXNET3State *s, uint32_t h, uint32_t l) 439 { 440 s->conf.macaddr.a[0] = VMXNET3_GET_BYTE(l, 0); 441 s->conf.macaddr.a[1] = VMXNET3_GET_BYTE(l, 1); 442 s->conf.macaddr.a[2] = VMXNET3_GET_BYTE(l, 2); 443 s->conf.macaddr.a[3] = VMXNET3_GET_BYTE(l, 3); 444 s->conf.macaddr.a[4] = VMXNET3_GET_BYTE(h, 0); 445 s->conf.macaddr.a[5] = VMXNET3_GET_BYTE(h, 1); 446 447 VMW_CFPRN("Variable MAC: " VMXNET_MF, VMXNET_MA(s->conf.macaddr.a)); 448 449 qemu_format_nic_info_str(qemu_get_queue(s->nic), s->conf.macaddr.a); 450 } 451 452 static uint64_t vmxnet3_get_mac_low(MACAddr *addr) 453 { 454 return VMXNET3_MAKE_BYTE(0, addr->a[0]) | 455 VMXNET3_MAKE_BYTE(1, addr->a[1]) | 456 VMXNET3_MAKE_BYTE(2, addr->a[2]) | 457 VMXNET3_MAKE_BYTE(3, addr->a[3]); 458 } 459 460 static uint64_t vmxnet3_get_mac_high(MACAddr *addr) 461 { 462 return VMXNET3_MAKE_BYTE(0, addr->a[4]) | 463 VMXNET3_MAKE_BYTE(1, addr->a[5]); 464 } 465 466 static void 467 vmxnet3_inc_tx_consumption_counter(VMXNET3State *s, int qidx) 468 { 469 vmxnet3_ring_inc(&s->txq_descr[qidx].tx_ring); 470 } 471 472 static inline void 473 vmxnet3_inc_rx_consumption_counter(VMXNET3State *s, int qidx, int ridx) 474 { 475 vmxnet3_ring_inc(&s->rxq_descr[qidx].rx_ring[ridx]); 476 } 477 478 static inline void 479 vmxnet3_inc_tx_completion_counter(VMXNET3State *s, int qidx) 480 { 481 vmxnet3_ring_inc(&s->txq_descr[qidx].comp_ring); 482 } 483 484 static void 485 vmxnet3_inc_rx_completion_counter(VMXNET3State *s, int qidx) 486 { 487 vmxnet3_ring_inc(&s->rxq_descr[qidx].comp_ring); 488 } 489 490 static void 491 vmxnet3_dec_rx_completion_counter(VMXNET3State *s, int qidx) 492 { 493 vmxnet3_ring_dec(&s->rxq_descr[qidx].comp_ring); 494 } 495 496 static void vmxnet3_complete_packet(VMXNET3State *s, int qidx, uint32 tx_ridx) 497 { 498 struct Vmxnet3_TxCompDesc txcq_descr; 499 500 VMXNET3_RING_DUMP(VMW_RIPRN, "TXC", qidx, &s->txq_descr[qidx].comp_ring); 501 502 txcq_descr.txdIdx = tx_ridx; 503 txcq_descr.gen = vmxnet3_ring_curr_gen(&s->txq_descr[qidx].comp_ring); 504 505 vmxnet3_ring_write_curr_cell(&s->txq_descr[qidx].comp_ring, &txcq_descr); 506 507 /* Flush changes in TX descriptor before changing the counter value */ 508 smp_wmb(); 509 510 vmxnet3_inc_tx_completion_counter(s, qidx); 511 vmxnet3_trigger_interrupt(s, s->txq_descr[qidx].intr_idx); 512 } 513 514 static bool 515 vmxnet3_setup_tx_offloads(VMXNET3State *s) 516 { 517 switch (s->offload_mode) { 518 case VMXNET3_OM_NONE: 519 vmxnet_tx_pkt_build_vheader(s->tx_pkt, false, false, 0); 520 break; 521 522 case VMXNET3_OM_CSUM: 523 vmxnet_tx_pkt_build_vheader(s->tx_pkt, false, true, 0); 524 VMW_PKPRN("L4 CSO requested\n"); 525 break; 526 527 case VMXNET3_OM_TSO: 528 vmxnet_tx_pkt_build_vheader(s->tx_pkt, true, true, 529 s->cso_or_gso_size); 530 vmxnet_tx_pkt_update_ip_checksums(s->tx_pkt); 531 VMW_PKPRN("GSO offload requested."); 532 break; 533 534 default: 535 g_assert_not_reached(); 536 return false; 537 } 538 539 return true; 540 } 541 542 static void 543 vmxnet3_tx_retrieve_metadata(VMXNET3State *s, 544 const struct Vmxnet3_TxDesc *txd) 545 { 546 s->offload_mode = txd->om; 547 s->cso_or_gso_size = txd->msscof; 548 s->tci = txd->tci; 549 s->needs_vlan = txd->ti; 550 } 551 552 typedef enum { 553 VMXNET3_PKT_STATUS_OK, 554 VMXNET3_PKT_STATUS_ERROR, 555 VMXNET3_PKT_STATUS_DISCARD,/* only for tx */ 556 VMXNET3_PKT_STATUS_OUT_OF_BUF /* only for rx */ 557 } Vmxnet3PktStatus; 558 559 static void 560 vmxnet3_on_tx_done_update_stats(VMXNET3State *s, int qidx, 561 Vmxnet3PktStatus status) 562 { 563 size_t tot_len = vmxnet_tx_pkt_get_total_len(s->tx_pkt); 564 struct UPT1_TxStats *stats = &s->txq_descr[qidx].txq_stats; 565 566 switch (status) { 567 case VMXNET3_PKT_STATUS_OK: 568 switch (vmxnet_tx_pkt_get_packet_type(s->tx_pkt)) { 569 case ETH_PKT_BCAST: 570 stats->bcastPktsTxOK++; 571 stats->bcastBytesTxOK += tot_len; 572 break; 573 case ETH_PKT_MCAST: 574 stats->mcastPktsTxOK++; 575 stats->mcastBytesTxOK += tot_len; 576 break; 577 case ETH_PKT_UCAST: 578 stats->ucastPktsTxOK++; 579 stats->ucastBytesTxOK += tot_len; 580 break; 581 default: 582 g_assert_not_reached(); 583 } 584 585 if (s->offload_mode == VMXNET3_OM_TSO) { 586 /* 587 * According to VMWARE headers this statistic is a number 588 * of packets after segmentation but since we don't have 589 * this information in QEMU model, the best we can do is to 590 * provide number of non-segmented packets 591 */ 592 stats->TSOPktsTxOK++; 593 stats->TSOBytesTxOK += tot_len; 594 } 595 break; 596 597 case VMXNET3_PKT_STATUS_DISCARD: 598 stats->pktsTxDiscard++; 599 break; 600 601 case VMXNET3_PKT_STATUS_ERROR: 602 stats->pktsTxError++; 603 break; 604 605 default: 606 g_assert_not_reached(); 607 } 608 } 609 610 static void 611 vmxnet3_on_rx_done_update_stats(VMXNET3State *s, 612 int qidx, 613 Vmxnet3PktStatus status) 614 { 615 struct UPT1_RxStats *stats = &s->rxq_descr[qidx].rxq_stats; 616 size_t tot_len = vmxnet_rx_pkt_get_total_len(s->rx_pkt); 617 618 switch (status) { 619 case VMXNET3_PKT_STATUS_OUT_OF_BUF: 620 stats->pktsRxOutOfBuf++; 621 break; 622 623 case VMXNET3_PKT_STATUS_ERROR: 624 stats->pktsRxError++; 625 break; 626 case VMXNET3_PKT_STATUS_OK: 627 switch (vmxnet_rx_pkt_get_packet_type(s->rx_pkt)) { 628 case ETH_PKT_BCAST: 629 stats->bcastPktsRxOK++; 630 stats->bcastBytesRxOK += tot_len; 631 break; 632 case ETH_PKT_MCAST: 633 stats->mcastPktsRxOK++; 634 stats->mcastBytesRxOK += tot_len; 635 break; 636 case ETH_PKT_UCAST: 637 stats->ucastPktsRxOK++; 638 stats->ucastBytesRxOK += tot_len; 639 break; 640 default: 641 g_assert_not_reached(); 642 } 643 644 if (tot_len > s->mtu) { 645 stats->LROPktsRxOK++; 646 stats->LROBytesRxOK += tot_len; 647 } 648 break; 649 default: 650 g_assert_not_reached(); 651 } 652 } 653 654 static inline bool 655 vmxnet3_pop_next_tx_descr(VMXNET3State *s, 656 int qidx, 657 struct Vmxnet3_TxDesc *txd, 658 uint32_t *descr_idx) 659 { 660 Vmxnet3Ring *ring = &s->txq_descr[qidx].tx_ring; 661 662 vmxnet3_ring_read_curr_cell(ring, txd); 663 if (txd->gen == vmxnet3_ring_curr_gen(ring)) { 664 /* Only read after generation field verification */ 665 smp_rmb(); 666 /* Re-read to be sure we got the latest version */ 667 vmxnet3_ring_read_curr_cell(ring, txd); 668 VMXNET3_RING_DUMP(VMW_RIPRN, "TX", qidx, ring); 669 *descr_idx = vmxnet3_ring_curr_cell_idx(ring); 670 vmxnet3_inc_tx_consumption_counter(s, qidx); 671 return true; 672 } 673 674 return false; 675 } 676 677 static bool 678 vmxnet3_send_packet(VMXNET3State *s, uint32_t qidx) 679 { 680 Vmxnet3PktStatus status = VMXNET3_PKT_STATUS_OK; 681 682 if (!vmxnet3_setup_tx_offloads(s)) { 683 status = VMXNET3_PKT_STATUS_ERROR; 684 goto func_exit; 685 } 686 687 /* debug prints */ 688 vmxnet3_dump_virt_hdr(vmxnet_tx_pkt_get_vhdr(s->tx_pkt)); 689 vmxnet_tx_pkt_dump(s->tx_pkt); 690 691 if (!vmxnet_tx_pkt_send(s->tx_pkt, qemu_get_queue(s->nic))) { 692 status = VMXNET3_PKT_STATUS_DISCARD; 693 goto func_exit; 694 } 695 696 func_exit: 697 vmxnet3_on_tx_done_update_stats(s, qidx, status); 698 return (status == VMXNET3_PKT_STATUS_OK); 699 } 700 701 static void vmxnet3_process_tx_queue(VMXNET3State *s, int qidx) 702 { 703 struct Vmxnet3_TxDesc txd; 704 uint32_t txd_idx; 705 uint32_t data_len; 706 hwaddr data_pa; 707 708 for (;;) { 709 if (!vmxnet3_pop_next_tx_descr(s, qidx, &txd, &txd_idx)) { 710 break; 711 } 712 713 vmxnet3_dump_tx_descr(&txd); 714 715 if (!s->skip_current_tx_pkt) { 716 data_len = (txd.len > 0) ? txd.len : VMXNET3_MAX_TX_BUF_SIZE; 717 data_pa = le64_to_cpu(txd.addr); 718 719 if (!vmxnet_tx_pkt_add_raw_fragment(s->tx_pkt, 720 data_pa, 721 data_len)) { 722 s->skip_current_tx_pkt = true; 723 } 724 } 725 726 if (s->tx_sop) { 727 vmxnet3_tx_retrieve_metadata(s, &txd); 728 s->tx_sop = false; 729 } 730 731 if (txd.eop) { 732 if (!s->skip_current_tx_pkt && vmxnet_tx_pkt_parse(s->tx_pkt)) { 733 if (s->needs_vlan) { 734 vmxnet_tx_pkt_setup_vlan_header(s->tx_pkt, s->tci); 735 } 736 737 vmxnet3_send_packet(s, qidx); 738 } else { 739 vmxnet3_on_tx_done_update_stats(s, qidx, 740 VMXNET3_PKT_STATUS_ERROR); 741 } 742 743 vmxnet3_complete_packet(s, qidx, txd_idx); 744 s->tx_sop = true; 745 s->skip_current_tx_pkt = false; 746 vmxnet_tx_pkt_reset(s->tx_pkt); 747 } 748 } 749 } 750 751 static inline void 752 vmxnet3_read_next_rx_descr(VMXNET3State *s, int qidx, int ridx, 753 struct Vmxnet3_RxDesc *dbuf, uint32_t *didx) 754 { 755 Vmxnet3Ring *ring = &s->rxq_descr[qidx].rx_ring[ridx]; 756 *didx = vmxnet3_ring_curr_cell_idx(ring); 757 vmxnet3_ring_read_curr_cell(ring, dbuf); 758 } 759 760 static inline uint8_t 761 vmxnet3_get_rx_ring_gen(VMXNET3State *s, int qidx, int ridx) 762 { 763 return s->rxq_descr[qidx].rx_ring[ridx].gen; 764 } 765 766 static inline hwaddr 767 vmxnet3_pop_rxc_descr(VMXNET3State *s, int qidx, uint32_t *descr_gen) 768 { 769 uint8_t ring_gen; 770 struct Vmxnet3_RxCompDesc rxcd; 771 772 hwaddr daddr = 773 vmxnet3_ring_curr_cell_pa(&s->rxq_descr[qidx].comp_ring); 774 775 cpu_physical_memory_read(daddr, &rxcd, sizeof(struct Vmxnet3_RxCompDesc)); 776 ring_gen = vmxnet3_ring_curr_gen(&s->rxq_descr[qidx].comp_ring); 777 778 if (rxcd.gen != ring_gen) { 779 *descr_gen = ring_gen; 780 vmxnet3_inc_rx_completion_counter(s, qidx); 781 return daddr; 782 } 783 784 return 0; 785 } 786 787 static inline void 788 vmxnet3_revert_rxc_descr(VMXNET3State *s, int qidx) 789 { 790 vmxnet3_dec_rx_completion_counter(s, qidx); 791 } 792 793 #define RXQ_IDX (0) 794 #define RX_HEAD_BODY_RING (0) 795 #define RX_BODY_ONLY_RING (1) 796 797 static bool 798 vmxnet3_get_next_head_rx_descr(VMXNET3State *s, 799 struct Vmxnet3_RxDesc *descr_buf, 800 uint32_t *descr_idx, 801 uint32_t *ridx) 802 { 803 for (;;) { 804 uint32_t ring_gen; 805 vmxnet3_read_next_rx_descr(s, RXQ_IDX, RX_HEAD_BODY_RING, 806 descr_buf, descr_idx); 807 808 /* If no more free descriptors - return */ 809 ring_gen = vmxnet3_get_rx_ring_gen(s, RXQ_IDX, RX_HEAD_BODY_RING); 810 if (descr_buf->gen != ring_gen) { 811 return false; 812 } 813 814 /* Only read after generation field verification */ 815 smp_rmb(); 816 /* Re-read to be sure we got the latest version */ 817 vmxnet3_read_next_rx_descr(s, RXQ_IDX, RX_HEAD_BODY_RING, 818 descr_buf, descr_idx); 819 820 /* Mark current descriptor as used/skipped */ 821 vmxnet3_inc_rx_consumption_counter(s, RXQ_IDX, RX_HEAD_BODY_RING); 822 823 /* If this is what we are looking for - return */ 824 if (descr_buf->btype == VMXNET3_RXD_BTYPE_HEAD) { 825 *ridx = RX_HEAD_BODY_RING; 826 return true; 827 } 828 } 829 } 830 831 static bool 832 vmxnet3_get_next_body_rx_descr(VMXNET3State *s, 833 struct Vmxnet3_RxDesc *d, 834 uint32_t *didx, 835 uint32_t *ridx) 836 { 837 vmxnet3_read_next_rx_descr(s, RXQ_IDX, RX_HEAD_BODY_RING, d, didx); 838 839 /* Try to find corresponding descriptor in head/body ring */ 840 if (d->gen == vmxnet3_get_rx_ring_gen(s, RXQ_IDX, RX_HEAD_BODY_RING)) { 841 /* Only read after generation field verification */ 842 smp_rmb(); 843 /* Re-read to be sure we got the latest version */ 844 vmxnet3_read_next_rx_descr(s, RXQ_IDX, RX_HEAD_BODY_RING, d, didx); 845 if (d->btype == VMXNET3_RXD_BTYPE_BODY) { 846 vmxnet3_inc_rx_consumption_counter(s, RXQ_IDX, RX_HEAD_BODY_RING); 847 *ridx = RX_HEAD_BODY_RING; 848 return true; 849 } 850 } 851 852 /* 853 * If there is no free descriptors on head/body ring or next free 854 * descriptor is a head descriptor switch to body only ring 855 */ 856 vmxnet3_read_next_rx_descr(s, RXQ_IDX, RX_BODY_ONLY_RING, d, didx); 857 858 /* If no more free descriptors - return */ 859 if (d->gen == vmxnet3_get_rx_ring_gen(s, RXQ_IDX, RX_BODY_ONLY_RING)) { 860 /* Only read after generation field verification */ 861 smp_rmb(); 862 /* Re-read to be sure we got the latest version */ 863 vmxnet3_read_next_rx_descr(s, RXQ_IDX, RX_BODY_ONLY_RING, d, didx); 864 assert(d->btype == VMXNET3_RXD_BTYPE_BODY); 865 *ridx = RX_BODY_ONLY_RING; 866 vmxnet3_inc_rx_consumption_counter(s, RXQ_IDX, RX_BODY_ONLY_RING); 867 return true; 868 } 869 870 return false; 871 } 872 873 static inline bool 874 vmxnet3_get_next_rx_descr(VMXNET3State *s, bool is_head, 875 struct Vmxnet3_RxDesc *descr_buf, 876 uint32_t *descr_idx, 877 uint32_t *ridx) 878 { 879 if (is_head || !s->rx_packets_compound) { 880 return vmxnet3_get_next_head_rx_descr(s, descr_buf, descr_idx, ridx); 881 } else { 882 return vmxnet3_get_next_body_rx_descr(s, descr_buf, descr_idx, ridx); 883 } 884 } 885 886 /* In case packet was csum offloaded (either NEEDS_CSUM or DATA_VALID), 887 * the implementation always passes an RxCompDesc with a "Checksum 888 * calculated and found correct" to the OS (cnc=0 and tuc=1, see 889 * vmxnet3_rx_update_descr). This emulates the observed ESXi behavior. 890 * 891 * Therefore, if packet has the NEEDS_CSUM set, we must calculate 892 * and place a fully computed checksum into the tcp/udp header. 893 * Otherwise, the OS driver will receive a checksum-correct indication 894 * (CHECKSUM_UNNECESSARY), but with the actual tcp/udp checksum field 895 * having just the pseudo header csum value. 896 * 897 * While this is not a problem if packet is destined for local delivery, 898 * in the case the host OS performs forwarding, it will forward an 899 * incorrectly checksummed packet. 900 */ 901 static void vmxnet3_rx_need_csum_calculate(struct VmxnetRxPkt *pkt, 902 const void *pkt_data, 903 size_t pkt_len) 904 { 905 struct virtio_net_hdr *vhdr; 906 bool isip4, isip6, istcp, isudp; 907 uint8_t *data; 908 int len; 909 910 if (!vmxnet_rx_pkt_has_virt_hdr(pkt)) { 911 return; 912 } 913 914 vhdr = vmxnet_rx_pkt_get_vhdr(pkt); 915 if (!VMXNET_FLAG_IS_SET(vhdr->flags, VIRTIO_NET_HDR_F_NEEDS_CSUM)) { 916 return; 917 } 918 919 vmxnet_rx_pkt_get_protocols(pkt, &isip4, &isip6, &isudp, &istcp); 920 if (!(isip4 || isip6) || !(istcp || isudp)) { 921 return; 922 } 923 924 vmxnet3_dump_virt_hdr(vhdr); 925 926 /* Validate packet len: csum_start + scum_offset + length of csum field */ 927 if (pkt_len < (vhdr->csum_start + vhdr->csum_offset + 2)) { 928 VMW_PKPRN("packet len:%lu < csum_start(%d) + csum_offset(%d) + 2, " 929 "cannot calculate checksum", 930 pkt_len, vhdr->csum_start, vhdr->csum_offset); 931 return; 932 } 933 934 data = (uint8_t *)pkt_data + vhdr->csum_start; 935 len = pkt_len - vhdr->csum_start; 936 /* Put the checksum obtained into the packet */ 937 stw_be_p(data + vhdr->csum_offset, net_raw_checksum(data, len)); 938 939 vhdr->flags &= ~VIRTIO_NET_HDR_F_NEEDS_CSUM; 940 vhdr->flags |= VIRTIO_NET_HDR_F_DATA_VALID; 941 } 942 943 static void vmxnet3_rx_update_descr(struct VmxnetRxPkt *pkt, 944 struct Vmxnet3_RxCompDesc *rxcd) 945 { 946 int csum_ok, is_gso; 947 bool isip4, isip6, istcp, isudp; 948 struct virtio_net_hdr *vhdr; 949 uint8_t offload_type; 950 951 if (vmxnet_rx_pkt_is_vlan_stripped(pkt)) { 952 rxcd->ts = 1; 953 rxcd->tci = vmxnet_rx_pkt_get_vlan_tag(pkt); 954 } 955 956 if (!vmxnet_rx_pkt_has_virt_hdr(pkt)) { 957 goto nocsum; 958 } 959 960 vhdr = vmxnet_rx_pkt_get_vhdr(pkt); 961 /* 962 * Checksum is valid when lower level tell so or when lower level 963 * requires checksum offload telling that packet produced/bridged 964 * locally and did travel over network after last checksum calculation 965 * or production 966 */ 967 csum_ok = VMXNET_FLAG_IS_SET(vhdr->flags, VIRTIO_NET_HDR_F_DATA_VALID) || 968 VMXNET_FLAG_IS_SET(vhdr->flags, VIRTIO_NET_HDR_F_NEEDS_CSUM); 969 970 offload_type = vhdr->gso_type & ~VIRTIO_NET_HDR_GSO_ECN; 971 is_gso = (offload_type != VIRTIO_NET_HDR_GSO_NONE) ? 1 : 0; 972 973 if (!csum_ok && !is_gso) { 974 goto nocsum; 975 } 976 977 vmxnet_rx_pkt_get_protocols(pkt, &isip4, &isip6, &isudp, &istcp); 978 if ((!istcp && !isudp) || (!isip4 && !isip6)) { 979 goto nocsum; 980 } 981 982 rxcd->cnc = 0; 983 rxcd->v4 = isip4 ? 1 : 0; 984 rxcd->v6 = isip6 ? 1 : 0; 985 rxcd->tcp = istcp ? 1 : 0; 986 rxcd->udp = isudp ? 1 : 0; 987 rxcd->fcs = rxcd->tuc = rxcd->ipc = 1; 988 return; 989 990 nocsum: 991 rxcd->cnc = 1; 992 return; 993 } 994 995 static void 996 vmxnet3_physical_memory_writev(const struct iovec *iov, 997 size_t start_iov_off, 998 hwaddr target_addr, 999 size_t bytes_to_copy) 1000 { 1001 size_t curr_off = 0; 1002 size_t copied = 0; 1003 1004 while (bytes_to_copy) { 1005 if (start_iov_off < (curr_off + iov->iov_len)) { 1006 size_t chunk_len = 1007 MIN((curr_off + iov->iov_len) - start_iov_off, bytes_to_copy); 1008 1009 cpu_physical_memory_write(target_addr + copied, 1010 iov->iov_base + start_iov_off - curr_off, 1011 chunk_len); 1012 1013 copied += chunk_len; 1014 start_iov_off += chunk_len; 1015 curr_off = start_iov_off; 1016 bytes_to_copy -= chunk_len; 1017 } else { 1018 curr_off += iov->iov_len; 1019 } 1020 iov++; 1021 } 1022 } 1023 1024 static bool 1025 vmxnet3_indicate_packet(VMXNET3State *s) 1026 { 1027 struct Vmxnet3_RxDesc rxd; 1028 bool is_head = true; 1029 uint32_t rxd_idx; 1030 uint32_t rx_ridx = 0; 1031 1032 struct Vmxnet3_RxCompDesc rxcd; 1033 uint32_t new_rxcd_gen = VMXNET3_INIT_GEN; 1034 hwaddr new_rxcd_pa = 0; 1035 hwaddr ready_rxcd_pa = 0; 1036 struct iovec *data = vmxnet_rx_pkt_get_iovec(s->rx_pkt); 1037 size_t bytes_copied = 0; 1038 size_t bytes_left = vmxnet_rx_pkt_get_total_len(s->rx_pkt); 1039 uint16_t num_frags = 0; 1040 size_t chunk_size; 1041 1042 vmxnet_rx_pkt_dump(s->rx_pkt); 1043 1044 while (bytes_left > 0) { 1045 1046 /* cannot add more frags to packet */ 1047 if (num_frags == s->max_rx_frags) { 1048 break; 1049 } 1050 1051 new_rxcd_pa = vmxnet3_pop_rxc_descr(s, RXQ_IDX, &new_rxcd_gen); 1052 if (!new_rxcd_pa) { 1053 break; 1054 } 1055 1056 if (!vmxnet3_get_next_rx_descr(s, is_head, &rxd, &rxd_idx, &rx_ridx)) { 1057 break; 1058 } 1059 1060 chunk_size = MIN(bytes_left, rxd.len); 1061 vmxnet3_physical_memory_writev(data, bytes_copied, 1062 le64_to_cpu(rxd.addr), chunk_size); 1063 bytes_copied += chunk_size; 1064 bytes_left -= chunk_size; 1065 1066 vmxnet3_dump_rx_descr(&rxd); 1067 1068 if (ready_rxcd_pa != 0) { 1069 cpu_physical_memory_write(ready_rxcd_pa, &rxcd, sizeof(rxcd)); 1070 } 1071 1072 memset(&rxcd, 0, sizeof(struct Vmxnet3_RxCompDesc)); 1073 rxcd.rxdIdx = rxd_idx; 1074 rxcd.len = chunk_size; 1075 rxcd.sop = is_head; 1076 rxcd.gen = new_rxcd_gen; 1077 rxcd.rqID = RXQ_IDX + rx_ridx * s->rxq_num; 1078 1079 if (bytes_left == 0) { 1080 vmxnet3_rx_update_descr(s->rx_pkt, &rxcd); 1081 } 1082 1083 VMW_RIPRN("RX Completion descriptor: rxRing: %lu rxIdx %lu len %lu " 1084 "sop %d csum_correct %lu", 1085 (unsigned long) rx_ridx, 1086 (unsigned long) rxcd.rxdIdx, 1087 (unsigned long) rxcd.len, 1088 (int) rxcd.sop, 1089 (unsigned long) rxcd.tuc); 1090 1091 is_head = false; 1092 ready_rxcd_pa = new_rxcd_pa; 1093 new_rxcd_pa = 0; 1094 num_frags++; 1095 } 1096 1097 if (ready_rxcd_pa != 0) { 1098 rxcd.eop = 1; 1099 rxcd.err = (bytes_left != 0); 1100 cpu_physical_memory_write(ready_rxcd_pa, &rxcd, sizeof(rxcd)); 1101 1102 /* Flush RX descriptor changes */ 1103 smp_wmb(); 1104 } 1105 1106 if (new_rxcd_pa != 0) { 1107 vmxnet3_revert_rxc_descr(s, RXQ_IDX); 1108 } 1109 1110 vmxnet3_trigger_interrupt(s, s->rxq_descr[RXQ_IDX].intr_idx); 1111 1112 if (bytes_left == 0) { 1113 vmxnet3_on_rx_done_update_stats(s, RXQ_IDX, VMXNET3_PKT_STATUS_OK); 1114 return true; 1115 } else if (num_frags == s->max_rx_frags) { 1116 vmxnet3_on_rx_done_update_stats(s, RXQ_IDX, VMXNET3_PKT_STATUS_ERROR); 1117 return false; 1118 } else { 1119 vmxnet3_on_rx_done_update_stats(s, RXQ_IDX, 1120 VMXNET3_PKT_STATUS_OUT_OF_BUF); 1121 return false; 1122 } 1123 } 1124 1125 static void 1126 vmxnet3_io_bar0_write(void *opaque, hwaddr addr, 1127 uint64_t val, unsigned size) 1128 { 1129 VMXNET3State *s = opaque; 1130 1131 if (VMW_IS_MULTIREG_ADDR(addr, VMXNET3_REG_TXPROD, 1132 VMXNET3_DEVICE_MAX_TX_QUEUES, VMXNET3_REG_ALIGN)) { 1133 int tx_queue_idx = 1134 VMW_MULTIREG_IDX_BY_ADDR(addr, VMXNET3_REG_TXPROD, 1135 VMXNET3_REG_ALIGN); 1136 assert(tx_queue_idx <= s->txq_num); 1137 vmxnet3_process_tx_queue(s, tx_queue_idx); 1138 return; 1139 } 1140 1141 if (VMW_IS_MULTIREG_ADDR(addr, VMXNET3_REG_IMR, 1142 VMXNET3_MAX_INTRS, VMXNET3_REG_ALIGN)) { 1143 int l = VMW_MULTIREG_IDX_BY_ADDR(addr, VMXNET3_REG_IMR, 1144 VMXNET3_REG_ALIGN); 1145 1146 VMW_CBPRN("Interrupt mask for line %d written: 0x%" PRIx64, l, val); 1147 1148 vmxnet3_on_interrupt_mask_changed(s, l, val); 1149 return; 1150 } 1151 1152 if (VMW_IS_MULTIREG_ADDR(addr, VMXNET3_REG_RXPROD, 1153 VMXNET3_DEVICE_MAX_RX_QUEUES, VMXNET3_REG_ALIGN) || 1154 VMW_IS_MULTIREG_ADDR(addr, VMXNET3_REG_RXPROD2, 1155 VMXNET3_DEVICE_MAX_RX_QUEUES, VMXNET3_REG_ALIGN)) { 1156 return; 1157 } 1158 1159 VMW_WRPRN("BAR0 unknown write [%" PRIx64 "] = %" PRIx64 ", size %d", 1160 (uint64_t) addr, val, size); 1161 } 1162 1163 static uint64_t 1164 vmxnet3_io_bar0_read(void *opaque, hwaddr addr, unsigned size) 1165 { 1166 VMXNET3State *s = opaque; 1167 1168 if (VMW_IS_MULTIREG_ADDR(addr, VMXNET3_REG_IMR, 1169 VMXNET3_MAX_INTRS, VMXNET3_REG_ALIGN)) { 1170 int l = VMW_MULTIREG_IDX_BY_ADDR(addr, VMXNET3_REG_IMR, 1171 VMXNET3_REG_ALIGN); 1172 return s->interrupt_states[l].is_masked; 1173 } 1174 1175 VMW_CBPRN("BAR0 unknown read [%" PRIx64 "], size %d", addr, size); 1176 return 0; 1177 } 1178 1179 static void vmxnet3_reset_interrupt_states(VMXNET3State *s) 1180 { 1181 int i; 1182 for (i = 0; i < ARRAY_SIZE(s->interrupt_states); i++) { 1183 s->interrupt_states[i].is_asserted = false; 1184 s->interrupt_states[i].is_pending = false; 1185 s->interrupt_states[i].is_masked = true; 1186 } 1187 } 1188 1189 static void vmxnet3_reset_mac(VMXNET3State *s) 1190 { 1191 memcpy(&s->conf.macaddr.a, &s->perm_mac.a, sizeof(s->perm_mac.a)); 1192 VMW_CFPRN("MAC address set to: " VMXNET_MF, VMXNET_MA(s->conf.macaddr.a)); 1193 } 1194 1195 static void vmxnet3_deactivate_device(VMXNET3State *s) 1196 { 1197 VMW_CBPRN("Deactivating vmxnet3..."); 1198 s->device_active = false; 1199 } 1200 1201 static void vmxnet3_reset(VMXNET3State *s) 1202 { 1203 VMW_CBPRN("Resetting vmxnet3..."); 1204 1205 vmxnet3_deactivate_device(s); 1206 vmxnet3_reset_interrupt_states(s); 1207 vmxnet_tx_pkt_reset(s->tx_pkt); 1208 s->drv_shmem = 0; 1209 s->tx_sop = true; 1210 s->skip_current_tx_pkt = false; 1211 } 1212 1213 static void vmxnet3_update_rx_mode(VMXNET3State *s) 1214 { 1215 s->rx_mode = VMXNET3_READ_DRV_SHARED32(s->drv_shmem, 1216 devRead.rxFilterConf.rxMode); 1217 VMW_CFPRN("RX mode: 0x%08X", s->rx_mode); 1218 } 1219 1220 static void vmxnet3_update_vlan_filters(VMXNET3State *s) 1221 { 1222 int i; 1223 1224 /* Copy configuration from shared memory */ 1225 VMXNET3_READ_DRV_SHARED(s->drv_shmem, 1226 devRead.rxFilterConf.vfTable, 1227 s->vlan_table, 1228 sizeof(s->vlan_table)); 1229 1230 /* Invert byte order when needed */ 1231 for (i = 0; i < ARRAY_SIZE(s->vlan_table); i++) { 1232 s->vlan_table[i] = le32_to_cpu(s->vlan_table[i]); 1233 } 1234 1235 /* Dump configuration for debugging purposes */ 1236 VMW_CFPRN("Configured VLANs:"); 1237 for (i = 0; i < sizeof(s->vlan_table) * 8; i++) { 1238 if (VMXNET3_VFTABLE_ENTRY_IS_SET(s->vlan_table, i)) { 1239 VMW_CFPRN("\tVLAN %d is present", i); 1240 } 1241 } 1242 } 1243 1244 static void vmxnet3_update_mcast_filters(VMXNET3State *s) 1245 { 1246 uint16_t list_bytes = 1247 VMXNET3_READ_DRV_SHARED16(s->drv_shmem, 1248 devRead.rxFilterConf.mfTableLen); 1249 1250 s->mcast_list_len = list_bytes / sizeof(s->mcast_list[0]); 1251 1252 s->mcast_list = g_realloc(s->mcast_list, list_bytes); 1253 if (!s->mcast_list) { 1254 if (s->mcast_list_len == 0) { 1255 VMW_CFPRN("Current multicast list is empty"); 1256 } else { 1257 VMW_ERPRN("Failed to allocate multicast list of %d elements", 1258 s->mcast_list_len); 1259 } 1260 s->mcast_list_len = 0; 1261 } else { 1262 int i; 1263 hwaddr mcast_list_pa = 1264 VMXNET3_READ_DRV_SHARED64(s->drv_shmem, 1265 devRead.rxFilterConf.mfTablePA); 1266 1267 cpu_physical_memory_read(mcast_list_pa, s->mcast_list, list_bytes); 1268 VMW_CFPRN("Current multicast list len is %d:", s->mcast_list_len); 1269 for (i = 0; i < s->mcast_list_len; i++) { 1270 VMW_CFPRN("\t" VMXNET_MF, VMXNET_MA(s->mcast_list[i].a)); 1271 } 1272 } 1273 } 1274 1275 static void vmxnet3_setup_rx_filtering(VMXNET3State *s) 1276 { 1277 vmxnet3_update_rx_mode(s); 1278 vmxnet3_update_vlan_filters(s); 1279 vmxnet3_update_mcast_filters(s); 1280 } 1281 1282 static uint32_t vmxnet3_get_interrupt_config(VMXNET3State *s) 1283 { 1284 uint32_t interrupt_mode = VMXNET3_IT_AUTO | (VMXNET3_IMM_AUTO << 2); 1285 VMW_CFPRN("Interrupt config is 0x%X", interrupt_mode); 1286 return interrupt_mode; 1287 } 1288 1289 static void vmxnet3_fill_stats(VMXNET3State *s) 1290 { 1291 int i; 1292 1293 if (!s->device_active) 1294 return; 1295 1296 for (i = 0; i < s->txq_num; i++) { 1297 cpu_physical_memory_write(s->txq_descr[i].tx_stats_pa, 1298 &s->txq_descr[i].txq_stats, 1299 sizeof(s->txq_descr[i].txq_stats)); 1300 } 1301 1302 for (i = 0; i < s->rxq_num; i++) { 1303 cpu_physical_memory_write(s->rxq_descr[i].rx_stats_pa, 1304 &s->rxq_descr[i].rxq_stats, 1305 sizeof(s->rxq_descr[i].rxq_stats)); 1306 } 1307 } 1308 1309 static void vmxnet3_adjust_by_guest_type(VMXNET3State *s) 1310 { 1311 struct Vmxnet3_GOSInfo gos; 1312 1313 VMXNET3_READ_DRV_SHARED(s->drv_shmem, devRead.misc.driverInfo.gos, 1314 &gos, sizeof(gos)); 1315 s->rx_packets_compound = 1316 (gos.gosType == VMXNET3_GOS_TYPE_WIN) ? false : true; 1317 1318 VMW_CFPRN("Guest type specifics: RXCOMPOUND: %d", s->rx_packets_compound); 1319 } 1320 1321 static void 1322 vmxnet3_dump_conf_descr(const char *name, 1323 struct Vmxnet3_VariableLenConfDesc *pm_descr) 1324 { 1325 VMW_CFPRN("%s descriptor dump: Version %u, Length %u", 1326 name, pm_descr->confVer, pm_descr->confLen); 1327 1328 }; 1329 1330 static void vmxnet3_update_pm_state(VMXNET3State *s) 1331 { 1332 struct Vmxnet3_VariableLenConfDesc pm_descr; 1333 1334 pm_descr.confLen = 1335 VMXNET3_READ_DRV_SHARED32(s->drv_shmem, devRead.pmConfDesc.confLen); 1336 pm_descr.confVer = 1337 VMXNET3_READ_DRV_SHARED32(s->drv_shmem, devRead.pmConfDesc.confVer); 1338 pm_descr.confPA = 1339 VMXNET3_READ_DRV_SHARED64(s->drv_shmem, devRead.pmConfDesc.confPA); 1340 1341 vmxnet3_dump_conf_descr("PM State", &pm_descr); 1342 } 1343 1344 static void vmxnet3_update_features(VMXNET3State *s) 1345 { 1346 uint32_t guest_features; 1347 int rxcso_supported; 1348 1349 guest_features = VMXNET3_READ_DRV_SHARED32(s->drv_shmem, 1350 devRead.misc.uptFeatures); 1351 1352 rxcso_supported = VMXNET_FLAG_IS_SET(guest_features, UPT1_F_RXCSUM); 1353 s->rx_vlan_stripping = VMXNET_FLAG_IS_SET(guest_features, UPT1_F_RXVLAN); 1354 s->lro_supported = VMXNET_FLAG_IS_SET(guest_features, UPT1_F_LRO); 1355 1356 VMW_CFPRN("Features configuration: LRO: %d, RXCSUM: %d, VLANSTRIP: %d", 1357 s->lro_supported, rxcso_supported, 1358 s->rx_vlan_stripping); 1359 if (s->peer_has_vhdr) { 1360 qemu_set_offload(qemu_get_queue(s->nic)->peer, 1361 rxcso_supported, 1362 s->lro_supported, 1363 s->lro_supported, 1364 0, 1365 0); 1366 } 1367 } 1368 1369 static bool vmxnet3_verify_intx(VMXNET3State *s, int intx) 1370 { 1371 return s->msix_used || s->msi_used || (intx == 1372 (pci_get_byte(s->parent_obj.config + PCI_INTERRUPT_PIN) - 1)); 1373 } 1374 1375 static void vmxnet3_validate_interrupt_idx(bool is_msix, int idx) 1376 { 1377 int max_ints = is_msix ? VMXNET3_MAX_INTRS : VMXNET3_MAX_NMSIX_INTRS; 1378 if (idx >= max_ints) { 1379 hw_error("Bad interrupt index: %d\n", idx); 1380 } 1381 } 1382 1383 static void vmxnet3_validate_interrupts(VMXNET3State *s) 1384 { 1385 int i; 1386 1387 VMW_CFPRN("Verifying event interrupt index (%d)", s->event_int_idx); 1388 vmxnet3_validate_interrupt_idx(s->msix_used, s->event_int_idx); 1389 1390 for (i = 0; i < s->txq_num; i++) { 1391 int idx = s->txq_descr[i].intr_idx; 1392 VMW_CFPRN("Verifying TX queue %d interrupt index (%d)", i, idx); 1393 vmxnet3_validate_interrupt_idx(s->msix_used, idx); 1394 } 1395 1396 for (i = 0; i < s->rxq_num; i++) { 1397 int idx = s->rxq_descr[i].intr_idx; 1398 VMW_CFPRN("Verifying RX queue %d interrupt index (%d)", i, idx); 1399 vmxnet3_validate_interrupt_idx(s->msix_used, idx); 1400 } 1401 } 1402 1403 static void vmxnet3_validate_queues(VMXNET3State *s) 1404 { 1405 /* 1406 * txq_num and rxq_num are total number of queues 1407 * configured by guest. These numbers must not 1408 * exceed corresponding maximal values. 1409 */ 1410 1411 if (s->txq_num > VMXNET3_DEVICE_MAX_TX_QUEUES) { 1412 hw_error("Bad TX queues number: %d\n", s->txq_num); 1413 } 1414 1415 if (s->rxq_num > VMXNET3_DEVICE_MAX_RX_QUEUES) { 1416 hw_error("Bad RX queues number: %d\n", s->rxq_num); 1417 } 1418 } 1419 1420 static void vmxnet3_activate_device(VMXNET3State *s) 1421 { 1422 int i; 1423 static const uint32_t VMXNET3_DEF_TX_THRESHOLD = 1; 1424 hwaddr qdescr_table_pa; 1425 uint64_t pa; 1426 uint32_t size; 1427 1428 /* Verify configuration consistency */ 1429 if (!vmxnet3_verify_driver_magic(s->drv_shmem)) { 1430 VMW_ERPRN("Device configuration received from driver is invalid"); 1431 return; 1432 } 1433 1434 vmxnet3_adjust_by_guest_type(s); 1435 vmxnet3_update_features(s); 1436 vmxnet3_update_pm_state(s); 1437 vmxnet3_setup_rx_filtering(s); 1438 /* Cache fields from shared memory */ 1439 s->mtu = VMXNET3_READ_DRV_SHARED32(s->drv_shmem, devRead.misc.mtu); 1440 VMW_CFPRN("MTU is %u", s->mtu); 1441 1442 s->max_rx_frags = 1443 VMXNET3_READ_DRV_SHARED16(s->drv_shmem, devRead.misc.maxNumRxSG); 1444 1445 if (s->max_rx_frags == 0) { 1446 s->max_rx_frags = 1; 1447 } 1448 1449 VMW_CFPRN("Max RX fragments is %u", s->max_rx_frags); 1450 1451 s->event_int_idx = 1452 VMXNET3_READ_DRV_SHARED8(s->drv_shmem, devRead.intrConf.eventIntrIdx); 1453 assert(vmxnet3_verify_intx(s, s->event_int_idx)); 1454 VMW_CFPRN("Events interrupt line is %u", s->event_int_idx); 1455 1456 s->auto_int_masking = 1457 VMXNET3_READ_DRV_SHARED8(s->drv_shmem, devRead.intrConf.autoMask); 1458 VMW_CFPRN("Automatic interrupt masking is %d", (int)s->auto_int_masking); 1459 1460 s->txq_num = 1461 VMXNET3_READ_DRV_SHARED8(s->drv_shmem, devRead.misc.numTxQueues); 1462 s->rxq_num = 1463 VMXNET3_READ_DRV_SHARED8(s->drv_shmem, devRead.misc.numRxQueues); 1464 1465 VMW_CFPRN("Number of TX/RX queues %u/%u", s->txq_num, s->rxq_num); 1466 vmxnet3_validate_queues(s); 1467 1468 qdescr_table_pa = 1469 VMXNET3_READ_DRV_SHARED64(s->drv_shmem, devRead.misc.queueDescPA); 1470 VMW_CFPRN("TX queues descriptors table is at 0x%" PRIx64, qdescr_table_pa); 1471 1472 /* 1473 * Worst-case scenario is a packet that holds all TX rings space so 1474 * we calculate total size of all TX rings for max TX fragments number 1475 */ 1476 s->max_tx_frags = 0; 1477 1478 /* TX queues */ 1479 for (i = 0; i < s->txq_num; i++) { 1480 hwaddr qdescr_pa = 1481 qdescr_table_pa + i * sizeof(struct Vmxnet3_TxQueueDesc); 1482 1483 /* Read interrupt number for this TX queue */ 1484 s->txq_descr[i].intr_idx = 1485 VMXNET3_READ_TX_QUEUE_DESCR8(qdescr_pa, conf.intrIdx); 1486 assert(vmxnet3_verify_intx(s, s->txq_descr[i].intr_idx)); 1487 1488 VMW_CFPRN("TX Queue %d interrupt: %d", i, s->txq_descr[i].intr_idx); 1489 1490 /* Read rings memory locations for TX queues */ 1491 pa = VMXNET3_READ_TX_QUEUE_DESCR64(qdescr_pa, conf.txRingBasePA); 1492 size = VMXNET3_READ_TX_QUEUE_DESCR32(qdescr_pa, conf.txRingSize); 1493 1494 vmxnet3_ring_init(&s->txq_descr[i].tx_ring, pa, size, 1495 sizeof(struct Vmxnet3_TxDesc), false); 1496 VMXNET3_RING_DUMP(VMW_CFPRN, "TX", i, &s->txq_descr[i].tx_ring); 1497 1498 s->max_tx_frags += size; 1499 1500 /* TXC ring */ 1501 pa = VMXNET3_READ_TX_QUEUE_DESCR64(qdescr_pa, conf.compRingBasePA); 1502 size = VMXNET3_READ_TX_QUEUE_DESCR32(qdescr_pa, conf.compRingSize); 1503 vmxnet3_ring_init(&s->txq_descr[i].comp_ring, pa, size, 1504 sizeof(struct Vmxnet3_TxCompDesc), true); 1505 VMXNET3_RING_DUMP(VMW_CFPRN, "TXC", i, &s->txq_descr[i].comp_ring); 1506 1507 s->txq_descr[i].tx_stats_pa = 1508 qdescr_pa + offsetof(struct Vmxnet3_TxQueueDesc, stats); 1509 1510 memset(&s->txq_descr[i].txq_stats, 0, 1511 sizeof(s->txq_descr[i].txq_stats)); 1512 1513 /* Fill device-managed parameters for queues */ 1514 VMXNET3_WRITE_TX_QUEUE_DESCR32(qdescr_pa, 1515 ctrl.txThreshold, 1516 VMXNET3_DEF_TX_THRESHOLD); 1517 } 1518 1519 /* Preallocate TX packet wrapper */ 1520 VMW_CFPRN("Max TX fragments is %u", s->max_tx_frags); 1521 vmxnet_tx_pkt_init(&s->tx_pkt, s->max_tx_frags, s->peer_has_vhdr); 1522 vmxnet_rx_pkt_init(&s->rx_pkt, s->peer_has_vhdr); 1523 1524 /* Read rings memory locations for RX queues */ 1525 for (i = 0; i < s->rxq_num; i++) { 1526 int j; 1527 hwaddr qd_pa = 1528 qdescr_table_pa + s->txq_num * sizeof(struct Vmxnet3_TxQueueDesc) + 1529 i * sizeof(struct Vmxnet3_RxQueueDesc); 1530 1531 /* Read interrupt number for this RX queue */ 1532 s->rxq_descr[i].intr_idx = 1533 VMXNET3_READ_TX_QUEUE_DESCR8(qd_pa, conf.intrIdx); 1534 assert(vmxnet3_verify_intx(s, s->rxq_descr[i].intr_idx)); 1535 1536 VMW_CFPRN("RX Queue %d interrupt: %d", i, s->rxq_descr[i].intr_idx); 1537 1538 /* Read rings memory locations */ 1539 for (j = 0; j < VMXNET3_RX_RINGS_PER_QUEUE; j++) { 1540 /* RX rings */ 1541 pa = VMXNET3_READ_RX_QUEUE_DESCR64(qd_pa, conf.rxRingBasePA[j]); 1542 size = VMXNET3_READ_RX_QUEUE_DESCR32(qd_pa, conf.rxRingSize[j]); 1543 vmxnet3_ring_init(&s->rxq_descr[i].rx_ring[j], pa, size, 1544 sizeof(struct Vmxnet3_RxDesc), false); 1545 VMW_CFPRN("RX queue %d:%d: Base: %" PRIx64 ", Size: %d", 1546 i, j, pa, size); 1547 } 1548 1549 /* RXC ring */ 1550 pa = VMXNET3_READ_RX_QUEUE_DESCR64(qd_pa, conf.compRingBasePA); 1551 size = VMXNET3_READ_RX_QUEUE_DESCR32(qd_pa, conf.compRingSize); 1552 vmxnet3_ring_init(&s->rxq_descr[i].comp_ring, pa, size, 1553 sizeof(struct Vmxnet3_RxCompDesc), true); 1554 VMW_CFPRN("RXC queue %d: Base: %" PRIx64 ", Size: %d", i, pa, size); 1555 1556 s->rxq_descr[i].rx_stats_pa = 1557 qd_pa + offsetof(struct Vmxnet3_RxQueueDesc, stats); 1558 memset(&s->rxq_descr[i].rxq_stats, 0, 1559 sizeof(s->rxq_descr[i].rxq_stats)); 1560 } 1561 1562 vmxnet3_validate_interrupts(s); 1563 1564 /* Make sure everything is in place before device activation */ 1565 smp_wmb(); 1566 1567 vmxnet3_reset_mac(s); 1568 1569 s->device_active = true; 1570 } 1571 1572 static void vmxnet3_handle_command(VMXNET3State *s, uint64_t cmd) 1573 { 1574 s->last_command = cmd; 1575 1576 switch (cmd) { 1577 case VMXNET3_CMD_GET_PERM_MAC_HI: 1578 VMW_CBPRN("Set: Get upper part of permanent MAC"); 1579 break; 1580 1581 case VMXNET3_CMD_GET_PERM_MAC_LO: 1582 VMW_CBPRN("Set: Get lower part of permanent MAC"); 1583 break; 1584 1585 case VMXNET3_CMD_GET_STATS: 1586 VMW_CBPRN("Set: Get device statistics"); 1587 vmxnet3_fill_stats(s); 1588 break; 1589 1590 case VMXNET3_CMD_ACTIVATE_DEV: 1591 VMW_CBPRN("Set: Activating vmxnet3 device"); 1592 vmxnet3_activate_device(s); 1593 break; 1594 1595 case VMXNET3_CMD_UPDATE_RX_MODE: 1596 VMW_CBPRN("Set: Update rx mode"); 1597 vmxnet3_update_rx_mode(s); 1598 break; 1599 1600 case VMXNET3_CMD_UPDATE_VLAN_FILTERS: 1601 VMW_CBPRN("Set: Update VLAN filters"); 1602 vmxnet3_update_vlan_filters(s); 1603 break; 1604 1605 case VMXNET3_CMD_UPDATE_MAC_FILTERS: 1606 VMW_CBPRN("Set: Update MAC filters"); 1607 vmxnet3_update_mcast_filters(s); 1608 break; 1609 1610 case VMXNET3_CMD_UPDATE_FEATURE: 1611 VMW_CBPRN("Set: Update features"); 1612 vmxnet3_update_features(s); 1613 break; 1614 1615 case VMXNET3_CMD_UPDATE_PMCFG: 1616 VMW_CBPRN("Set: Update power management config"); 1617 vmxnet3_update_pm_state(s); 1618 break; 1619 1620 case VMXNET3_CMD_GET_LINK: 1621 VMW_CBPRN("Set: Get link"); 1622 break; 1623 1624 case VMXNET3_CMD_RESET_DEV: 1625 VMW_CBPRN("Set: Reset device"); 1626 vmxnet3_reset(s); 1627 break; 1628 1629 case VMXNET3_CMD_QUIESCE_DEV: 1630 VMW_CBPRN("Set: VMXNET3_CMD_QUIESCE_DEV - pause the device"); 1631 vmxnet3_deactivate_device(s); 1632 break; 1633 1634 case VMXNET3_CMD_GET_CONF_INTR: 1635 VMW_CBPRN("Set: VMXNET3_CMD_GET_CONF_INTR - interrupt configuration"); 1636 break; 1637 1638 case VMXNET3_CMD_GET_ADAPTIVE_RING_INFO: 1639 VMW_CBPRN("Set: VMXNET3_CMD_GET_ADAPTIVE_RING_INFO - " 1640 "adaptive ring info flags"); 1641 break; 1642 1643 default: 1644 VMW_CBPRN("Received unknown command: %" PRIx64, cmd); 1645 break; 1646 } 1647 } 1648 1649 static uint64_t vmxnet3_get_command_status(VMXNET3State *s) 1650 { 1651 uint64_t ret; 1652 1653 switch (s->last_command) { 1654 case VMXNET3_CMD_ACTIVATE_DEV: 1655 ret = (s->device_active) ? 0 : -1; 1656 VMW_CFPRN("Device active: %" PRIx64, ret); 1657 break; 1658 1659 case VMXNET3_CMD_RESET_DEV: 1660 case VMXNET3_CMD_QUIESCE_DEV: 1661 case VMXNET3_CMD_GET_QUEUE_STATUS: 1662 ret = 0; 1663 break; 1664 1665 case VMXNET3_CMD_GET_LINK: 1666 ret = s->link_status_and_speed; 1667 VMW_CFPRN("Link and speed: %" PRIx64, ret); 1668 break; 1669 1670 case VMXNET3_CMD_GET_PERM_MAC_LO: 1671 ret = vmxnet3_get_mac_low(&s->perm_mac); 1672 break; 1673 1674 case VMXNET3_CMD_GET_PERM_MAC_HI: 1675 ret = vmxnet3_get_mac_high(&s->perm_mac); 1676 break; 1677 1678 case VMXNET3_CMD_GET_CONF_INTR: 1679 ret = vmxnet3_get_interrupt_config(s); 1680 break; 1681 1682 case VMXNET3_CMD_GET_ADAPTIVE_RING_INFO: 1683 ret = VMXNET3_DISABLE_ADAPTIVE_RING; 1684 break; 1685 1686 default: 1687 VMW_WRPRN("Received request for unknown command: %x", s->last_command); 1688 ret = -1; 1689 break; 1690 } 1691 1692 return ret; 1693 } 1694 1695 static void vmxnet3_set_events(VMXNET3State *s, uint32_t val) 1696 { 1697 uint32_t events; 1698 1699 VMW_CBPRN("Setting events: 0x%x", val); 1700 events = VMXNET3_READ_DRV_SHARED32(s->drv_shmem, ecr) | val; 1701 VMXNET3_WRITE_DRV_SHARED32(s->drv_shmem, ecr, events); 1702 } 1703 1704 static void vmxnet3_ack_events(VMXNET3State *s, uint32_t val) 1705 { 1706 uint32_t events; 1707 1708 VMW_CBPRN("Clearing events: 0x%x", val); 1709 events = VMXNET3_READ_DRV_SHARED32(s->drv_shmem, ecr) & ~val; 1710 VMXNET3_WRITE_DRV_SHARED32(s->drv_shmem, ecr, events); 1711 } 1712 1713 static void 1714 vmxnet3_io_bar1_write(void *opaque, 1715 hwaddr addr, 1716 uint64_t val, 1717 unsigned size) 1718 { 1719 VMXNET3State *s = opaque; 1720 1721 switch (addr) { 1722 /* Vmxnet3 Revision Report Selection */ 1723 case VMXNET3_REG_VRRS: 1724 VMW_CBPRN("Write BAR1 [VMXNET3_REG_VRRS] = %" PRIx64 ", size %d", 1725 val, size); 1726 break; 1727 1728 /* UPT Version Report Selection */ 1729 case VMXNET3_REG_UVRS: 1730 VMW_CBPRN("Write BAR1 [VMXNET3_REG_UVRS] = %" PRIx64 ", size %d", 1731 val, size); 1732 break; 1733 1734 /* Driver Shared Address Low */ 1735 case VMXNET3_REG_DSAL: 1736 VMW_CBPRN("Write BAR1 [VMXNET3_REG_DSAL] = %" PRIx64 ", size %d", 1737 val, size); 1738 /* 1739 * Guest driver will first write the low part of the shared 1740 * memory address. We save it to temp variable and set the 1741 * shared address only after we get the high part 1742 */ 1743 if (val == 0) { 1744 s->device_active = false; 1745 } 1746 s->temp_shared_guest_driver_memory = val; 1747 s->drv_shmem = 0; 1748 break; 1749 1750 /* Driver Shared Address High */ 1751 case VMXNET3_REG_DSAH: 1752 VMW_CBPRN("Write BAR1 [VMXNET3_REG_DSAH] = %" PRIx64 ", size %d", 1753 val, size); 1754 /* 1755 * Set the shared memory between guest driver and device. 1756 * We already should have low address part. 1757 */ 1758 s->drv_shmem = s->temp_shared_guest_driver_memory | (val << 32); 1759 break; 1760 1761 /* Command */ 1762 case VMXNET3_REG_CMD: 1763 VMW_CBPRN("Write BAR1 [VMXNET3_REG_CMD] = %" PRIx64 ", size %d", 1764 val, size); 1765 vmxnet3_handle_command(s, val); 1766 break; 1767 1768 /* MAC Address Low */ 1769 case VMXNET3_REG_MACL: 1770 VMW_CBPRN("Write BAR1 [VMXNET3_REG_MACL] = %" PRIx64 ", size %d", 1771 val, size); 1772 s->temp_mac = val; 1773 break; 1774 1775 /* MAC Address High */ 1776 case VMXNET3_REG_MACH: 1777 VMW_CBPRN("Write BAR1 [VMXNET3_REG_MACH] = %" PRIx64 ", size %d", 1778 val, size); 1779 vmxnet3_set_variable_mac(s, val, s->temp_mac); 1780 break; 1781 1782 /* Interrupt Cause Register */ 1783 case VMXNET3_REG_ICR: 1784 VMW_CBPRN("Write BAR1 [VMXNET3_REG_ICR] = %" PRIx64 ", size %d", 1785 val, size); 1786 g_assert_not_reached(); 1787 break; 1788 1789 /* Event Cause Register */ 1790 case VMXNET3_REG_ECR: 1791 VMW_CBPRN("Write BAR1 [VMXNET3_REG_ECR] = %" PRIx64 ", size %d", 1792 val, size); 1793 vmxnet3_ack_events(s, val); 1794 break; 1795 1796 default: 1797 VMW_CBPRN("Unknown Write to BAR1 [%" PRIx64 "] = %" PRIx64 ", size %d", 1798 addr, val, size); 1799 break; 1800 } 1801 } 1802 1803 static uint64_t 1804 vmxnet3_io_bar1_read(void *opaque, hwaddr addr, unsigned size) 1805 { 1806 VMXNET3State *s = opaque; 1807 uint64_t ret = 0; 1808 1809 switch (addr) { 1810 /* Vmxnet3 Revision Report Selection */ 1811 case VMXNET3_REG_VRRS: 1812 VMW_CBPRN("Read BAR1 [VMXNET3_REG_VRRS], size %d", size); 1813 ret = VMXNET3_DEVICE_REVISION; 1814 break; 1815 1816 /* UPT Version Report Selection */ 1817 case VMXNET3_REG_UVRS: 1818 VMW_CBPRN("Read BAR1 [VMXNET3_REG_UVRS], size %d", size); 1819 ret = VMXNET3_DEVICE_VERSION; 1820 break; 1821 1822 /* Command */ 1823 case VMXNET3_REG_CMD: 1824 VMW_CBPRN("Read BAR1 [VMXNET3_REG_CMD], size %d", size); 1825 ret = vmxnet3_get_command_status(s); 1826 break; 1827 1828 /* MAC Address Low */ 1829 case VMXNET3_REG_MACL: 1830 VMW_CBPRN("Read BAR1 [VMXNET3_REG_MACL], size %d", size); 1831 ret = vmxnet3_get_mac_low(&s->conf.macaddr); 1832 break; 1833 1834 /* MAC Address High */ 1835 case VMXNET3_REG_MACH: 1836 VMW_CBPRN("Read BAR1 [VMXNET3_REG_MACH], size %d", size); 1837 ret = vmxnet3_get_mac_high(&s->conf.macaddr); 1838 break; 1839 1840 /* 1841 * Interrupt Cause Register 1842 * Used for legacy interrupts only so interrupt index always 0 1843 */ 1844 case VMXNET3_REG_ICR: 1845 VMW_CBPRN("Read BAR1 [VMXNET3_REG_ICR], size %d", size); 1846 if (vmxnet3_interrupt_asserted(s, 0)) { 1847 vmxnet3_clear_interrupt(s, 0); 1848 ret = true; 1849 } else { 1850 ret = false; 1851 } 1852 break; 1853 1854 default: 1855 VMW_CBPRN("Unknow read BAR1[%" PRIx64 "], %d bytes", addr, size); 1856 break; 1857 } 1858 1859 return ret; 1860 } 1861 1862 static int 1863 vmxnet3_can_receive(NetClientState *nc) 1864 { 1865 VMXNET3State *s = qemu_get_nic_opaque(nc); 1866 return s->device_active && 1867 VMXNET_FLAG_IS_SET(s->link_status_and_speed, VMXNET3_LINK_STATUS_UP); 1868 } 1869 1870 static inline bool 1871 vmxnet3_is_registered_vlan(VMXNET3State *s, const void *data) 1872 { 1873 uint16_t vlan_tag = eth_get_pkt_tci(data) & VLAN_VID_MASK; 1874 if (IS_SPECIAL_VLAN_ID(vlan_tag)) { 1875 return true; 1876 } 1877 1878 return VMXNET3_VFTABLE_ENTRY_IS_SET(s->vlan_table, vlan_tag); 1879 } 1880 1881 static bool 1882 vmxnet3_is_allowed_mcast_group(VMXNET3State *s, const uint8_t *group_mac) 1883 { 1884 int i; 1885 for (i = 0; i < s->mcast_list_len; i++) { 1886 if (!memcmp(group_mac, s->mcast_list[i].a, sizeof(s->mcast_list[i]))) { 1887 return true; 1888 } 1889 } 1890 return false; 1891 } 1892 1893 static bool 1894 vmxnet3_rx_filter_may_indicate(VMXNET3State *s, const void *data, 1895 size_t size) 1896 { 1897 struct eth_header *ehdr = PKT_GET_ETH_HDR(data); 1898 1899 if (VMXNET_FLAG_IS_SET(s->rx_mode, VMXNET3_RXM_PROMISC)) { 1900 return true; 1901 } 1902 1903 if (!vmxnet3_is_registered_vlan(s, data)) { 1904 return false; 1905 } 1906 1907 switch (vmxnet_rx_pkt_get_packet_type(s->rx_pkt)) { 1908 case ETH_PKT_UCAST: 1909 if (!VMXNET_FLAG_IS_SET(s->rx_mode, VMXNET3_RXM_UCAST)) { 1910 return false; 1911 } 1912 if (memcmp(s->conf.macaddr.a, ehdr->h_dest, ETH_ALEN)) { 1913 return false; 1914 } 1915 break; 1916 1917 case ETH_PKT_BCAST: 1918 if (!VMXNET_FLAG_IS_SET(s->rx_mode, VMXNET3_RXM_BCAST)) { 1919 return false; 1920 } 1921 break; 1922 1923 case ETH_PKT_MCAST: 1924 if (VMXNET_FLAG_IS_SET(s->rx_mode, VMXNET3_RXM_ALL_MULTI)) { 1925 return true; 1926 } 1927 if (!VMXNET_FLAG_IS_SET(s->rx_mode, VMXNET3_RXM_MCAST)) { 1928 return false; 1929 } 1930 if (!vmxnet3_is_allowed_mcast_group(s, ehdr->h_dest)) { 1931 return false; 1932 } 1933 break; 1934 1935 default: 1936 g_assert_not_reached(); 1937 } 1938 1939 return true; 1940 } 1941 1942 static ssize_t 1943 vmxnet3_receive(NetClientState *nc, const uint8_t *buf, size_t size) 1944 { 1945 VMXNET3State *s = qemu_get_nic_opaque(nc); 1946 size_t bytes_indicated; 1947 uint8_t min_buf[MIN_BUF_SIZE]; 1948 1949 if (!vmxnet3_can_receive(nc)) { 1950 VMW_PKPRN("Cannot receive now"); 1951 return -1; 1952 } 1953 1954 if (s->peer_has_vhdr) { 1955 vmxnet_rx_pkt_set_vhdr(s->rx_pkt, (struct virtio_net_hdr *)buf); 1956 buf += sizeof(struct virtio_net_hdr); 1957 size -= sizeof(struct virtio_net_hdr); 1958 } 1959 1960 /* Pad to minimum Ethernet frame length */ 1961 if (size < sizeof(min_buf)) { 1962 memcpy(min_buf, buf, size); 1963 memset(&min_buf[size], 0, sizeof(min_buf) - size); 1964 buf = min_buf; 1965 size = sizeof(min_buf); 1966 } 1967 1968 vmxnet_rx_pkt_set_packet_type(s->rx_pkt, 1969 get_eth_packet_type(PKT_GET_ETH_HDR(buf))); 1970 1971 if (vmxnet3_rx_filter_may_indicate(s, buf, size)) { 1972 vmxnet_rx_pkt_set_protocols(s->rx_pkt, buf, size); 1973 vmxnet3_rx_need_csum_calculate(s->rx_pkt, buf, size); 1974 vmxnet_rx_pkt_attach_data(s->rx_pkt, buf, size, s->rx_vlan_stripping); 1975 bytes_indicated = vmxnet3_indicate_packet(s) ? size : -1; 1976 if (bytes_indicated < size) { 1977 VMW_PKPRN("RX: %lu of %lu bytes indicated", bytes_indicated, size); 1978 } 1979 } else { 1980 VMW_PKPRN("Packet dropped by RX filter"); 1981 bytes_indicated = size; 1982 } 1983 1984 assert(size > 0); 1985 assert(bytes_indicated != 0); 1986 return bytes_indicated; 1987 } 1988 1989 static void vmxnet3_set_link_status(NetClientState *nc) 1990 { 1991 VMXNET3State *s = qemu_get_nic_opaque(nc); 1992 1993 if (nc->link_down) { 1994 s->link_status_and_speed &= ~VMXNET3_LINK_STATUS_UP; 1995 } else { 1996 s->link_status_and_speed |= VMXNET3_LINK_STATUS_UP; 1997 } 1998 1999 vmxnet3_set_events(s, VMXNET3_ECR_LINK); 2000 vmxnet3_trigger_interrupt(s, s->event_int_idx); 2001 } 2002 2003 static NetClientInfo net_vmxnet3_info = { 2004 .type = NET_CLIENT_OPTIONS_KIND_NIC, 2005 .size = sizeof(NICState), 2006 .receive = vmxnet3_receive, 2007 .link_status_changed = vmxnet3_set_link_status, 2008 }; 2009 2010 static bool vmxnet3_peer_has_vnet_hdr(VMXNET3State *s) 2011 { 2012 NetClientState *nc = qemu_get_queue(s->nic); 2013 2014 if (qemu_has_vnet_hdr(nc->peer)) { 2015 return true; 2016 } 2017 2018 return false; 2019 } 2020 2021 static void vmxnet3_net_uninit(VMXNET3State *s) 2022 { 2023 g_free(s->mcast_list); 2024 vmxnet_tx_pkt_reset(s->tx_pkt); 2025 vmxnet_tx_pkt_uninit(s->tx_pkt); 2026 vmxnet_rx_pkt_uninit(s->rx_pkt); 2027 qemu_del_nic(s->nic); 2028 } 2029 2030 static void vmxnet3_net_init(VMXNET3State *s) 2031 { 2032 DeviceState *d = DEVICE(s); 2033 2034 VMW_CBPRN("vmxnet3_net_init called..."); 2035 2036 qemu_macaddr_default_if_unset(&s->conf.macaddr); 2037 2038 /* Windows guest will query the address that was set on init */ 2039 memcpy(&s->perm_mac.a, &s->conf.macaddr.a, sizeof(s->perm_mac.a)); 2040 2041 s->mcast_list = NULL; 2042 s->mcast_list_len = 0; 2043 2044 s->link_status_and_speed = VMXNET3_LINK_SPEED | VMXNET3_LINK_STATUS_UP; 2045 2046 VMW_CFPRN("Permanent MAC: " MAC_FMT, MAC_ARG(s->perm_mac.a)); 2047 2048 s->nic = qemu_new_nic(&net_vmxnet3_info, &s->conf, 2049 object_get_typename(OBJECT(s)), 2050 d->id, s); 2051 2052 s->peer_has_vhdr = vmxnet3_peer_has_vnet_hdr(s); 2053 s->tx_sop = true; 2054 s->skip_current_tx_pkt = false; 2055 s->tx_pkt = NULL; 2056 s->rx_pkt = NULL; 2057 s->rx_vlan_stripping = false; 2058 s->lro_supported = false; 2059 2060 if (s->peer_has_vhdr) { 2061 qemu_set_vnet_hdr_len(qemu_get_queue(s->nic)->peer, 2062 sizeof(struct virtio_net_hdr)); 2063 2064 qemu_using_vnet_hdr(qemu_get_queue(s->nic)->peer, 1); 2065 } 2066 2067 qemu_format_nic_info_str(qemu_get_queue(s->nic), s->conf.macaddr.a); 2068 } 2069 2070 static void 2071 vmxnet3_unuse_msix_vectors(VMXNET3State *s, int num_vectors) 2072 { 2073 PCIDevice *d = PCI_DEVICE(s); 2074 int i; 2075 for (i = 0; i < num_vectors; i++) { 2076 msix_vector_unuse(d, i); 2077 } 2078 } 2079 2080 static bool 2081 vmxnet3_use_msix_vectors(VMXNET3State *s, int num_vectors) 2082 { 2083 PCIDevice *d = PCI_DEVICE(s); 2084 int i; 2085 for (i = 0; i < num_vectors; i++) { 2086 int res = msix_vector_use(d, i); 2087 if (0 > res) { 2088 VMW_WRPRN("Failed to use MSI-X vector %d, error %d", i, res); 2089 vmxnet3_unuse_msix_vectors(s, i); 2090 return false; 2091 } 2092 } 2093 return true; 2094 } 2095 2096 static bool 2097 vmxnet3_init_msix(VMXNET3State *s) 2098 { 2099 PCIDevice *d = PCI_DEVICE(s); 2100 int res = msix_init(d, VMXNET3_MAX_INTRS, 2101 &s->msix_bar, 2102 VMXNET3_MSIX_BAR_IDX, VMXNET3_OFF_MSIX_TABLE, 2103 &s->msix_bar, 2104 VMXNET3_MSIX_BAR_IDX, VMXNET3_OFF_MSIX_PBA, 2105 0); 2106 2107 if (0 > res) { 2108 VMW_WRPRN("Failed to initialize MSI-X, error %d", res); 2109 s->msix_used = false; 2110 } else { 2111 if (!vmxnet3_use_msix_vectors(s, VMXNET3_MAX_INTRS)) { 2112 VMW_WRPRN("Failed to use MSI-X vectors, error %d", res); 2113 msix_uninit(d, &s->msix_bar, &s->msix_bar); 2114 s->msix_used = false; 2115 } else { 2116 s->msix_used = true; 2117 } 2118 } 2119 return s->msix_used; 2120 } 2121 2122 static void 2123 vmxnet3_cleanup_msix(VMXNET3State *s) 2124 { 2125 PCIDevice *d = PCI_DEVICE(s); 2126 2127 if (s->msix_used) { 2128 vmxnet3_unuse_msix_vectors(s, VMXNET3_MAX_INTRS); 2129 msix_uninit(d, &s->msix_bar, &s->msix_bar); 2130 } 2131 } 2132 2133 #define VMXNET3_MSI_OFFSET (0x50) 2134 #define VMXNET3_USE_64BIT (true) 2135 #define VMXNET3_PER_VECTOR_MASK (false) 2136 2137 static bool 2138 vmxnet3_init_msi(VMXNET3State *s) 2139 { 2140 PCIDevice *d = PCI_DEVICE(s); 2141 int res; 2142 2143 res = msi_init(d, VMXNET3_MSI_OFFSET, VMXNET3_MAX_NMSIX_INTRS, 2144 VMXNET3_USE_64BIT, VMXNET3_PER_VECTOR_MASK); 2145 if (0 > res) { 2146 VMW_WRPRN("Failed to initialize MSI, error %d", res); 2147 s->msi_used = false; 2148 } else { 2149 s->msi_used = true; 2150 } 2151 2152 return s->msi_used; 2153 } 2154 2155 static void 2156 vmxnet3_cleanup_msi(VMXNET3State *s) 2157 { 2158 PCIDevice *d = PCI_DEVICE(s); 2159 2160 if (s->msi_used) { 2161 msi_uninit(d); 2162 } 2163 } 2164 2165 static void 2166 vmxnet3_msix_save(QEMUFile *f, void *opaque) 2167 { 2168 PCIDevice *d = PCI_DEVICE(opaque); 2169 msix_save(d, f); 2170 } 2171 2172 static int 2173 vmxnet3_msix_load(QEMUFile *f, void *opaque, int version_id) 2174 { 2175 PCIDevice *d = PCI_DEVICE(opaque); 2176 msix_load(d, f); 2177 return 0; 2178 } 2179 2180 static const MemoryRegionOps b0_ops = { 2181 .read = vmxnet3_io_bar0_read, 2182 .write = vmxnet3_io_bar0_write, 2183 .endianness = DEVICE_LITTLE_ENDIAN, 2184 .impl = { 2185 .min_access_size = 4, 2186 .max_access_size = 4, 2187 }, 2188 }; 2189 2190 static const MemoryRegionOps b1_ops = { 2191 .read = vmxnet3_io_bar1_read, 2192 .write = vmxnet3_io_bar1_write, 2193 .endianness = DEVICE_LITTLE_ENDIAN, 2194 .impl = { 2195 .min_access_size = 4, 2196 .max_access_size = 4, 2197 }, 2198 }; 2199 2200 static void vmxnet3_pci_realize(PCIDevice *pci_dev, Error **errp) 2201 { 2202 DeviceState *dev = DEVICE(pci_dev); 2203 VMXNET3State *s = VMXNET3(pci_dev); 2204 2205 VMW_CBPRN("Starting init..."); 2206 2207 memory_region_init_io(&s->bar0, OBJECT(s), &b0_ops, s, 2208 "vmxnet3-b0", VMXNET3_PT_REG_SIZE); 2209 pci_register_bar(pci_dev, VMXNET3_BAR0_IDX, 2210 PCI_BASE_ADDRESS_SPACE_MEMORY, &s->bar0); 2211 2212 memory_region_init_io(&s->bar1, OBJECT(s), &b1_ops, s, 2213 "vmxnet3-b1", VMXNET3_VD_REG_SIZE); 2214 pci_register_bar(pci_dev, VMXNET3_BAR1_IDX, 2215 PCI_BASE_ADDRESS_SPACE_MEMORY, &s->bar1); 2216 2217 memory_region_init(&s->msix_bar, OBJECT(s), "vmxnet3-msix-bar", 2218 VMXNET3_MSIX_BAR_SIZE); 2219 pci_register_bar(pci_dev, VMXNET3_MSIX_BAR_IDX, 2220 PCI_BASE_ADDRESS_SPACE_MEMORY, &s->msix_bar); 2221 2222 vmxnet3_reset_interrupt_states(s); 2223 2224 /* Interrupt pin A */ 2225 pci_dev->config[PCI_INTERRUPT_PIN] = 0x01; 2226 2227 if (!vmxnet3_init_msix(s)) { 2228 VMW_WRPRN("Failed to initialize MSI-X, configuration is inconsistent."); 2229 } 2230 2231 if (!vmxnet3_init_msi(s)) { 2232 VMW_WRPRN("Failed to initialize MSI, configuration is inconsistent."); 2233 } 2234 2235 vmxnet3_net_init(s); 2236 2237 register_savevm(dev, "vmxnet3-msix", -1, 1, 2238 vmxnet3_msix_save, vmxnet3_msix_load, s); 2239 } 2240 2241 static void vmxnet3_instance_init(Object *obj) 2242 { 2243 VMXNET3State *s = VMXNET3(obj); 2244 device_add_bootindex_property(obj, &s->conf.bootindex, 2245 "bootindex", "/ethernet-phy@0", 2246 DEVICE(obj), NULL); 2247 } 2248 2249 static void vmxnet3_pci_uninit(PCIDevice *pci_dev) 2250 { 2251 DeviceState *dev = DEVICE(pci_dev); 2252 VMXNET3State *s = VMXNET3(pci_dev); 2253 2254 VMW_CBPRN("Starting uninit..."); 2255 2256 unregister_savevm(dev, "vmxnet3-msix", s); 2257 2258 vmxnet3_net_uninit(s); 2259 2260 vmxnet3_cleanup_msix(s); 2261 2262 vmxnet3_cleanup_msi(s); 2263 } 2264 2265 static void vmxnet3_qdev_reset(DeviceState *dev) 2266 { 2267 PCIDevice *d = PCI_DEVICE(dev); 2268 VMXNET3State *s = VMXNET3(d); 2269 2270 VMW_CBPRN("Starting QDEV reset..."); 2271 vmxnet3_reset(s); 2272 } 2273 2274 static bool vmxnet3_mc_list_needed(void *opaque) 2275 { 2276 return true; 2277 } 2278 2279 static int vmxnet3_mcast_list_pre_load(void *opaque) 2280 { 2281 VMXNET3State *s = opaque; 2282 2283 s->mcast_list = g_malloc(s->mcast_list_buff_size); 2284 2285 return 0; 2286 } 2287 2288 2289 static void vmxnet3_pre_save(void *opaque) 2290 { 2291 VMXNET3State *s = opaque; 2292 2293 s->mcast_list_buff_size = s->mcast_list_len * sizeof(MACAddr); 2294 } 2295 2296 static const VMStateDescription vmxstate_vmxnet3_mcast_list = { 2297 .name = "vmxnet3/mcast_list", 2298 .version_id = 1, 2299 .minimum_version_id = 1, 2300 .pre_load = vmxnet3_mcast_list_pre_load, 2301 .needed = vmxnet3_mc_list_needed, 2302 .fields = (VMStateField[]) { 2303 VMSTATE_VBUFFER_UINT32(mcast_list, VMXNET3State, 0, NULL, 0, 2304 mcast_list_buff_size), 2305 VMSTATE_END_OF_LIST() 2306 } 2307 }; 2308 2309 static void vmxnet3_get_ring_from_file(QEMUFile *f, Vmxnet3Ring *r) 2310 { 2311 r->pa = qemu_get_be64(f); 2312 r->size = qemu_get_be32(f); 2313 r->cell_size = qemu_get_be32(f); 2314 r->next = qemu_get_be32(f); 2315 r->gen = qemu_get_byte(f); 2316 } 2317 2318 static void vmxnet3_put_ring_to_file(QEMUFile *f, Vmxnet3Ring *r) 2319 { 2320 qemu_put_be64(f, r->pa); 2321 qemu_put_be32(f, r->size); 2322 qemu_put_be32(f, r->cell_size); 2323 qemu_put_be32(f, r->next); 2324 qemu_put_byte(f, r->gen); 2325 } 2326 2327 static void vmxnet3_get_tx_stats_from_file(QEMUFile *f, 2328 struct UPT1_TxStats *tx_stat) 2329 { 2330 tx_stat->TSOPktsTxOK = qemu_get_be64(f); 2331 tx_stat->TSOBytesTxOK = qemu_get_be64(f); 2332 tx_stat->ucastPktsTxOK = qemu_get_be64(f); 2333 tx_stat->ucastBytesTxOK = qemu_get_be64(f); 2334 tx_stat->mcastPktsTxOK = qemu_get_be64(f); 2335 tx_stat->mcastBytesTxOK = qemu_get_be64(f); 2336 tx_stat->bcastPktsTxOK = qemu_get_be64(f); 2337 tx_stat->bcastBytesTxOK = qemu_get_be64(f); 2338 tx_stat->pktsTxError = qemu_get_be64(f); 2339 tx_stat->pktsTxDiscard = qemu_get_be64(f); 2340 } 2341 2342 static void vmxnet3_put_tx_stats_to_file(QEMUFile *f, 2343 struct UPT1_TxStats *tx_stat) 2344 { 2345 qemu_put_be64(f, tx_stat->TSOPktsTxOK); 2346 qemu_put_be64(f, tx_stat->TSOBytesTxOK); 2347 qemu_put_be64(f, tx_stat->ucastPktsTxOK); 2348 qemu_put_be64(f, tx_stat->ucastBytesTxOK); 2349 qemu_put_be64(f, tx_stat->mcastPktsTxOK); 2350 qemu_put_be64(f, tx_stat->mcastBytesTxOK); 2351 qemu_put_be64(f, tx_stat->bcastPktsTxOK); 2352 qemu_put_be64(f, tx_stat->bcastBytesTxOK); 2353 qemu_put_be64(f, tx_stat->pktsTxError); 2354 qemu_put_be64(f, tx_stat->pktsTxDiscard); 2355 } 2356 2357 static int vmxnet3_get_txq_descr(QEMUFile *f, void *pv, size_t size) 2358 { 2359 Vmxnet3TxqDescr *r = pv; 2360 2361 vmxnet3_get_ring_from_file(f, &r->tx_ring); 2362 vmxnet3_get_ring_from_file(f, &r->comp_ring); 2363 r->intr_idx = qemu_get_byte(f); 2364 r->tx_stats_pa = qemu_get_be64(f); 2365 2366 vmxnet3_get_tx_stats_from_file(f, &r->txq_stats); 2367 2368 return 0; 2369 } 2370 2371 static void vmxnet3_put_txq_descr(QEMUFile *f, void *pv, size_t size) 2372 { 2373 Vmxnet3TxqDescr *r = pv; 2374 2375 vmxnet3_put_ring_to_file(f, &r->tx_ring); 2376 vmxnet3_put_ring_to_file(f, &r->comp_ring); 2377 qemu_put_byte(f, r->intr_idx); 2378 qemu_put_be64(f, r->tx_stats_pa); 2379 vmxnet3_put_tx_stats_to_file(f, &r->txq_stats); 2380 } 2381 2382 static const VMStateInfo txq_descr_info = { 2383 .name = "txq_descr", 2384 .get = vmxnet3_get_txq_descr, 2385 .put = vmxnet3_put_txq_descr 2386 }; 2387 2388 static void vmxnet3_get_rx_stats_from_file(QEMUFile *f, 2389 struct UPT1_RxStats *rx_stat) 2390 { 2391 rx_stat->LROPktsRxOK = qemu_get_be64(f); 2392 rx_stat->LROBytesRxOK = qemu_get_be64(f); 2393 rx_stat->ucastPktsRxOK = qemu_get_be64(f); 2394 rx_stat->ucastBytesRxOK = qemu_get_be64(f); 2395 rx_stat->mcastPktsRxOK = qemu_get_be64(f); 2396 rx_stat->mcastBytesRxOK = qemu_get_be64(f); 2397 rx_stat->bcastPktsRxOK = qemu_get_be64(f); 2398 rx_stat->bcastBytesRxOK = qemu_get_be64(f); 2399 rx_stat->pktsRxOutOfBuf = qemu_get_be64(f); 2400 rx_stat->pktsRxError = qemu_get_be64(f); 2401 } 2402 2403 static void vmxnet3_put_rx_stats_to_file(QEMUFile *f, 2404 struct UPT1_RxStats *rx_stat) 2405 { 2406 qemu_put_be64(f, rx_stat->LROPktsRxOK); 2407 qemu_put_be64(f, rx_stat->LROBytesRxOK); 2408 qemu_put_be64(f, rx_stat->ucastPktsRxOK); 2409 qemu_put_be64(f, rx_stat->ucastBytesRxOK); 2410 qemu_put_be64(f, rx_stat->mcastPktsRxOK); 2411 qemu_put_be64(f, rx_stat->mcastBytesRxOK); 2412 qemu_put_be64(f, rx_stat->bcastPktsRxOK); 2413 qemu_put_be64(f, rx_stat->bcastBytesRxOK); 2414 qemu_put_be64(f, rx_stat->pktsRxOutOfBuf); 2415 qemu_put_be64(f, rx_stat->pktsRxError); 2416 } 2417 2418 static int vmxnet3_get_rxq_descr(QEMUFile *f, void *pv, size_t size) 2419 { 2420 Vmxnet3RxqDescr *r = pv; 2421 int i; 2422 2423 for (i = 0; i < VMXNET3_RX_RINGS_PER_QUEUE; i++) { 2424 vmxnet3_get_ring_from_file(f, &r->rx_ring[i]); 2425 } 2426 2427 vmxnet3_get_ring_from_file(f, &r->comp_ring); 2428 r->intr_idx = qemu_get_byte(f); 2429 r->rx_stats_pa = qemu_get_be64(f); 2430 2431 vmxnet3_get_rx_stats_from_file(f, &r->rxq_stats); 2432 2433 return 0; 2434 } 2435 2436 static void vmxnet3_put_rxq_descr(QEMUFile *f, void *pv, size_t size) 2437 { 2438 Vmxnet3RxqDescr *r = pv; 2439 int i; 2440 2441 for (i = 0; i < VMXNET3_RX_RINGS_PER_QUEUE; i++) { 2442 vmxnet3_put_ring_to_file(f, &r->rx_ring[i]); 2443 } 2444 2445 vmxnet3_put_ring_to_file(f, &r->comp_ring); 2446 qemu_put_byte(f, r->intr_idx); 2447 qemu_put_be64(f, r->rx_stats_pa); 2448 vmxnet3_put_rx_stats_to_file(f, &r->rxq_stats); 2449 } 2450 2451 static int vmxnet3_post_load(void *opaque, int version_id) 2452 { 2453 VMXNET3State *s = opaque; 2454 PCIDevice *d = PCI_DEVICE(s); 2455 2456 vmxnet_tx_pkt_init(&s->tx_pkt, s->max_tx_frags, s->peer_has_vhdr); 2457 vmxnet_rx_pkt_init(&s->rx_pkt, s->peer_has_vhdr); 2458 2459 if (s->msix_used) { 2460 if (!vmxnet3_use_msix_vectors(s, VMXNET3_MAX_INTRS)) { 2461 VMW_WRPRN("Failed to re-use MSI-X vectors"); 2462 msix_uninit(d, &s->msix_bar, &s->msix_bar); 2463 s->msix_used = false; 2464 return -1; 2465 } 2466 } 2467 2468 vmxnet3_validate_queues(s); 2469 vmxnet3_validate_interrupts(s); 2470 2471 return 0; 2472 } 2473 2474 static const VMStateInfo rxq_descr_info = { 2475 .name = "rxq_descr", 2476 .get = vmxnet3_get_rxq_descr, 2477 .put = vmxnet3_put_rxq_descr 2478 }; 2479 2480 static int vmxnet3_get_int_state(QEMUFile *f, void *pv, size_t size) 2481 { 2482 Vmxnet3IntState *r = pv; 2483 2484 r->is_masked = qemu_get_byte(f); 2485 r->is_pending = qemu_get_byte(f); 2486 r->is_asserted = qemu_get_byte(f); 2487 2488 return 0; 2489 } 2490 2491 static void vmxnet3_put_int_state(QEMUFile *f, void *pv, size_t size) 2492 { 2493 Vmxnet3IntState *r = pv; 2494 2495 qemu_put_byte(f, r->is_masked); 2496 qemu_put_byte(f, r->is_pending); 2497 qemu_put_byte(f, r->is_asserted); 2498 } 2499 2500 static const VMStateInfo int_state_info = { 2501 .name = "int_state", 2502 .get = vmxnet3_get_int_state, 2503 .put = vmxnet3_put_int_state 2504 }; 2505 2506 static const VMStateDescription vmstate_vmxnet3 = { 2507 .name = "vmxnet3", 2508 .version_id = 1, 2509 .minimum_version_id = 1, 2510 .pre_save = vmxnet3_pre_save, 2511 .post_load = vmxnet3_post_load, 2512 .fields = (VMStateField[]) { 2513 VMSTATE_PCI_DEVICE(parent_obj, VMXNET3State), 2514 VMSTATE_BOOL(rx_packets_compound, VMXNET3State), 2515 VMSTATE_BOOL(rx_vlan_stripping, VMXNET3State), 2516 VMSTATE_BOOL(lro_supported, VMXNET3State), 2517 VMSTATE_UINT32(rx_mode, VMXNET3State), 2518 VMSTATE_UINT32(mcast_list_len, VMXNET3State), 2519 VMSTATE_UINT32(mcast_list_buff_size, VMXNET3State), 2520 VMSTATE_UINT32_ARRAY(vlan_table, VMXNET3State, VMXNET3_VFT_SIZE), 2521 VMSTATE_UINT32(mtu, VMXNET3State), 2522 VMSTATE_UINT16(max_rx_frags, VMXNET3State), 2523 VMSTATE_UINT32(max_tx_frags, VMXNET3State), 2524 VMSTATE_UINT8(event_int_idx, VMXNET3State), 2525 VMSTATE_BOOL(auto_int_masking, VMXNET3State), 2526 VMSTATE_UINT8(txq_num, VMXNET3State), 2527 VMSTATE_UINT8(rxq_num, VMXNET3State), 2528 VMSTATE_UINT32(device_active, VMXNET3State), 2529 VMSTATE_UINT32(last_command, VMXNET3State), 2530 VMSTATE_UINT32(link_status_and_speed, VMXNET3State), 2531 VMSTATE_UINT32(temp_mac, VMXNET3State), 2532 VMSTATE_UINT64(drv_shmem, VMXNET3State), 2533 VMSTATE_UINT64(temp_shared_guest_driver_memory, VMXNET3State), 2534 2535 VMSTATE_ARRAY(txq_descr, VMXNET3State, 2536 VMXNET3_DEVICE_MAX_TX_QUEUES, 0, txq_descr_info, 2537 Vmxnet3TxqDescr), 2538 VMSTATE_ARRAY(rxq_descr, VMXNET3State, 2539 VMXNET3_DEVICE_MAX_RX_QUEUES, 0, rxq_descr_info, 2540 Vmxnet3RxqDescr), 2541 VMSTATE_ARRAY(interrupt_states, VMXNET3State, VMXNET3_MAX_INTRS, 2542 0, int_state_info, Vmxnet3IntState), 2543 2544 VMSTATE_END_OF_LIST() 2545 }, 2546 .subsections = (const VMStateDescription*[]) { 2547 &vmxstate_vmxnet3_mcast_list, 2548 NULL 2549 } 2550 }; 2551 2552 static Property vmxnet3_properties[] = { 2553 DEFINE_NIC_PROPERTIES(VMXNET3State, conf), 2554 DEFINE_PROP_END_OF_LIST(), 2555 }; 2556 2557 static void vmxnet3_class_init(ObjectClass *class, void *data) 2558 { 2559 DeviceClass *dc = DEVICE_CLASS(class); 2560 PCIDeviceClass *c = PCI_DEVICE_CLASS(class); 2561 2562 c->realize = vmxnet3_pci_realize; 2563 c->exit = vmxnet3_pci_uninit; 2564 c->vendor_id = PCI_VENDOR_ID_VMWARE; 2565 c->device_id = PCI_DEVICE_ID_VMWARE_VMXNET3; 2566 c->revision = PCI_DEVICE_ID_VMWARE_VMXNET3_REVISION; 2567 c->class_id = PCI_CLASS_NETWORK_ETHERNET; 2568 c->subsystem_vendor_id = PCI_VENDOR_ID_VMWARE; 2569 c->subsystem_id = PCI_DEVICE_ID_VMWARE_VMXNET3; 2570 dc->desc = "VMWare Paravirtualized Ethernet v3"; 2571 dc->reset = vmxnet3_qdev_reset; 2572 dc->vmsd = &vmstate_vmxnet3; 2573 dc->props = vmxnet3_properties; 2574 set_bit(DEVICE_CATEGORY_NETWORK, dc->categories); 2575 } 2576 2577 static const TypeInfo vmxnet3_info = { 2578 .name = TYPE_VMXNET3, 2579 .parent = TYPE_PCI_DEVICE, 2580 .instance_size = sizeof(VMXNET3State), 2581 .class_init = vmxnet3_class_init, 2582 .instance_init = vmxnet3_instance_init, 2583 }; 2584 2585 static void vmxnet3_register_types(void) 2586 { 2587 VMW_CBPRN("vmxnet3_register_types called..."); 2588 type_register_static(&vmxnet3_info); 2589 } 2590 2591 type_init(vmxnet3_register_types) 2592