xref: /openbmc/qemu/hw/net/vmxnet3.c (revision 1b111dc1)
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 
38 #define VMXNET3_BAR0_IDX      (0)
39 #define VMXNET3_BAR1_IDX      (1)
40 #define VMXNET3_MSIX_BAR_IDX  (2)
41 
42 #define VMXNET3_OFF_MSIX_TABLE (0x000)
43 #define VMXNET3_OFF_MSIX_PBA   (0x800)
44 
45 /* Link speed in Mbps should be shifted by 16 */
46 #define VMXNET3_LINK_SPEED      (1000 << 16)
47 
48 /* Link status: 1 - up, 0 - down. */
49 #define VMXNET3_LINK_STATUS_UP  0x1
50 
51 /* Least significant bit should be set for revision and version */
52 #define VMXNET3_DEVICE_VERSION    0x1
53 #define VMXNET3_DEVICE_REVISION   0x1
54 
55 /* Macros for rings descriptors access */
56 #define VMXNET3_READ_TX_QUEUE_DESCR8(dpa, field) \
57     (vmw_shmem_ld8(dpa + offsetof(struct Vmxnet3_TxQueueDesc, field)))
58 
59 #define VMXNET3_WRITE_TX_QUEUE_DESCR8(dpa, field, value) \
60     (vmw_shmem_st8(dpa + offsetof(struct Vmxnet3_TxQueueDesc, field, value)))
61 
62 #define VMXNET3_READ_TX_QUEUE_DESCR32(dpa, field) \
63     (vmw_shmem_ld32(dpa + offsetof(struct Vmxnet3_TxQueueDesc, field)))
64 
65 #define VMXNET3_WRITE_TX_QUEUE_DESCR32(dpa, field, value) \
66     (vmw_shmem_st32(dpa + offsetof(struct Vmxnet3_TxQueueDesc, field), value))
67 
68 #define VMXNET3_READ_TX_QUEUE_DESCR64(dpa, field) \
69     (vmw_shmem_ld64(dpa + offsetof(struct Vmxnet3_TxQueueDesc, field)))
70 
71 #define VMXNET3_WRITE_TX_QUEUE_DESCR64(dpa, field, value) \
72     (vmw_shmem_st64(dpa + offsetof(struct Vmxnet3_TxQueueDesc, field), value))
73 
74 #define VMXNET3_READ_RX_QUEUE_DESCR64(dpa, field) \
75     (vmw_shmem_ld64(dpa + offsetof(struct Vmxnet3_RxQueueDesc, field)))
76 
77 #define VMXNET3_READ_RX_QUEUE_DESCR32(dpa, field) \
78     (vmw_shmem_ld32(dpa + offsetof(struct Vmxnet3_RxQueueDesc, field)))
79 
80 #define VMXNET3_WRITE_RX_QUEUE_DESCR64(dpa, field, value) \
81     (vmw_shmem_st64(dpa + offsetof(struct Vmxnet3_RxQueueDesc, field), value))
82 
83 #define VMXNET3_WRITE_RX_QUEUE_DESCR8(dpa, field, value) \
84     (vmw_shmem_st8(dpa + offsetof(struct Vmxnet3_RxQueueDesc, field), value))
85 
86 /* Macros for guest driver shared area access */
87 #define VMXNET3_READ_DRV_SHARED64(shpa, field) \
88     (vmw_shmem_ld64(shpa + offsetof(struct Vmxnet3_DriverShared, field)))
89 
90 #define VMXNET3_READ_DRV_SHARED32(shpa, field) \
91     (vmw_shmem_ld32(shpa + offsetof(struct Vmxnet3_DriverShared, field)))
92 
93 #define VMXNET3_WRITE_DRV_SHARED32(shpa, field, val) \
94     (vmw_shmem_st32(shpa + offsetof(struct Vmxnet3_DriverShared, field), val))
95 
96 #define VMXNET3_READ_DRV_SHARED16(shpa, field) \
97     (vmw_shmem_ld16(shpa + offsetof(struct Vmxnet3_DriverShared, field)))
98 
99 #define VMXNET3_READ_DRV_SHARED8(shpa, field) \
100     (vmw_shmem_ld8(shpa + offsetof(struct Vmxnet3_DriverShared, field)))
101 
102 #define VMXNET3_READ_DRV_SHARED(shpa, field, b, l) \
103     (vmw_shmem_read(shpa + offsetof(struct Vmxnet3_DriverShared, field), b, l))
104 
105 #define VMXNET_FLAG_IS_SET(field, flag) (((field) & (flag)) == (flag))
106 
107 #define TYPE_VMXNET3 "vmxnet3"
108 #define VMXNET3(obj) OBJECT_CHECK(VMXNET3State, (obj), TYPE_VMXNET3)
109 
110 /* Cyclic ring abstraction */
111 typedef struct {
112     hwaddr pa;
113     size_t size;
114     size_t cell_size;
115     size_t next;
116     uint8_t gen;
117 } Vmxnet3Ring;
118 
119 static inline void vmxnet3_ring_init(Vmxnet3Ring *ring,
120                                      hwaddr pa,
121                                      size_t size,
122                                      size_t cell_size,
123                                      bool zero_region)
124 {
125     ring->pa = pa;
126     ring->size = size;
127     ring->cell_size = cell_size;
128     ring->gen = VMXNET3_INIT_GEN;
129     ring->next = 0;
130 
131     if (zero_region) {
132         vmw_shmem_set(pa, 0, size * cell_size);
133     }
134 }
135 
136 #define VMXNET3_RING_DUMP(macro, ring_name, ridx, r)                         \
137     macro("%s#%d: base %" PRIx64 " size %lu cell_size %lu gen %d next %lu",  \
138           (ring_name), (ridx),                                               \
139           (r)->pa, (r)->size, (r)->cell_size, (r)->gen, (r)->next)
140 
141 static inline void vmxnet3_ring_inc(Vmxnet3Ring *ring)
142 {
143     if (++ring->next >= ring->size) {
144         ring->next = 0;
145         ring->gen ^= 1;
146     }
147 }
148 
149 static inline void vmxnet3_ring_dec(Vmxnet3Ring *ring)
150 {
151     if (ring->next-- == 0) {
152         ring->next = ring->size - 1;
153         ring->gen ^= 1;
154     }
155 }
156 
157 static inline hwaddr vmxnet3_ring_curr_cell_pa(Vmxnet3Ring *ring)
158 {
159     return ring->pa + ring->next * ring->cell_size;
160 }
161 
162 static inline void vmxnet3_ring_read_curr_cell(Vmxnet3Ring *ring, void *buff)
163 {
164     vmw_shmem_read(vmxnet3_ring_curr_cell_pa(ring), buff, ring->cell_size);
165 }
166 
167 static inline void vmxnet3_ring_write_curr_cell(Vmxnet3Ring *ring, void *buff)
168 {
169     vmw_shmem_write(vmxnet3_ring_curr_cell_pa(ring), buff, ring->cell_size);
170 }
171 
172 static inline size_t vmxnet3_ring_curr_cell_idx(Vmxnet3Ring *ring)
173 {
174     return ring->next;
175 }
176 
177 static inline uint8_t vmxnet3_ring_curr_gen(Vmxnet3Ring *ring)
178 {
179     return ring->gen;
180 }
181 
182 /* Debug trace-related functions */
183 static inline void
184 vmxnet3_dump_tx_descr(struct Vmxnet3_TxDesc *descr)
185 {
186     VMW_PKPRN("TX DESCR: "
187               "addr %" PRIx64 ", len: %d, gen: %d, rsvd: %d, "
188               "dtype: %d, ext1: %d, msscof: %d, hlen: %d, om: %d, "
189               "eop: %d, cq: %d, ext2: %d, ti: %d, tci: %d",
190               le64_to_cpu(descr->addr), descr->len, descr->gen, descr->rsvd,
191               descr->dtype, descr->ext1, descr->msscof, descr->hlen, descr->om,
192               descr->eop, descr->cq, descr->ext2, descr->ti, descr->tci);
193 }
194 
195 static inline void
196 vmxnet3_dump_virt_hdr(struct virtio_net_hdr *vhdr)
197 {
198     VMW_PKPRN("VHDR: flags 0x%x, gso_type: 0x%x, hdr_len: %d, gso_size: %d, "
199               "csum_start: %d, csum_offset: %d",
200               vhdr->flags, vhdr->gso_type, vhdr->hdr_len, vhdr->gso_size,
201               vhdr->csum_start, vhdr->csum_offset);
202 }
203 
204 static inline void
205 vmxnet3_dump_rx_descr(struct Vmxnet3_RxDesc *descr)
206 {
207     VMW_PKPRN("RX DESCR: addr %" PRIx64 ", len: %d, gen: %d, rsvd: %d, "
208               "dtype: %d, ext1: %d, btype: %d",
209               le64_to_cpu(descr->addr), descr->len, descr->gen,
210               descr->rsvd, descr->dtype, descr->ext1, descr->btype);
211 }
212 
213 /* Device state and helper functions */
214 #define VMXNET3_RX_RINGS_PER_QUEUE (2)
215 
216 typedef struct {
217     Vmxnet3Ring tx_ring;
218     Vmxnet3Ring comp_ring;
219 
220     uint8_t intr_idx;
221     hwaddr tx_stats_pa;
222     struct UPT1_TxStats txq_stats;
223 } Vmxnet3TxqDescr;
224 
225 typedef struct {
226     Vmxnet3Ring rx_ring[VMXNET3_RX_RINGS_PER_QUEUE];
227     Vmxnet3Ring comp_ring;
228     uint8_t intr_idx;
229     hwaddr rx_stats_pa;
230     struct UPT1_RxStats rxq_stats;
231 } Vmxnet3RxqDescr;
232 
233 typedef struct {
234     bool is_masked;
235     bool is_pending;
236     bool is_asserted;
237 } Vmxnet3IntState;
238 
239 typedef struct {
240         PCIDevice parent_obj;
241         NICState *nic;
242         NICConf conf;
243         MemoryRegion bar0;
244         MemoryRegion bar1;
245         MemoryRegion msix_bar;
246 
247         Vmxnet3RxqDescr rxq_descr[VMXNET3_DEVICE_MAX_RX_QUEUES];
248         Vmxnet3TxqDescr txq_descr[VMXNET3_DEVICE_MAX_TX_QUEUES];
249 
250         /* Whether MSI-X support was installed successfully */
251         bool msix_used;
252         /* Whether MSI support was installed successfully */
253         bool msi_used;
254         hwaddr drv_shmem;
255         hwaddr temp_shared_guest_driver_memory;
256 
257         uint8_t txq_num;
258 
259         /* This boolean tells whether RX packet being indicated has to */
260         /* be split into head and body chunks from different RX rings  */
261         bool rx_packets_compound;
262 
263         bool rx_vlan_stripping;
264         bool lro_supported;
265 
266         uint8_t rxq_num;
267 
268         /* Network MTU */
269         uint32_t mtu;
270 
271         /* Maximum number of fragments for indicated TX packets */
272         uint32_t max_tx_frags;
273 
274         /* Maximum number of fragments for indicated RX packets */
275         uint16_t max_rx_frags;
276 
277         /* Index for events interrupt */
278         uint8_t event_int_idx;
279 
280         /* Whether automatic interrupts masking enabled */
281         bool auto_int_masking;
282 
283         bool peer_has_vhdr;
284 
285         /* TX packets to QEMU interface */
286         struct VmxnetTxPkt *tx_pkt;
287         uint32_t offload_mode;
288         uint32_t cso_or_gso_size;
289         uint16_t tci;
290         bool needs_vlan;
291 
292         struct VmxnetRxPkt *rx_pkt;
293 
294         bool tx_sop;
295         bool skip_current_tx_pkt;
296 
297         uint32_t device_active;
298         uint32_t last_command;
299 
300         uint32_t link_status_and_speed;
301 
302         Vmxnet3IntState interrupt_states[VMXNET3_MAX_INTRS];
303 
304         uint32_t temp_mac;   /* To store the low part first */
305 
306         MACAddr perm_mac;
307         uint32_t vlan_table[VMXNET3_VFT_SIZE];
308         uint32_t rx_mode;
309         MACAddr *mcast_list;
310         uint32_t mcast_list_len;
311         uint32_t mcast_list_buff_size; /* needed for live migration. */
312 } VMXNET3State;
313 
314 /* Interrupt management */
315 
316 /*
317  *This function returns sign whether interrupt line is in asserted state
318  * This depends on the type of interrupt used. For INTX interrupt line will
319  * be asserted until explicit deassertion, for MSI(X) interrupt line will
320  * be deasserted automatically due to notification semantics of the MSI(X)
321  * interrupts
322  */
323 static bool _vmxnet3_assert_interrupt_line(VMXNET3State *s, uint32_t int_idx)
324 {
325     PCIDevice *d = PCI_DEVICE(s);
326 
327     if (s->msix_used && msix_enabled(d)) {
328         VMW_IRPRN("Sending MSI-X notification for vector %u", int_idx);
329         msix_notify(d, int_idx);
330         return false;
331     }
332     if (s->msi_used && msi_enabled(d)) {
333         VMW_IRPRN("Sending MSI notification for vector %u", int_idx);
334         msi_notify(d, int_idx);
335         return false;
336     }
337 
338     VMW_IRPRN("Asserting line for interrupt %u", int_idx);
339     pci_irq_assert(d);
340     return true;
341 }
342 
343 static void _vmxnet3_deassert_interrupt_line(VMXNET3State *s, int lidx)
344 {
345     PCIDevice *d = PCI_DEVICE(s);
346 
347     /*
348      * This function should never be called for MSI(X) interrupts
349      * because deassertion never required for message interrupts
350      */
351     assert(!s->msix_used || !msix_enabled(d));
352     /*
353      * This function should never be called for MSI(X) interrupts
354      * because deassertion never required for message interrupts
355      */
356     assert(!s->msi_used || !msi_enabled(d));
357 
358     VMW_IRPRN("Deasserting line for interrupt %u", lidx);
359     pci_irq_deassert(d);
360 }
361 
362 static void vmxnet3_update_interrupt_line_state(VMXNET3State *s, int lidx)
363 {
364     if (!s->interrupt_states[lidx].is_pending &&
365        s->interrupt_states[lidx].is_asserted) {
366         VMW_IRPRN("New interrupt line state for index %d is DOWN", lidx);
367         _vmxnet3_deassert_interrupt_line(s, lidx);
368         s->interrupt_states[lidx].is_asserted = false;
369         return;
370     }
371 
372     if (s->interrupt_states[lidx].is_pending &&
373        !s->interrupt_states[lidx].is_masked &&
374        !s->interrupt_states[lidx].is_asserted) {
375         VMW_IRPRN("New interrupt line state for index %d is UP", lidx);
376         s->interrupt_states[lidx].is_asserted =
377             _vmxnet3_assert_interrupt_line(s, lidx);
378         s->interrupt_states[lidx].is_pending = false;
379         return;
380     }
381 }
382 
383 static void vmxnet3_trigger_interrupt(VMXNET3State *s, int lidx)
384 {
385     PCIDevice *d = PCI_DEVICE(s);
386     s->interrupt_states[lidx].is_pending = true;
387     vmxnet3_update_interrupt_line_state(s, lidx);
388 
389     if (s->msix_used && msix_enabled(d) && s->auto_int_masking) {
390         goto do_automask;
391     }
392 
393     if (s->msi_used && msi_enabled(d) && s->auto_int_masking) {
394         goto do_automask;
395     }
396 
397     return;
398 
399 do_automask:
400     s->interrupt_states[lidx].is_masked = true;
401     vmxnet3_update_interrupt_line_state(s, lidx);
402 }
403 
404 static bool vmxnet3_interrupt_asserted(VMXNET3State *s, int lidx)
405 {
406     return s->interrupt_states[lidx].is_asserted;
407 }
408 
409 static void vmxnet3_clear_interrupt(VMXNET3State *s, int int_idx)
410 {
411     s->interrupt_states[int_idx].is_pending = false;
412     if (s->auto_int_masking) {
413         s->interrupt_states[int_idx].is_masked = true;
414     }
415     vmxnet3_update_interrupt_line_state(s, int_idx);
416 }
417 
418 static void
419 vmxnet3_on_interrupt_mask_changed(VMXNET3State *s, int lidx, bool is_masked)
420 {
421     s->interrupt_states[lidx].is_masked = is_masked;
422     vmxnet3_update_interrupt_line_state(s, lidx);
423 }
424 
425 static bool vmxnet3_verify_driver_magic(hwaddr dshmem)
426 {
427     return (VMXNET3_READ_DRV_SHARED32(dshmem, magic) == VMXNET3_REV1_MAGIC);
428 }
429 
430 #define VMXNET3_GET_BYTE(x, byte_num) (((x) >> (byte_num)*8) & 0xFF)
431 #define VMXNET3_MAKE_BYTE(byte_num, val) \
432     (((uint32_t)((val) & 0xFF)) << (byte_num)*8)
433 
434 static void vmxnet3_set_variable_mac(VMXNET3State *s, uint32_t h, uint32_t l)
435 {
436     s->conf.macaddr.a[0] = VMXNET3_GET_BYTE(l,  0);
437     s->conf.macaddr.a[1] = VMXNET3_GET_BYTE(l,  1);
438     s->conf.macaddr.a[2] = VMXNET3_GET_BYTE(l,  2);
439     s->conf.macaddr.a[3] = VMXNET3_GET_BYTE(l,  3);
440     s->conf.macaddr.a[4] = VMXNET3_GET_BYTE(h, 0);
441     s->conf.macaddr.a[5] = VMXNET3_GET_BYTE(h, 1);
442 
443     VMW_CFPRN("Variable MAC: " VMXNET_MF, VMXNET_MA(s->conf.macaddr.a));
444 
445     qemu_format_nic_info_str(qemu_get_queue(s->nic), s->conf.macaddr.a);
446 }
447 
448 static uint64_t vmxnet3_get_mac_low(MACAddr *addr)
449 {
450     return VMXNET3_MAKE_BYTE(0, addr->a[0]) |
451            VMXNET3_MAKE_BYTE(1, addr->a[1]) |
452            VMXNET3_MAKE_BYTE(2, addr->a[2]) |
453            VMXNET3_MAKE_BYTE(3, addr->a[3]);
454 }
455 
456 static uint64_t vmxnet3_get_mac_high(MACAddr *addr)
457 {
458     return VMXNET3_MAKE_BYTE(0, addr->a[4]) |
459            VMXNET3_MAKE_BYTE(1, addr->a[5]);
460 }
461 
462 static void
463 vmxnet3_inc_tx_consumption_counter(VMXNET3State *s, int qidx)
464 {
465     vmxnet3_ring_inc(&s->txq_descr[qidx].tx_ring);
466 }
467 
468 static inline void
469 vmxnet3_inc_rx_consumption_counter(VMXNET3State *s, int qidx, int ridx)
470 {
471     vmxnet3_ring_inc(&s->rxq_descr[qidx].rx_ring[ridx]);
472 }
473 
474 static inline void
475 vmxnet3_inc_tx_completion_counter(VMXNET3State *s, int qidx)
476 {
477     vmxnet3_ring_inc(&s->txq_descr[qidx].comp_ring);
478 }
479 
480 static void
481 vmxnet3_inc_rx_completion_counter(VMXNET3State *s, int qidx)
482 {
483     vmxnet3_ring_inc(&s->rxq_descr[qidx].comp_ring);
484 }
485 
486 static void
487 vmxnet3_dec_rx_completion_counter(VMXNET3State *s, int qidx)
488 {
489     vmxnet3_ring_dec(&s->rxq_descr[qidx].comp_ring);
490 }
491 
492 static void vmxnet3_complete_packet(VMXNET3State *s, int qidx, uint32 tx_ridx)
493 {
494     struct Vmxnet3_TxCompDesc txcq_descr;
495 
496     VMXNET3_RING_DUMP(VMW_RIPRN, "TXC", qidx, &s->txq_descr[qidx].comp_ring);
497 
498     txcq_descr.txdIdx = tx_ridx;
499     txcq_descr.gen = vmxnet3_ring_curr_gen(&s->txq_descr[qidx].comp_ring);
500 
501     vmxnet3_ring_write_curr_cell(&s->txq_descr[qidx].comp_ring, &txcq_descr);
502 
503     /* Flush changes in TX descriptor before changing the counter value */
504     smp_wmb();
505 
506     vmxnet3_inc_tx_completion_counter(s, qidx);
507     vmxnet3_trigger_interrupt(s, s->txq_descr[qidx].intr_idx);
508 }
509 
510 static bool
511 vmxnet3_setup_tx_offloads(VMXNET3State *s)
512 {
513     switch (s->offload_mode) {
514     case VMXNET3_OM_NONE:
515         vmxnet_tx_pkt_build_vheader(s->tx_pkt, false, false, 0);
516         break;
517 
518     case VMXNET3_OM_CSUM:
519         vmxnet_tx_pkt_build_vheader(s->tx_pkt, false, true, 0);
520         VMW_PKPRN("L4 CSO requested\n");
521         break;
522 
523     case VMXNET3_OM_TSO:
524         vmxnet_tx_pkt_build_vheader(s->tx_pkt, true, true,
525             s->cso_or_gso_size);
526         vmxnet_tx_pkt_update_ip_checksums(s->tx_pkt);
527         VMW_PKPRN("GSO offload requested.");
528         break;
529 
530     default:
531         g_assert_not_reached();
532         return false;
533     }
534 
535     return true;
536 }
537 
538 static void
539 vmxnet3_tx_retrieve_metadata(VMXNET3State *s,
540                              const struct Vmxnet3_TxDesc *txd)
541 {
542     s->offload_mode = txd->om;
543     s->cso_or_gso_size = txd->msscof;
544     s->tci = txd->tci;
545     s->needs_vlan = txd->ti;
546 }
547 
548 typedef enum {
549     VMXNET3_PKT_STATUS_OK,
550     VMXNET3_PKT_STATUS_ERROR,
551     VMXNET3_PKT_STATUS_DISCARD,/* only for tx */
552     VMXNET3_PKT_STATUS_OUT_OF_BUF /* only for rx */
553 } Vmxnet3PktStatus;
554 
555 static void
556 vmxnet3_on_tx_done_update_stats(VMXNET3State *s, int qidx,
557     Vmxnet3PktStatus status)
558 {
559     size_t tot_len = vmxnet_tx_pkt_get_total_len(s->tx_pkt);
560     struct UPT1_TxStats *stats = &s->txq_descr[qidx].txq_stats;
561 
562     switch (status) {
563     case VMXNET3_PKT_STATUS_OK:
564         switch (vmxnet_tx_pkt_get_packet_type(s->tx_pkt)) {
565         case ETH_PKT_BCAST:
566             stats->bcastPktsTxOK++;
567             stats->bcastBytesTxOK += tot_len;
568             break;
569         case ETH_PKT_MCAST:
570             stats->mcastPktsTxOK++;
571             stats->mcastBytesTxOK += tot_len;
572             break;
573         case ETH_PKT_UCAST:
574             stats->ucastPktsTxOK++;
575             stats->ucastBytesTxOK += tot_len;
576             break;
577         default:
578             g_assert_not_reached();
579         }
580 
581         if (s->offload_mode == VMXNET3_OM_TSO) {
582             /*
583              * According to VMWARE headers this statistic is a number
584              * of packets after segmentation but since we don't have
585              * this information in QEMU model, the best we can do is to
586              * provide number of non-segmented packets
587              */
588             stats->TSOPktsTxOK++;
589             stats->TSOBytesTxOK += tot_len;
590         }
591         break;
592 
593     case VMXNET3_PKT_STATUS_DISCARD:
594         stats->pktsTxDiscard++;
595         break;
596 
597     case VMXNET3_PKT_STATUS_ERROR:
598         stats->pktsTxError++;
599         break;
600 
601     default:
602         g_assert_not_reached();
603     }
604 }
605 
606 static void
607 vmxnet3_on_rx_done_update_stats(VMXNET3State *s,
608                                 int qidx,
609                                 Vmxnet3PktStatus status)
610 {
611     struct UPT1_RxStats *stats = &s->rxq_descr[qidx].rxq_stats;
612     size_t tot_len = vmxnet_rx_pkt_get_total_len(s->rx_pkt);
613 
614     switch (status) {
615     case VMXNET3_PKT_STATUS_OUT_OF_BUF:
616         stats->pktsRxOutOfBuf++;
617         break;
618 
619     case VMXNET3_PKT_STATUS_ERROR:
620         stats->pktsRxError++;
621         break;
622     case VMXNET3_PKT_STATUS_OK:
623         switch (vmxnet_rx_pkt_get_packet_type(s->rx_pkt)) {
624         case ETH_PKT_BCAST:
625             stats->bcastPktsRxOK++;
626             stats->bcastBytesRxOK += tot_len;
627             break;
628         case ETH_PKT_MCAST:
629             stats->mcastPktsRxOK++;
630             stats->mcastBytesRxOK += tot_len;
631             break;
632         case ETH_PKT_UCAST:
633             stats->ucastPktsRxOK++;
634             stats->ucastBytesRxOK += tot_len;
635             break;
636         default:
637             g_assert_not_reached();
638         }
639 
640         if (tot_len > s->mtu) {
641             stats->LROPktsRxOK++;
642             stats->LROBytesRxOK += tot_len;
643         }
644         break;
645     default:
646         g_assert_not_reached();
647     }
648 }
649 
650 static inline bool
651 vmxnet3_pop_next_tx_descr(VMXNET3State *s,
652                           int qidx,
653                           struct Vmxnet3_TxDesc *txd,
654                           uint32_t *descr_idx)
655 {
656     Vmxnet3Ring *ring = &s->txq_descr[qidx].tx_ring;
657 
658     vmxnet3_ring_read_curr_cell(ring, txd);
659     if (txd->gen == vmxnet3_ring_curr_gen(ring)) {
660         /* Only read after generation field verification */
661         smp_rmb();
662         /* Re-read to be sure we got the latest version */
663         vmxnet3_ring_read_curr_cell(ring, txd);
664         VMXNET3_RING_DUMP(VMW_RIPRN, "TX", qidx, ring);
665         *descr_idx = vmxnet3_ring_curr_cell_idx(ring);
666         vmxnet3_inc_tx_consumption_counter(s, qidx);
667         return true;
668     }
669 
670     return false;
671 }
672 
673 static bool
674 vmxnet3_send_packet(VMXNET3State *s, uint32_t qidx)
675 {
676     Vmxnet3PktStatus status = VMXNET3_PKT_STATUS_OK;
677 
678     if (!vmxnet3_setup_tx_offloads(s)) {
679         status = VMXNET3_PKT_STATUS_ERROR;
680         goto func_exit;
681     }
682 
683     /* debug prints */
684     vmxnet3_dump_virt_hdr(vmxnet_tx_pkt_get_vhdr(s->tx_pkt));
685     vmxnet_tx_pkt_dump(s->tx_pkt);
686 
687     if (!vmxnet_tx_pkt_send(s->tx_pkt, qemu_get_queue(s->nic))) {
688         status = VMXNET3_PKT_STATUS_DISCARD;
689         goto func_exit;
690     }
691 
692 func_exit:
693     vmxnet3_on_tx_done_update_stats(s, qidx, status);
694     return (status == VMXNET3_PKT_STATUS_OK);
695 }
696 
697 static void vmxnet3_process_tx_queue(VMXNET3State *s, int qidx)
698 {
699     struct Vmxnet3_TxDesc txd;
700     uint32_t txd_idx;
701     uint32_t data_len;
702     hwaddr data_pa;
703 
704     for (;;) {
705         if (!vmxnet3_pop_next_tx_descr(s, qidx, &txd, &txd_idx)) {
706             break;
707         }
708 
709         vmxnet3_dump_tx_descr(&txd);
710 
711         if (!s->skip_current_tx_pkt) {
712             data_len = (txd.len > 0) ? txd.len : VMXNET3_MAX_TX_BUF_SIZE;
713             data_pa = le64_to_cpu(txd.addr);
714 
715             if (!vmxnet_tx_pkt_add_raw_fragment(s->tx_pkt,
716                                                 data_pa,
717                                                 data_len)) {
718                 s->skip_current_tx_pkt = true;
719             }
720         }
721 
722         if (s->tx_sop) {
723             vmxnet3_tx_retrieve_metadata(s, &txd);
724             s->tx_sop = false;
725         }
726 
727         if (txd.eop) {
728             if (!s->skip_current_tx_pkt) {
729                 vmxnet_tx_pkt_parse(s->tx_pkt);
730 
731                 if (s->needs_vlan) {
732                     vmxnet_tx_pkt_setup_vlan_header(s->tx_pkt, s->tci);
733                 }
734 
735                 vmxnet3_send_packet(s, qidx);
736             } else {
737                 vmxnet3_on_tx_done_update_stats(s, qidx,
738                                                 VMXNET3_PKT_STATUS_ERROR);
739             }
740 
741             vmxnet3_complete_packet(s, qidx, txd_idx);
742             s->tx_sop = true;
743             s->skip_current_tx_pkt = false;
744             vmxnet_tx_pkt_reset(s->tx_pkt);
745         }
746     }
747 }
748 
749 static inline void
750 vmxnet3_read_next_rx_descr(VMXNET3State *s, int qidx, int ridx,
751                            struct Vmxnet3_RxDesc *dbuf, uint32_t *didx)
752 {
753     Vmxnet3Ring *ring = &s->rxq_descr[qidx].rx_ring[ridx];
754     *didx = vmxnet3_ring_curr_cell_idx(ring);
755     vmxnet3_ring_read_curr_cell(ring, dbuf);
756 }
757 
758 static inline uint8_t
759 vmxnet3_get_rx_ring_gen(VMXNET3State *s, int qidx, int ridx)
760 {
761     return s->rxq_descr[qidx].rx_ring[ridx].gen;
762 }
763 
764 static inline hwaddr
765 vmxnet3_pop_rxc_descr(VMXNET3State *s, int qidx, uint32_t *descr_gen)
766 {
767     uint8_t ring_gen;
768     struct Vmxnet3_RxCompDesc rxcd;
769 
770     hwaddr daddr =
771         vmxnet3_ring_curr_cell_pa(&s->rxq_descr[qidx].comp_ring);
772 
773     cpu_physical_memory_read(daddr, &rxcd, sizeof(struct Vmxnet3_RxCompDesc));
774     ring_gen = vmxnet3_ring_curr_gen(&s->rxq_descr[qidx].comp_ring);
775 
776     if (rxcd.gen != ring_gen) {
777         *descr_gen = ring_gen;
778         vmxnet3_inc_rx_completion_counter(s, qidx);
779         return daddr;
780     }
781 
782     return 0;
783 }
784 
785 static inline void
786 vmxnet3_revert_rxc_descr(VMXNET3State *s, int qidx)
787 {
788     vmxnet3_dec_rx_completion_counter(s, qidx);
789 }
790 
791 #define RXQ_IDX      (0)
792 #define RX_HEAD_BODY_RING (0)
793 #define RX_BODY_ONLY_RING (1)
794 
795 static bool
796 vmxnet3_get_next_head_rx_descr(VMXNET3State *s,
797                                struct Vmxnet3_RxDesc *descr_buf,
798                                uint32_t *descr_idx,
799                                uint32_t *ridx)
800 {
801     for (;;) {
802         uint32_t ring_gen;
803         vmxnet3_read_next_rx_descr(s, RXQ_IDX, RX_HEAD_BODY_RING,
804                                    descr_buf, descr_idx);
805 
806         /* If no more free descriptors - return */
807         ring_gen = vmxnet3_get_rx_ring_gen(s, RXQ_IDX, RX_HEAD_BODY_RING);
808         if (descr_buf->gen != ring_gen) {
809             return false;
810         }
811 
812         /* Only read after generation field verification */
813         smp_rmb();
814         /* Re-read to be sure we got the latest version */
815         vmxnet3_read_next_rx_descr(s, RXQ_IDX, RX_HEAD_BODY_RING,
816                                    descr_buf, descr_idx);
817 
818         /* Mark current descriptor as used/skipped */
819         vmxnet3_inc_rx_consumption_counter(s, RXQ_IDX, RX_HEAD_BODY_RING);
820 
821         /* If this is what we are looking for - return */
822         if (descr_buf->btype == VMXNET3_RXD_BTYPE_HEAD) {
823             *ridx = RX_HEAD_BODY_RING;
824             return true;
825         }
826     }
827 }
828 
829 static bool
830 vmxnet3_get_next_body_rx_descr(VMXNET3State *s,
831                                struct Vmxnet3_RxDesc *d,
832                                uint32_t *didx,
833                                uint32_t *ridx)
834 {
835     vmxnet3_read_next_rx_descr(s, RXQ_IDX, RX_HEAD_BODY_RING, d, didx);
836 
837     /* Try to find corresponding descriptor in head/body ring */
838     if (d->gen == vmxnet3_get_rx_ring_gen(s, RXQ_IDX, RX_HEAD_BODY_RING)) {
839         /* Only read after generation field verification */
840         smp_rmb();
841         /* Re-read to be sure we got the latest version */
842         vmxnet3_read_next_rx_descr(s, RXQ_IDX, RX_HEAD_BODY_RING, d, didx);
843         if (d->btype == VMXNET3_RXD_BTYPE_BODY) {
844             vmxnet3_inc_rx_consumption_counter(s, RXQ_IDX, RX_HEAD_BODY_RING);
845             *ridx = RX_HEAD_BODY_RING;
846             return true;
847         }
848     }
849 
850     /*
851      * If there is no free descriptors on head/body ring or next free
852      * descriptor is a head descriptor switch to body only ring
853      */
854     vmxnet3_read_next_rx_descr(s, RXQ_IDX, RX_BODY_ONLY_RING, d, didx);
855 
856     /* If no more free descriptors - return */
857     if (d->gen == vmxnet3_get_rx_ring_gen(s, RXQ_IDX, RX_BODY_ONLY_RING)) {
858         /* Only read after generation field verification */
859         smp_rmb();
860         /* Re-read to be sure we got the latest version */
861         vmxnet3_read_next_rx_descr(s, RXQ_IDX, RX_BODY_ONLY_RING, d, didx);
862         assert(d->btype == VMXNET3_RXD_BTYPE_BODY);
863         *ridx = RX_BODY_ONLY_RING;
864         vmxnet3_inc_rx_consumption_counter(s, RXQ_IDX, RX_BODY_ONLY_RING);
865         return true;
866     }
867 
868     return false;
869 }
870 
871 static inline bool
872 vmxnet3_get_next_rx_descr(VMXNET3State *s, bool is_head,
873                           struct Vmxnet3_RxDesc *descr_buf,
874                           uint32_t *descr_idx,
875                           uint32_t *ridx)
876 {
877     if (is_head || !s->rx_packets_compound) {
878         return vmxnet3_get_next_head_rx_descr(s, descr_buf, descr_idx, ridx);
879     } else {
880         return vmxnet3_get_next_body_rx_descr(s, descr_buf, descr_idx, ridx);
881     }
882 }
883 
884 static void vmxnet3_rx_update_descr(struct VmxnetRxPkt *pkt,
885     struct Vmxnet3_RxCompDesc *rxcd)
886 {
887     int csum_ok, is_gso;
888     bool isip4, isip6, istcp, isudp;
889     struct virtio_net_hdr *vhdr;
890     uint8_t offload_type;
891 
892     if (vmxnet_rx_pkt_is_vlan_stripped(pkt)) {
893         rxcd->ts = 1;
894         rxcd->tci = vmxnet_rx_pkt_get_vlan_tag(pkt);
895     }
896 
897     if (!vmxnet_rx_pkt_has_virt_hdr(pkt)) {
898         goto nocsum;
899     }
900 
901     vhdr = vmxnet_rx_pkt_get_vhdr(pkt);
902     /*
903      * Checksum is valid when lower level tell so or when lower level
904      * requires checksum offload telling that packet produced/bridged
905      * locally and did travel over network after last checksum calculation
906      * or production
907      */
908     csum_ok = VMXNET_FLAG_IS_SET(vhdr->flags, VIRTIO_NET_HDR_F_DATA_VALID) ||
909               VMXNET_FLAG_IS_SET(vhdr->flags, VIRTIO_NET_HDR_F_NEEDS_CSUM);
910 
911     offload_type = vhdr->gso_type & ~VIRTIO_NET_HDR_GSO_ECN;
912     is_gso = (offload_type != VIRTIO_NET_HDR_GSO_NONE) ? 1 : 0;
913 
914     if (!csum_ok && !is_gso) {
915         goto nocsum;
916     }
917 
918     vmxnet_rx_pkt_get_protocols(pkt, &isip4, &isip6, &isudp, &istcp);
919     if ((!istcp && !isudp) || (!isip4 && !isip6)) {
920         goto nocsum;
921     }
922 
923     rxcd->cnc = 0;
924     rxcd->v4 = isip4 ? 1 : 0;
925     rxcd->v6 = isip6 ? 1 : 0;
926     rxcd->tcp = istcp ? 1 : 0;
927     rxcd->udp = isudp ? 1 : 0;
928     rxcd->fcs = rxcd->tuc = rxcd->ipc = 1;
929     return;
930 
931 nocsum:
932     rxcd->cnc = 1;
933     return;
934 }
935 
936 static void
937 vmxnet3_physical_memory_writev(const struct iovec *iov,
938                                size_t start_iov_off,
939                                hwaddr target_addr,
940                                size_t bytes_to_copy)
941 {
942     size_t curr_off = 0;
943     size_t copied = 0;
944 
945     while (bytes_to_copy) {
946         if (start_iov_off < (curr_off + iov->iov_len)) {
947             size_t chunk_len =
948                 MIN((curr_off + iov->iov_len) - start_iov_off, bytes_to_copy);
949 
950             cpu_physical_memory_write(target_addr + copied,
951                                       iov->iov_base + start_iov_off - curr_off,
952                                       chunk_len);
953 
954             copied += chunk_len;
955             start_iov_off += chunk_len;
956             curr_off = start_iov_off;
957             bytes_to_copy -= chunk_len;
958         } else {
959             curr_off += iov->iov_len;
960         }
961         iov++;
962     }
963 }
964 
965 static bool
966 vmxnet3_indicate_packet(VMXNET3State *s)
967 {
968     struct Vmxnet3_RxDesc rxd;
969     bool is_head = true;
970     uint32_t rxd_idx;
971     uint32_t rx_ridx = 0;
972 
973     struct Vmxnet3_RxCompDesc rxcd;
974     uint32_t new_rxcd_gen = VMXNET3_INIT_GEN;
975     hwaddr new_rxcd_pa = 0;
976     hwaddr ready_rxcd_pa = 0;
977     struct iovec *data = vmxnet_rx_pkt_get_iovec(s->rx_pkt);
978     size_t bytes_copied = 0;
979     size_t bytes_left = vmxnet_rx_pkt_get_total_len(s->rx_pkt);
980     uint16_t num_frags = 0;
981     size_t chunk_size;
982 
983     vmxnet_rx_pkt_dump(s->rx_pkt);
984 
985     while (bytes_left > 0) {
986 
987         /* cannot add more frags to packet */
988         if (num_frags == s->max_rx_frags) {
989             break;
990         }
991 
992         new_rxcd_pa = vmxnet3_pop_rxc_descr(s, RXQ_IDX, &new_rxcd_gen);
993         if (!new_rxcd_pa) {
994             break;
995         }
996 
997         if (!vmxnet3_get_next_rx_descr(s, is_head, &rxd, &rxd_idx, &rx_ridx)) {
998             break;
999         }
1000 
1001         chunk_size = MIN(bytes_left, rxd.len);
1002         vmxnet3_physical_memory_writev(data, bytes_copied,
1003                                        le64_to_cpu(rxd.addr), chunk_size);
1004         bytes_copied += chunk_size;
1005         bytes_left -= chunk_size;
1006 
1007         vmxnet3_dump_rx_descr(&rxd);
1008 
1009         if (0 != ready_rxcd_pa) {
1010             cpu_physical_memory_write(ready_rxcd_pa, &rxcd, sizeof(rxcd));
1011         }
1012 
1013         memset(&rxcd, 0, sizeof(struct Vmxnet3_RxCompDesc));
1014         rxcd.rxdIdx = rxd_idx;
1015         rxcd.len = chunk_size;
1016         rxcd.sop = is_head;
1017         rxcd.gen = new_rxcd_gen;
1018         rxcd.rqID = RXQ_IDX + rx_ridx * s->rxq_num;
1019 
1020         if (0 == bytes_left) {
1021             vmxnet3_rx_update_descr(s->rx_pkt, &rxcd);
1022         }
1023 
1024         VMW_RIPRN("RX Completion descriptor: rxRing: %lu rxIdx %lu len %lu "
1025                   "sop %d csum_correct %lu",
1026                   (unsigned long) rx_ridx,
1027                   (unsigned long) rxcd.rxdIdx,
1028                   (unsigned long) rxcd.len,
1029                   (int) rxcd.sop,
1030                   (unsigned long) rxcd.tuc);
1031 
1032         is_head = false;
1033         ready_rxcd_pa = new_rxcd_pa;
1034         new_rxcd_pa = 0;
1035         num_frags++;
1036     }
1037 
1038     if (0 != ready_rxcd_pa) {
1039         rxcd.eop = 1;
1040         rxcd.err = (0 != bytes_left);
1041         cpu_physical_memory_write(ready_rxcd_pa, &rxcd, sizeof(rxcd));
1042 
1043         /* Flush RX descriptor changes */
1044         smp_wmb();
1045     }
1046 
1047     if (0 != new_rxcd_pa) {
1048         vmxnet3_revert_rxc_descr(s, RXQ_IDX);
1049     }
1050 
1051     vmxnet3_trigger_interrupt(s, s->rxq_descr[RXQ_IDX].intr_idx);
1052 
1053     if (bytes_left == 0) {
1054         vmxnet3_on_rx_done_update_stats(s, RXQ_IDX, VMXNET3_PKT_STATUS_OK);
1055         return true;
1056     } else if (num_frags == s->max_rx_frags) {
1057         vmxnet3_on_rx_done_update_stats(s, RXQ_IDX, VMXNET3_PKT_STATUS_ERROR);
1058         return false;
1059     } else {
1060         vmxnet3_on_rx_done_update_stats(s, RXQ_IDX,
1061                                         VMXNET3_PKT_STATUS_OUT_OF_BUF);
1062         return false;
1063     }
1064 }
1065 
1066 static void
1067 vmxnet3_io_bar0_write(void *opaque, hwaddr addr,
1068                       uint64_t val, unsigned size)
1069 {
1070     VMXNET3State *s = opaque;
1071 
1072     if (VMW_IS_MULTIREG_ADDR(addr, VMXNET3_REG_TXPROD,
1073                         VMXNET3_DEVICE_MAX_TX_QUEUES, VMXNET3_REG_ALIGN)) {
1074         int tx_queue_idx =
1075             VMW_MULTIREG_IDX_BY_ADDR(addr, VMXNET3_REG_TXPROD,
1076                                      VMXNET3_REG_ALIGN);
1077         assert(tx_queue_idx <= s->txq_num);
1078         vmxnet3_process_tx_queue(s, tx_queue_idx);
1079         return;
1080     }
1081 
1082     if (VMW_IS_MULTIREG_ADDR(addr, VMXNET3_REG_IMR,
1083                         VMXNET3_MAX_INTRS, VMXNET3_REG_ALIGN)) {
1084         int l = VMW_MULTIREG_IDX_BY_ADDR(addr, VMXNET3_REG_IMR,
1085                                          VMXNET3_REG_ALIGN);
1086 
1087         VMW_CBPRN("Interrupt mask for line %d written: 0x%" PRIx64, l, val);
1088 
1089         vmxnet3_on_interrupt_mask_changed(s, l, val);
1090         return;
1091     }
1092 
1093     if (VMW_IS_MULTIREG_ADDR(addr, VMXNET3_REG_RXPROD,
1094                         VMXNET3_DEVICE_MAX_RX_QUEUES, VMXNET3_REG_ALIGN) ||
1095        VMW_IS_MULTIREG_ADDR(addr, VMXNET3_REG_RXPROD2,
1096                         VMXNET3_DEVICE_MAX_RX_QUEUES, VMXNET3_REG_ALIGN)) {
1097         return;
1098     }
1099 
1100     VMW_WRPRN("BAR0 unknown write [%" PRIx64 "] = %" PRIx64 ", size %d",
1101               (uint64_t) addr, val, size);
1102 }
1103 
1104 static uint64_t
1105 vmxnet3_io_bar0_read(void *opaque, hwaddr addr, unsigned size)
1106 {
1107     if (VMW_IS_MULTIREG_ADDR(addr, VMXNET3_REG_IMR,
1108                         VMXNET3_MAX_INTRS, VMXNET3_REG_ALIGN)) {
1109         g_assert_not_reached();
1110     }
1111 
1112     VMW_CBPRN("BAR0 unknown read [%" PRIx64 "], size %d", addr, size);
1113     return 0;
1114 }
1115 
1116 static void vmxnet3_reset_interrupt_states(VMXNET3State *s)
1117 {
1118     int i;
1119     for (i = 0; i < ARRAY_SIZE(s->interrupt_states); i++) {
1120         s->interrupt_states[i].is_asserted = false;
1121         s->interrupt_states[i].is_pending = false;
1122         s->interrupt_states[i].is_masked = true;
1123     }
1124 }
1125 
1126 static void vmxnet3_reset_mac(VMXNET3State *s)
1127 {
1128     memcpy(&s->conf.macaddr.a, &s->perm_mac.a, sizeof(s->perm_mac.a));
1129     VMW_CFPRN("MAC address set to: " VMXNET_MF, VMXNET_MA(s->conf.macaddr.a));
1130 }
1131 
1132 static void vmxnet3_deactivate_device(VMXNET3State *s)
1133 {
1134     VMW_CBPRN("Deactivating vmxnet3...");
1135     s->device_active = false;
1136 }
1137 
1138 static void vmxnet3_reset(VMXNET3State *s)
1139 {
1140     VMW_CBPRN("Resetting vmxnet3...");
1141 
1142     vmxnet3_deactivate_device(s);
1143     vmxnet3_reset_interrupt_states(s);
1144     vmxnet_tx_pkt_reset(s->tx_pkt);
1145     s->drv_shmem = 0;
1146     s->tx_sop = true;
1147     s->skip_current_tx_pkt = false;
1148 }
1149 
1150 static void vmxnet3_update_rx_mode(VMXNET3State *s)
1151 {
1152     s->rx_mode = VMXNET3_READ_DRV_SHARED32(s->drv_shmem,
1153                                            devRead.rxFilterConf.rxMode);
1154     VMW_CFPRN("RX mode: 0x%08X", s->rx_mode);
1155 }
1156 
1157 static void vmxnet3_update_vlan_filters(VMXNET3State *s)
1158 {
1159     int i;
1160 
1161     /* Copy configuration from shared memory */
1162     VMXNET3_READ_DRV_SHARED(s->drv_shmem,
1163                             devRead.rxFilterConf.vfTable,
1164                             s->vlan_table,
1165                             sizeof(s->vlan_table));
1166 
1167     /* Invert byte order when needed */
1168     for (i = 0; i < ARRAY_SIZE(s->vlan_table); i++) {
1169         s->vlan_table[i] = le32_to_cpu(s->vlan_table[i]);
1170     }
1171 
1172     /* Dump configuration for debugging purposes */
1173     VMW_CFPRN("Configured VLANs:");
1174     for (i = 0; i < sizeof(s->vlan_table) * 8; i++) {
1175         if (VMXNET3_VFTABLE_ENTRY_IS_SET(s->vlan_table, i)) {
1176             VMW_CFPRN("\tVLAN %d is present", i);
1177         }
1178     }
1179 }
1180 
1181 static void vmxnet3_update_mcast_filters(VMXNET3State *s)
1182 {
1183     uint16_t list_bytes =
1184         VMXNET3_READ_DRV_SHARED16(s->drv_shmem,
1185                                   devRead.rxFilterConf.mfTableLen);
1186 
1187     s->mcast_list_len = list_bytes / sizeof(s->mcast_list[0]);
1188 
1189     s->mcast_list = g_realloc(s->mcast_list, list_bytes);
1190     if (NULL == s->mcast_list) {
1191         if (0 == s->mcast_list_len) {
1192             VMW_CFPRN("Current multicast list is empty");
1193         } else {
1194             VMW_ERPRN("Failed to allocate multicast list of %d elements",
1195                       s->mcast_list_len);
1196         }
1197         s->mcast_list_len = 0;
1198     } else {
1199         int i;
1200         hwaddr mcast_list_pa =
1201             VMXNET3_READ_DRV_SHARED64(s->drv_shmem,
1202                                       devRead.rxFilterConf.mfTablePA);
1203 
1204         cpu_physical_memory_read(mcast_list_pa, s->mcast_list, list_bytes);
1205         VMW_CFPRN("Current multicast list len is %d:", s->mcast_list_len);
1206         for (i = 0; i < s->mcast_list_len; i++) {
1207             VMW_CFPRN("\t" VMXNET_MF, VMXNET_MA(s->mcast_list[i].a));
1208         }
1209     }
1210 }
1211 
1212 static void vmxnet3_setup_rx_filtering(VMXNET3State *s)
1213 {
1214     vmxnet3_update_rx_mode(s);
1215     vmxnet3_update_vlan_filters(s);
1216     vmxnet3_update_mcast_filters(s);
1217 }
1218 
1219 static uint32_t vmxnet3_get_interrupt_config(VMXNET3State *s)
1220 {
1221     uint32_t interrupt_mode = VMXNET3_IT_AUTO | (VMXNET3_IMM_AUTO << 2);
1222     VMW_CFPRN("Interrupt config is 0x%X", interrupt_mode);
1223     return interrupt_mode;
1224 }
1225 
1226 static void vmxnet3_fill_stats(VMXNET3State *s)
1227 {
1228     int i;
1229     for (i = 0; i < s->txq_num; i++) {
1230         cpu_physical_memory_write(s->txq_descr[i].tx_stats_pa,
1231                                   &s->txq_descr[i].txq_stats,
1232                                   sizeof(s->txq_descr[i].txq_stats));
1233     }
1234 
1235     for (i = 0; i < s->rxq_num; i++) {
1236         cpu_physical_memory_write(s->rxq_descr[i].rx_stats_pa,
1237                                   &s->rxq_descr[i].rxq_stats,
1238                                   sizeof(s->rxq_descr[i].rxq_stats));
1239     }
1240 }
1241 
1242 static void vmxnet3_adjust_by_guest_type(VMXNET3State *s)
1243 {
1244     struct Vmxnet3_GOSInfo gos;
1245 
1246     VMXNET3_READ_DRV_SHARED(s->drv_shmem, devRead.misc.driverInfo.gos,
1247                             &gos, sizeof(gos));
1248     s->rx_packets_compound =
1249         (gos.gosType == VMXNET3_GOS_TYPE_WIN) ? false : true;
1250 
1251     VMW_CFPRN("Guest type specifics: RXCOMPOUND: %d", s->rx_packets_compound);
1252 }
1253 
1254 static void
1255 vmxnet3_dump_conf_descr(const char *name,
1256                         struct Vmxnet3_VariableLenConfDesc *pm_descr)
1257 {
1258     VMW_CFPRN("%s descriptor dump: Version %u, Length %u",
1259               name, pm_descr->confVer, pm_descr->confLen);
1260 
1261 };
1262 
1263 static void vmxnet3_update_pm_state(VMXNET3State *s)
1264 {
1265     struct Vmxnet3_VariableLenConfDesc pm_descr;
1266 
1267     pm_descr.confLen =
1268         VMXNET3_READ_DRV_SHARED32(s->drv_shmem, devRead.pmConfDesc.confLen);
1269     pm_descr.confVer =
1270         VMXNET3_READ_DRV_SHARED32(s->drv_shmem, devRead.pmConfDesc.confVer);
1271     pm_descr.confPA =
1272         VMXNET3_READ_DRV_SHARED64(s->drv_shmem, devRead.pmConfDesc.confPA);
1273 
1274     vmxnet3_dump_conf_descr("PM State", &pm_descr);
1275 }
1276 
1277 static void vmxnet3_update_features(VMXNET3State *s)
1278 {
1279     uint32_t guest_features;
1280     int rxcso_supported;
1281 
1282     guest_features = VMXNET3_READ_DRV_SHARED32(s->drv_shmem,
1283                                                devRead.misc.uptFeatures);
1284 
1285     rxcso_supported = VMXNET_FLAG_IS_SET(guest_features, UPT1_F_RXCSUM);
1286     s->rx_vlan_stripping = VMXNET_FLAG_IS_SET(guest_features, UPT1_F_RXVLAN);
1287     s->lro_supported = VMXNET_FLAG_IS_SET(guest_features, UPT1_F_LRO);
1288 
1289     VMW_CFPRN("Features configuration: LRO: %d, RXCSUM: %d, VLANSTRIP: %d",
1290               s->lro_supported, rxcso_supported,
1291               s->rx_vlan_stripping);
1292     if (s->peer_has_vhdr) {
1293         tap_set_offload(qemu_get_queue(s->nic)->peer,
1294                         rxcso_supported,
1295                         s->lro_supported,
1296                         s->lro_supported,
1297                         0,
1298                         0);
1299     }
1300 }
1301 
1302 static bool vmxnet3_verify_intx(VMXNET3State *s, int intx)
1303 {
1304     return s->msix_used || s->msi_used || (intx ==
1305            (pci_get_byte(s->parent_obj.config + PCI_INTERRUPT_PIN) - 1));
1306 }
1307 
1308 static void vmxnet3_activate_device(VMXNET3State *s)
1309 {
1310     int i;
1311     static const uint32_t VMXNET3_DEF_TX_THRESHOLD = 1;
1312     hwaddr qdescr_table_pa;
1313     uint64_t pa;
1314     uint32_t size;
1315 
1316     /* Verify configuration consistency */
1317     if (!vmxnet3_verify_driver_magic(s->drv_shmem)) {
1318         VMW_ERPRN("Device configuration received from driver is invalid");
1319         return;
1320     }
1321 
1322     vmxnet3_adjust_by_guest_type(s);
1323     vmxnet3_update_features(s);
1324     vmxnet3_update_pm_state(s);
1325     vmxnet3_setup_rx_filtering(s);
1326     /* Cache fields from shared memory */
1327     s->mtu = VMXNET3_READ_DRV_SHARED32(s->drv_shmem, devRead.misc.mtu);
1328     VMW_CFPRN("MTU is %u", s->mtu);
1329 
1330     s->max_rx_frags =
1331         VMXNET3_READ_DRV_SHARED16(s->drv_shmem, devRead.misc.maxNumRxSG);
1332 
1333     if (s->max_rx_frags == 0) {
1334         s->max_rx_frags = 1;
1335     }
1336 
1337     VMW_CFPRN("Max RX fragments is %u", s->max_rx_frags);
1338 
1339     s->event_int_idx =
1340         VMXNET3_READ_DRV_SHARED8(s->drv_shmem, devRead.intrConf.eventIntrIdx);
1341     assert(vmxnet3_verify_intx(s, s->event_int_idx));
1342     VMW_CFPRN("Events interrupt line is %u", s->event_int_idx);
1343 
1344     s->auto_int_masking =
1345         VMXNET3_READ_DRV_SHARED8(s->drv_shmem, devRead.intrConf.autoMask);
1346     VMW_CFPRN("Automatic interrupt masking is %d", (int)s->auto_int_masking);
1347 
1348     s->txq_num =
1349         VMXNET3_READ_DRV_SHARED8(s->drv_shmem, devRead.misc.numTxQueues);
1350     s->rxq_num =
1351         VMXNET3_READ_DRV_SHARED8(s->drv_shmem, devRead.misc.numRxQueues);
1352 
1353     VMW_CFPRN("Number of TX/RX queues %u/%u", s->txq_num, s->rxq_num);
1354     assert(s->txq_num <= VMXNET3_DEVICE_MAX_TX_QUEUES);
1355 
1356     qdescr_table_pa =
1357         VMXNET3_READ_DRV_SHARED64(s->drv_shmem, devRead.misc.queueDescPA);
1358     VMW_CFPRN("TX queues descriptors table is at 0x%" PRIx64, qdescr_table_pa);
1359 
1360     /*
1361      * Worst-case scenario is a packet that holds all TX rings space so
1362      * we calculate total size of all TX rings for max TX fragments number
1363      */
1364     s->max_tx_frags = 0;
1365 
1366     /* TX queues */
1367     for (i = 0; i < s->txq_num; i++) {
1368         hwaddr qdescr_pa =
1369             qdescr_table_pa + i * sizeof(struct Vmxnet3_TxQueueDesc);
1370 
1371         /* Read interrupt number for this TX queue */
1372         s->txq_descr[i].intr_idx =
1373             VMXNET3_READ_TX_QUEUE_DESCR8(qdescr_pa, conf.intrIdx);
1374         assert(vmxnet3_verify_intx(s, s->txq_descr[i].intr_idx));
1375 
1376         VMW_CFPRN("TX Queue %d interrupt: %d", i, s->txq_descr[i].intr_idx);
1377 
1378         /* Read rings memory locations for TX queues */
1379         pa = VMXNET3_READ_TX_QUEUE_DESCR64(qdescr_pa, conf.txRingBasePA);
1380         size = VMXNET3_READ_TX_QUEUE_DESCR32(qdescr_pa, conf.txRingSize);
1381 
1382         vmxnet3_ring_init(&s->txq_descr[i].tx_ring, pa, size,
1383                           sizeof(struct Vmxnet3_TxDesc), false);
1384         VMXNET3_RING_DUMP(VMW_CFPRN, "TX", i, &s->txq_descr[i].tx_ring);
1385 
1386         s->max_tx_frags += size;
1387 
1388         /* TXC ring */
1389         pa = VMXNET3_READ_TX_QUEUE_DESCR64(qdescr_pa, conf.compRingBasePA);
1390         size = VMXNET3_READ_TX_QUEUE_DESCR32(qdescr_pa, conf.compRingSize);
1391         vmxnet3_ring_init(&s->txq_descr[i].comp_ring, pa, size,
1392                           sizeof(struct Vmxnet3_TxCompDesc), true);
1393         VMXNET3_RING_DUMP(VMW_CFPRN, "TXC", i, &s->txq_descr[i].comp_ring);
1394 
1395         s->txq_descr[i].tx_stats_pa =
1396             qdescr_pa + offsetof(struct Vmxnet3_TxQueueDesc, stats);
1397 
1398         memset(&s->txq_descr[i].txq_stats, 0,
1399                sizeof(s->txq_descr[i].txq_stats));
1400 
1401         /* Fill device-managed parameters for queues */
1402         VMXNET3_WRITE_TX_QUEUE_DESCR32(qdescr_pa,
1403                                        ctrl.txThreshold,
1404                                        VMXNET3_DEF_TX_THRESHOLD);
1405     }
1406 
1407     /* Preallocate TX packet wrapper */
1408     VMW_CFPRN("Max TX fragments is %u", s->max_tx_frags);
1409     vmxnet_tx_pkt_init(&s->tx_pkt, s->max_tx_frags, s->peer_has_vhdr);
1410     vmxnet_rx_pkt_init(&s->rx_pkt, s->peer_has_vhdr);
1411 
1412     /* Read rings memory locations for RX queues */
1413     for (i = 0; i < s->rxq_num; i++) {
1414         int j;
1415         hwaddr qd_pa =
1416             qdescr_table_pa + s->txq_num * sizeof(struct Vmxnet3_TxQueueDesc) +
1417             i * sizeof(struct Vmxnet3_RxQueueDesc);
1418 
1419         /* Read interrupt number for this RX queue */
1420         s->rxq_descr[i].intr_idx =
1421             VMXNET3_READ_TX_QUEUE_DESCR8(qd_pa, conf.intrIdx);
1422         assert(vmxnet3_verify_intx(s, s->rxq_descr[i].intr_idx));
1423 
1424         VMW_CFPRN("RX Queue %d interrupt: %d", i, s->rxq_descr[i].intr_idx);
1425 
1426         /* Read rings memory locations */
1427         for (j = 0; j < VMXNET3_RX_RINGS_PER_QUEUE; j++) {
1428             /* RX rings */
1429             pa = VMXNET3_READ_RX_QUEUE_DESCR64(qd_pa, conf.rxRingBasePA[j]);
1430             size = VMXNET3_READ_RX_QUEUE_DESCR32(qd_pa, conf.rxRingSize[j]);
1431             vmxnet3_ring_init(&s->rxq_descr[i].rx_ring[j], pa, size,
1432                               sizeof(struct Vmxnet3_RxDesc), false);
1433             VMW_CFPRN("RX queue %d:%d: Base: %" PRIx64 ", Size: %d",
1434                       i, j, pa, size);
1435         }
1436 
1437         /* RXC ring */
1438         pa = VMXNET3_READ_RX_QUEUE_DESCR64(qd_pa, conf.compRingBasePA);
1439         size = VMXNET3_READ_RX_QUEUE_DESCR32(qd_pa, conf.compRingSize);
1440         vmxnet3_ring_init(&s->rxq_descr[i].comp_ring, pa, size,
1441                           sizeof(struct Vmxnet3_RxCompDesc), true);
1442         VMW_CFPRN("RXC queue %d: Base: %" PRIx64 ", Size: %d", i, pa, size);
1443 
1444         s->rxq_descr[i].rx_stats_pa =
1445             qd_pa + offsetof(struct Vmxnet3_RxQueueDesc, stats);
1446         memset(&s->rxq_descr[i].rxq_stats, 0,
1447                sizeof(s->rxq_descr[i].rxq_stats));
1448     }
1449 
1450     /* Make sure everything is in place before device activation */
1451     smp_wmb();
1452 
1453     vmxnet3_reset_mac(s);
1454 
1455     s->device_active = true;
1456 }
1457 
1458 static void vmxnet3_handle_command(VMXNET3State *s, uint64_t cmd)
1459 {
1460     s->last_command = cmd;
1461 
1462     switch (cmd) {
1463     case VMXNET3_CMD_GET_PERM_MAC_HI:
1464         VMW_CBPRN("Set: Get upper part of permanent MAC");
1465         break;
1466 
1467     case VMXNET3_CMD_GET_PERM_MAC_LO:
1468         VMW_CBPRN("Set: Get lower part of permanent MAC");
1469         break;
1470 
1471     case VMXNET3_CMD_GET_STATS:
1472         VMW_CBPRN("Set: Get device statistics");
1473         vmxnet3_fill_stats(s);
1474         break;
1475 
1476     case VMXNET3_CMD_ACTIVATE_DEV:
1477         VMW_CBPRN("Set: Activating vmxnet3 device");
1478         vmxnet3_activate_device(s);
1479         break;
1480 
1481     case VMXNET3_CMD_UPDATE_RX_MODE:
1482         VMW_CBPRN("Set: Update rx mode");
1483         vmxnet3_update_rx_mode(s);
1484         break;
1485 
1486     case VMXNET3_CMD_UPDATE_VLAN_FILTERS:
1487         VMW_CBPRN("Set: Update VLAN filters");
1488         vmxnet3_update_vlan_filters(s);
1489         break;
1490 
1491     case VMXNET3_CMD_UPDATE_MAC_FILTERS:
1492         VMW_CBPRN("Set: Update MAC filters");
1493         vmxnet3_update_mcast_filters(s);
1494         break;
1495 
1496     case VMXNET3_CMD_UPDATE_FEATURE:
1497         VMW_CBPRN("Set: Update features");
1498         vmxnet3_update_features(s);
1499         break;
1500 
1501     case VMXNET3_CMD_UPDATE_PMCFG:
1502         VMW_CBPRN("Set: Update power management config");
1503         vmxnet3_update_pm_state(s);
1504         break;
1505 
1506     case VMXNET3_CMD_GET_LINK:
1507         VMW_CBPRN("Set: Get link");
1508         break;
1509 
1510     case VMXNET3_CMD_RESET_DEV:
1511         VMW_CBPRN("Set: Reset device");
1512         vmxnet3_reset(s);
1513         break;
1514 
1515     case VMXNET3_CMD_QUIESCE_DEV:
1516         VMW_CBPRN("Set: VMXNET3_CMD_QUIESCE_DEV - pause the device");
1517         vmxnet3_deactivate_device(s);
1518         break;
1519 
1520     case VMXNET3_CMD_GET_CONF_INTR:
1521         VMW_CBPRN("Set: VMXNET3_CMD_GET_CONF_INTR - interrupt configuration");
1522         break;
1523 
1524     default:
1525         VMW_CBPRN("Received unknown command: %" PRIx64, cmd);
1526         break;
1527     }
1528 }
1529 
1530 static uint64_t vmxnet3_get_command_status(VMXNET3State *s)
1531 {
1532     uint64_t ret;
1533 
1534     switch (s->last_command) {
1535     case VMXNET3_CMD_ACTIVATE_DEV:
1536         ret = (s->device_active) ? 0 : -1;
1537         VMW_CFPRN("Device active: %" PRIx64, ret);
1538         break;
1539 
1540     case VMXNET3_CMD_RESET_DEV:
1541     case VMXNET3_CMD_QUIESCE_DEV:
1542     case VMXNET3_CMD_GET_QUEUE_STATUS:
1543         ret = 0;
1544         break;
1545 
1546     case VMXNET3_CMD_GET_LINK:
1547         ret = s->link_status_and_speed;
1548         VMW_CFPRN("Link and speed: %" PRIx64, ret);
1549         break;
1550 
1551     case VMXNET3_CMD_GET_PERM_MAC_LO:
1552         ret = vmxnet3_get_mac_low(&s->perm_mac);
1553         break;
1554 
1555     case VMXNET3_CMD_GET_PERM_MAC_HI:
1556         ret = vmxnet3_get_mac_high(&s->perm_mac);
1557         break;
1558 
1559     case VMXNET3_CMD_GET_CONF_INTR:
1560         ret = vmxnet3_get_interrupt_config(s);
1561         break;
1562 
1563     default:
1564         VMW_WRPRN("Received request for unknown command: %x", s->last_command);
1565         ret = -1;
1566         break;
1567     }
1568 
1569     return ret;
1570 }
1571 
1572 static void vmxnet3_set_events(VMXNET3State *s, uint32_t val)
1573 {
1574     uint32_t events;
1575 
1576     VMW_CBPRN("Setting events: 0x%x", val);
1577     events = VMXNET3_READ_DRV_SHARED32(s->drv_shmem, ecr) | val;
1578     VMXNET3_WRITE_DRV_SHARED32(s->drv_shmem, ecr, events);
1579 }
1580 
1581 static void vmxnet3_ack_events(VMXNET3State *s, uint32_t val)
1582 {
1583     uint32_t events;
1584 
1585     VMW_CBPRN("Clearing events: 0x%x", val);
1586     events = VMXNET3_READ_DRV_SHARED32(s->drv_shmem, ecr) & ~val;
1587     VMXNET3_WRITE_DRV_SHARED32(s->drv_shmem, ecr, events);
1588 }
1589 
1590 static void
1591 vmxnet3_io_bar1_write(void *opaque,
1592                       hwaddr addr,
1593                       uint64_t val,
1594                       unsigned size)
1595 {
1596     VMXNET3State *s = opaque;
1597 
1598     switch (addr) {
1599     /* Vmxnet3 Revision Report Selection */
1600     case VMXNET3_REG_VRRS:
1601         VMW_CBPRN("Write BAR1 [VMXNET3_REG_VRRS] = %" PRIx64 ", size %d",
1602                   val, size);
1603         break;
1604 
1605     /* UPT Version Report Selection */
1606     case VMXNET3_REG_UVRS:
1607         VMW_CBPRN("Write BAR1 [VMXNET3_REG_UVRS] = %" PRIx64 ", size %d",
1608                   val, size);
1609         break;
1610 
1611     /* Driver Shared Address Low */
1612     case VMXNET3_REG_DSAL:
1613         VMW_CBPRN("Write BAR1 [VMXNET3_REG_DSAL] = %" PRIx64 ", size %d",
1614                   val, size);
1615         /*
1616          * Guest driver will first write the low part of the shared
1617          * memory address. We save it to temp variable and set the
1618          * shared address only after we get the high part
1619          */
1620         if (0 == val) {
1621             s->device_active = false;
1622         }
1623         s->temp_shared_guest_driver_memory = val;
1624         s->drv_shmem = 0;
1625         break;
1626 
1627     /* Driver Shared Address High */
1628     case VMXNET3_REG_DSAH:
1629         VMW_CBPRN("Write BAR1 [VMXNET3_REG_DSAH] = %" PRIx64 ", size %d",
1630                   val, size);
1631         /*
1632          * Set the shared memory between guest driver and device.
1633          * We already should have low address part.
1634          */
1635         s->drv_shmem = s->temp_shared_guest_driver_memory | (val << 32);
1636         break;
1637 
1638     /* Command */
1639     case VMXNET3_REG_CMD:
1640         VMW_CBPRN("Write BAR1 [VMXNET3_REG_CMD] = %" PRIx64 ", size %d",
1641                   val, size);
1642         vmxnet3_handle_command(s, val);
1643         break;
1644 
1645     /* MAC Address Low */
1646     case VMXNET3_REG_MACL:
1647         VMW_CBPRN("Write BAR1 [VMXNET3_REG_MACL] = %" PRIx64 ", size %d",
1648                   val, size);
1649         s->temp_mac = val;
1650         break;
1651 
1652     /* MAC Address High */
1653     case VMXNET3_REG_MACH:
1654         VMW_CBPRN("Write BAR1 [VMXNET3_REG_MACH] = %" PRIx64 ", size %d",
1655                   val, size);
1656         vmxnet3_set_variable_mac(s, val, s->temp_mac);
1657         break;
1658 
1659     /* Interrupt Cause Register */
1660     case VMXNET3_REG_ICR:
1661         VMW_CBPRN("Write BAR1 [VMXNET3_REG_ICR] = %" PRIx64 ", size %d",
1662                   val, size);
1663         g_assert_not_reached();
1664         break;
1665 
1666     /* Event Cause Register */
1667     case VMXNET3_REG_ECR:
1668         VMW_CBPRN("Write BAR1 [VMXNET3_REG_ECR] = %" PRIx64 ", size %d",
1669                   val, size);
1670         vmxnet3_ack_events(s, val);
1671         break;
1672 
1673     default:
1674         VMW_CBPRN("Unknown Write to BAR1 [%" PRIx64 "] = %" PRIx64 ", size %d",
1675                   addr, val, size);
1676         break;
1677     }
1678 }
1679 
1680 static uint64_t
1681 vmxnet3_io_bar1_read(void *opaque, hwaddr addr, unsigned size)
1682 {
1683         VMXNET3State *s = opaque;
1684         uint64_t ret = 0;
1685 
1686         switch (addr) {
1687         /* Vmxnet3 Revision Report Selection */
1688         case VMXNET3_REG_VRRS:
1689             VMW_CBPRN("Read BAR1 [VMXNET3_REG_VRRS], size %d", size);
1690             ret = VMXNET3_DEVICE_REVISION;
1691             break;
1692 
1693         /* UPT Version Report Selection */
1694         case VMXNET3_REG_UVRS:
1695             VMW_CBPRN("Read BAR1 [VMXNET3_REG_UVRS], size %d", size);
1696             ret = VMXNET3_DEVICE_VERSION;
1697             break;
1698 
1699         /* Command */
1700         case VMXNET3_REG_CMD:
1701             VMW_CBPRN("Read BAR1 [VMXNET3_REG_CMD], size %d", size);
1702             ret = vmxnet3_get_command_status(s);
1703             break;
1704 
1705         /* MAC Address Low */
1706         case VMXNET3_REG_MACL:
1707             VMW_CBPRN("Read BAR1 [VMXNET3_REG_MACL], size %d", size);
1708             ret = vmxnet3_get_mac_low(&s->conf.macaddr);
1709             break;
1710 
1711         /* MAC Address High */
1712         case VMXNET3_REG_MACH:
1713             VMW_CBPRN("Read BAR1 [VMXNET3_REG_MACH], size %d", size);
1714             ret = vmxnet3_get_mac_high(&s->conf.macaddr);
1715             break;
1716 
1717         /*
1718          * Interrupt Cause Register
1719          * Used for legacy interrupts only so interrupt index always 0
1720          */
1721         case VMXNET3_REG_ICR:
1722             VMW_CBPRN("Read BAR1 [VMXNET3_REG_ICR], size %d", size);
1723             if (vmxnet3_interrupt_asserted(s, 0)) {
1724                 vmxnet3_clear_interrupt(s, 0);
1725                 ret = true;
1726             } else {
1727                 ret = false;
1728             }
1729             break;
1730 
1731         default:
1732             VMW_CBPRN("Unknow read BAR1[%" PRIx64 "], %d bytes", addr, size);
1733             break;
1734         }
1735 
1736         return ret;
1737 }
1738 
1739 static int
1740 vmxnet3_can_receive(NetClientState *nc)
1741 {
1742     VMXNET3State *s = qemu_get_nic_opaque(nc);
1743     return s->device_active &&
1744            VMXNET_FLAG_IS_SET(s->link_status_and_speed, VMXNET3_LINK_STATUS_UP);
1745 }
1746 
1747 static inline bool
1748 vmxnet3_is_registered_vlan(VMXNET3State *s, const void *data)
1749 {
1750     uint16_t vlan_tag = eth_get_pkt_tci(data) & VLAN_VID_MASK;
1751     if (IS_SPECIAL_VLAN_ID(vlan_tag)) {
1752         return true;
1753     }
1754 
1755     return VMXNET3_VFTABLE_ENTRY_IS_SET(s->vlan_table, vlan_tag);
1756 }
1757 
1758 static bool
1759 vmxnet3_is_allowed_mcast_group(VMXNET3State *s, const uint8_t *group_mac)
1760 {
1761     int i;
1762     for (i = 0; i < s->mcast_list_len; i++) {
1763         if (!memcmp(group_mac, s->mcast_list[i].a, sizeof(s->mcast_list[i]))) {
1764             return true;
1765         }
1766     }
1767     return false;
1768 }
1769 
1770 static bool
1771 vmxnet3_rx_filter_may_indicate(VMXNET3State *s, const void *data,
1772     size_t size)
1773 {
1774     struct eth_header *ehdr = PKT_GET_ETH_HDR(data);
1775 
1776     if (VMXNET_FLAG_IS_SET(s->rx_mode, VMXNET3_RXM_PROMISC)) {
1777         return true;
1778     }
1779 
1780     if (!vmxnet3_is_registered_vlan(s, data)) {
1781         return false;
1782     }
1783 
1784     switch (vmxnet_rx_pkt_get_packet_type(s->rx_pkt)) {
1785     case ETH_PKT_UCAST:
1786         if (!VMXNET_FLAG_IS_SET(s->rx_mode, VMXNET3_RXM_UCAST)) {
1787             return false;
1788         }
1789         if (memcmp(s->conf.macaddr.a, ehdr->h_dest, ETH_ALEN)) {
1790             return false;
1791         }
1792         break;
1793 
1794     case ETH_PKT_BCAST:
1795         if (!VMXNET_FLAG_IS_SET(s->rx_mode, VMXNET3_RXM_BCAST)) {
1796             return false;
1797         }
1798         break;
1799 
1800     case ETH_PKT_MCAST:
1801         if (VMXNET_FLAG_IS_SET(s->rx_mode, VMXNET3_RXM_ALL_MULTI)) {
1802             return true;
1803         }
1804         if (!VMXNET_FLAG_IS_SET(s->rx_mode, VMXNET3_RXM_MCAST)) {
1805             return false;
1806         }
1807         if (!vmxnet3_is_allowed_mcast_group(s, ehdr->h_dest)) {
1808             return false;
1809         }
1810         break;
1811 
1812     default:
1813         g_assert_not_reached();
1814     }
1815 
1816     return true;
1817 }
1818 
1819 static ssize_t
1820 vmxnet3_receive(NetClientState *nc, const uint8_t *buf, size_t size)
1821 {
1822     VMXNET3State *s = qemu_get_nic_opaque(nc);
1823     size_t bytes_indicated;
1824 
1825     if (!vmxnet3_can_receive(nc)) {
1826         VMW_PKPRN("Cannot receive now");
1827         return -1;
1828     }
1829 
1830     if (s->peer_has_vhdr) {
1831         vmxnet_rx_pkt_set_vhdr(s->rx_pkt, (struct virtio_net_hdr *)buf);
1832         buf += sizeof(struct virtio_net_hdr);
1833         size -= sizeof(struct virtio_net_hdr);
1834     }
1835 
1836     vmxnet_rx_pkt_set_packet_type(s->rx_pkt,
1837         get_eth_packet_type(PKT_GET_ETH_HDR(buf)));
1838 
1839     if (vmxnet3_rx_filter_may_indicate(s, buf, size)) {
1840         vmxnet_rx_pkt_attach_data(s->rx_pkt, buf, size, s->rx_vlan_stripping);
1841         bytes_indicated = vmxnet3_indicate_packet(s) ? size : -1;
1842         if (bytes_indicated < size) {
1843             VMW_PKPRN("RX: %lu of %lu bytes indicated", bytes_indicated, size);
1844         }
1845     } else {
1846         VMW_PKPRN("Packet dropped by RX filter");
1847         bytes_indicated = size;
1848     }
1849 
1850     assert(size > 0);
1851     assert(bytes_indicated != 0);
1852     return bytes_indicated;
1853 }
1854 
1855 static void vmxnet3_cleanup(NetClientState *nc)
1856 {
1857     VMXNET3State *s = qemu_get_nic_opaque(nc);
1858     s->nic = NULL;
1859 }
1860 
1861 static void vmxnet3_set_link_status(NetClientState *nc)
1862 {
1863     VMXNET3State *s = qemu_get_nic_opaque(nc);
1864 
1865     if (nc->link_down) {
1866         s->link_status_and_speed &= ~VMXNET3_LINK_STATUS_UP;
1867     } else {
1868         s->link_status_and_speed |= VMXNET3_LINK_STATUS_UP;
1869     }
1870 
1871     vmxnet3_set_events(s, VMXNET3_ECR_LINK);
1872     vmxnet3_trigger_interrupt(s, s->event_int_idx);
1873 }
1874 
1875 static NetClientInfo net_vmxnet3_info = {
1876         .type = NET_CLIENT_OPTIONS_KIND_NIC,
1877         .size = sizeof(NICState),
1878         .can_receive = vmxnet3_can_receive,
1879         .receive = vmxnet3_receive,
1880         .cleanup = vmxnet3_cleanup,
1881         .link_status_changed = vmxnet3_set_link_status,
1882 };
1883 
1884 static bool vmxnet3_peer_has_vnet_hdr(VMXNET3State *s)
1885 {
1886     NetClientState *peer = qemu_get_queue(s->nic)->peer;
1887 
1888     if ((NULL != peer)                              &&
1889         (peer->info->type == NET_CLIENT_OPTIONS_KIND_TAP)   &&
1890         tap_has_vnet_hdr(peer)) {
1891         return true;
1892     }
1893 
1894     VMW_WRPRN("Peer has no virtio extension. Task offloads will be emulated.");
1895     return false;
1896 }
1897 
1898 static void vmxnet3_net_uninit(VMXNET3State *s)
1899 {
1900     g_free(s->mcast_list);
1901     vmxnet_tx_pkt_reset(s->tx_pkt);
1902     vmxnet_tx_pkt_uninit(s->tx_pkt);
1903     vmxnet_rx_pkt_uninit(s->rx_pkt);
1904     qemu_del_nic(s->nic);
1905 }
1906 
1907 static void vmxnet3_net_init(VMXNET3State *s)
1908 {
1909     DeviceState *d = DEVICE(s);
1910 
1911     VMW_CBPRN("vmxnet3_net_init called...");
1912 
1913     qemu_macaddr_default_if_unset(&s->conf.macaddr);
1914 
1915     /* Windows guest will query the address that was set on init */
1916     memcpy(&s->perm_mac.a, &s->conf.macaddr.a, sizeof(s->perm_mac.a));
1917 
1918     s->mcast_list = NULL;
1919     s->mcast_list_len = 0;
1920 
1921     s->link_status_and_speed = VMXNET3_LINK_SPEED | VMXNET3_LINK_STATUS_UP;
1922 
1923     VMW_CFPRN("Permanent MAC: " MAC_FMT, MAC_ARG(s->perm_mac.a));
1924 
1925     s->nic = qemu_new_nic(&net_vmxnet3_info, &s->conf,
1926                           object_get_typename(OBJECT(s)),
1927                           d->id, s);
1928 
1929     s->peer_has_vhdr = vmxnet3_peer_has_vnet_hdr(s);
1930     s->tx_sop = true;
1931     s->skip_current_tx_pkt = false;
1932     s->tx_pkt = NULL;
1933     s->rx_pkt = NULL;
1934     s->rx_vlan_stripping = false;
1935     s->lro_supported = false;
1936 
1937     if (s->peer_has_vhdr) {
1938         tap_set_vnet_hdr_len(qemu_get_queue(s->nic)->peer,
1939             sizeof(struct virtio_net_hdr));
1940 
1941         tap_using_vnet_hdr(qemu_get_queue(s->nic)->peer, 1);
1942     }
1943 
1944     qemu_format_nic_info_str(qemu_get_queue(s->nic), s->conf.macaddr.a);
1945 }
1946 
1947 static void
1948 vmxnet3_unuse_msix_vectors(VMXNET3State *s, int num_vectors)
1949 {
1950     PCIDevice *d = PCI_DEVICE(s);
1951     int i;
1952     for (i = 0; i < num_vectors; i++) {
1953         msix_vector_unuse(d, i);
1954     }
1955 }
1956 
1957 static bool
1958 vmxnet3_use_msix_vectors(VMXNET3State *s, int num_vectors)
1959 {
1960     PCIDevice *d = PCI_DEVICE(s);
1961     int i;
1962     for (i = 0; i < num_vectors; i++) {
1963         int res = msix_vector_use(d, i);
1964         if (0 > res) {
1965             VMW_WRPRN("Failed to use MSI-X vector %d, error %d", i, res);
1966             vmxnet3_unuse_msix_vectors(s, i);
1967             return false;
1968         }
1969     }
1970     return true;
1971 }
1972 
1973 static bool
1974 vmxnet3_init_msix(VMXNET3State *s)
1975 {
1976     PCIDevice *d = PCI_DEVICE(s);
1977     int res = msix_init(d, VMXNET3_MAX_INTRS,
1978                         &s->msix_bar,
1979                         VMXNET3_MSIX_BAR_IDX, VMXNET3_OFF_MSIX_TABLE,
1980                         &s->msix_bar,
1981                         VMXNET3_MSIX_BAR_IDX, VMXNET3_OFF_MSIX_PBA,
1982                         0);
1983 
1984     if (0 > res) {
1985         VMW_WRPRN("Failed to initialize MSI-X, error %d", res);
1986         s->msix_used = false;
1987     } else {
1988         if (!vmxnet3_use_msix_vectors(s, VMXNET3_MAX_INTRS)) {
1989             VMW_WRPRN("Failed to use MSI-X vectors, error %d", res);
1990             msix_uninit(d, &s->msix_bar, &s->msix_bar);
1991             s->msix_used = false;
1992         } else {
1993             s->msix_used = true;
1994         }
1995     }
1996     return s->msix_used;
1997 }
1998 
1999 static void
2000 vmxnet3_cleanup_msix(VMXNET3State *s)
2001 {
2002     PCIDevice *d = PCI_DEVICE(s);
2003 
2004     if (s->msix_used) {
2005         msix_vector_unuse(d, VMXNET3_MAX_INTRS);
2006         msix_uninit(d, &s->msix_bar, &s->msix_bar);
2007     }
2008 }
2009 
2010 #define VMXNET3_MSI_NUM_VECTORS   (1)
2011 #define VMXNET3_MSI_OFFSET        (0x50)
2012 #define VMXNET3_USE_64BIT         (true)
2013 #define VMXNET3_PER_VECTOR_MASK   (false)
2014 
2015 static bool
2016 vmxnet3_init_msi(VMXNET3State *s)
2017 {
2018     PCIDevice *d = PCI_DEVICE(s);
2019     int res;
2020 
2021     res = msi_init(d, VMXNET3_MSI_OFFSET, VMXNET3_MSI_NUM_VECTORS,
2022                    VMXNET3_USE_64BIT, VMXNET3_PER_VECTOR_MASK);
2023     if (0 > res) {
2024         VMW_WRPRN("Failed to initialize MSI, error %d", res);
2025         s->msi_used = false;
2026     } else {
2027         s->msi_used = true;
2028     }
2029 
2030     return s->msi_used;
2031 }
2032 
2033 static void
2034 vmxnet3_cleanup_msi(VMXNET3State *s)
2035 {
2036     PCIDevice *d = PCI_DEVICE(s);
2037 
2038     if (s->msi_used) {
2039         msi_uninit(d);
2040     }
2041 }
2042 
2043 static void
2044 vmxnet3_msix_save(QEMUFile *f, void *opaque)
2045 {
2046     PCIDevice *d = PCI_DEVICE(opaque);
2047     msix_save(d, f);
2048 }
2049 
2050 static int
2051 vmxnet3_msix_load(QEMUFile *f, void *opaque, int version_id)
2052 {
2053     PCIDevice *d = PCI_DEVICE(opaque);
2054     msix_load(d, f);
2055     return 0;
2056 }
2057 
2058 static const MemoryRegionOps b0_ops = {
2059     .read = vmxnet3_io_bar0_read,
2060     .write = vmxnet3_io_bar0_write,
2061     .endianness = DEVICE_LITTLE_ENDIAN,
2062     .impl = {
2063             .min_access_size = 4,
2064             .max_access_size = 4,
2065     },
2066 };
2067 
2068 static const MemoryRegionOps b1_ops = {
2069     .read = vmxnet3_io_bar1_read,
2070     .write = vmxnet3_io_bar1_write,
2071     .endianness = DEVICE_LITTLE_ENDIAN,
2072     .impl = {
2073             .min_access_size = 4,
2074             .max_access_size = 4,
2075     },
2076 };
2077 
2078 static int vmxnet3_pci_init(PCIDevice *pci_dev)
2079 {
2080     DeviceState *dev = DEVICE(pci_dev);
2081     VMXNET3State *s = VMXNET3(pci_dev);
2082 
2083     VMW_CBPRN("Starting init...");
2084 
2085     memory_region_init_io(&s->bar0, OBJECT(s), &b0_ops, s,
2086                           "vmxnet3-b0", VMXNET3_PT_REG_SIZE);
2087     pci_register_bar(pci_dev, VMXNET3_BAR0_IDX,
2088                      PCI_BASE_ADDRESS_SPACE_MEMORY, &s->bar0);
2089 
2090     memory_region_init_io(&s->bar1, OBJECT(s), &b1_ops, s,
2091                           "vmxnet3-b1", VMXNET3_VD_REG_SIZE);
2092     pci_register_bar(pci_dev, VMXNET3_BAR1_IDX,
2093                      PCI_BASE_ADDRESS_SPACE_MEMORY, &s->bar1);
2094 
2095     memory_region_init(&s->msix_bar, OBJECT(s), "vmxnet3-msix-bar",
2096                        VMXNET3_MSIX_BAR_SIZE);
2097     pci_register_bar(pci_dev, VMXNET3_MSIX_BAR_IDX,
2098                      PCI_BASE_ADDRESS_SPACE_MEMORY, &s->msix_bar);
2099 
2100     vmxnet3_reset_interrupt_states(s);
2101 
2102     /* Interrupt pin A */
2103     pci_dev->config[PCI_INTERRUPT_PIN] = 0x01;
2104 
2105     if (!vmxnet3_init_msix(s)) {
2106         VMW_WRPRN("Failed to initialize MSI-X, configuration is inconsistent.");
2107     }
2108 
2109     if (!vmxnet3_init_msi(s)) {
2110         VMW_WRPRN("Failed to initialize MSI, configuration is inconsistent.");
2111     }
2112 
2113     vmxnet3_net_init(s);
2114 
2115     register_savevm(dev, "vmxnet3-msix", -1, 1,
2116                     vmxnet3_msix_save, vmxnet3_msix_load, s);
2117 
2118     add_boot_device_path(s->conf.bootindex, dev, "/ethernet-phy@0");
2119 
2120     return 0;
2121 }
2122 
2123 
2124 static void vmxnet3_pci_uninit(PCIDevice *pci_dev)
2125 {
2126     DeviceState *dev = DEVICE(pci_dev);
2127     VMXNET3State *s = VMXNET3(pci_dev);
2128 
2129     VMW_CBPRN("Starting uninit...");
2130 
2131     unregister_savevm(dev, "vmxnet3-msix", s);
2132 
2133     vmxnet3_net_uninit(s);
2134 
2135     vmxnet3_cleanup_msix(s);
2136 
2137     vmxnet3_cleanup_msi(s);
2138 
2139     memory_region_destroy(&s->bar0);
2140     memory_region_destroy(&s->bar1);
2141     memory_region_destroy(&s->msix_bar);
2142 }
2143 
2144 static void vmxnet3_qdev_reset(DeviceState *dev)
2145 {
2146     PCIDevice *d = PCI_DEVICE(dev);
2147     VMXNET3State *s = VMXNET3(d);
2148 
2149     VMW_CBPRN("Starting QDEV reset...");
2150     vmxnet3_reset(s);
2151 }
2152 
2153 static bool vmxnet3_mc_list_needed(void *opaque)
2154 {
2155     return true;
2156 }
2157 
2158 static int vmxnet3_mcast_list_pre_load(void *opaque)
2159 {
2160     VMXNET3State *s = opaque;
2161 
2162     s->mcast_list = g_malloc(s->mcast_list_buff_size);
2163 
2164     return 0;
2165 }
2166 
2167 
2168 static void vmxnet3_pre_save(void *opaque)
2169 {
2170     VMXNET3State *s = opaque;
2171 
2172     s->mcast_list_buff_size = s->mcast_list_len * sizeof(MACAddr);
2173 }
2174 
2175 static const VMStateDescription vmxstate_vmxnet3_mcast_list = {
2176     .name = "vmxnet3/mcast_list",
2177     .version_id = 1,
2178     .minimum_version_id = 1,
2179     .minimum_version_id_old = 1,
2180     .pre_load = vmxnet3_mcast_list_pre_load,
2181     .fields = (VMStateField[]) {
2182         VMSTATE_VBUFFER_UINT32(mcast_list, VMXNET3State, 0, NULL, 0,
2183             mcast_list_buff_size),
2184         VMSTATE_END_OF_LIST()
2185     }
2186 };
2187 
2188 static void vmxnet3_get_ring_from_file(QEMUFile *f, Vmxnet3Ring *r)
2189 {
2190     r->pa = qemu_get_be64(f);
2191     r->size = qemu_get_be32(f);
2192     r->cell_size = qemu_get_be32(f);
2193     r->next = qemu_get_be32(f);
2194     r->gen = qemu_get_byte(f);
2195 }
2196 
2197 static void vmxnet3_put_ring_to_file(QEMUFile *f, Vmxnet3Ring *r)
2198 {
2199     qemu_put_be64(f, r->pa);
2200     qemu_put_be32(f, r->size);
2201     qemu_put_be32(f, r->cell_size);
2202     qemu_put_be32(f, r->next);
2203     qemu_put_byte(f, r->gen);
2204 }
2205 
2206 static void vmxnet3_get_tx_stats_from_file(QEMUFile *f,
2207     struct UPT1_TxStats *tx_stat)
2208 {
2209     tx_stat->TSOPktsTxOK = qemu_get_be64(f);
2210     tx_stat->TSOBytesTxOK = qemu_get_be64(f);
2211     tx_stat->ucastPktsTxOK = qemu_get_be64(f);
2212     tx_stat->ucastBytesTxOK = qemu_get_be64(f);
2213     tx_stat->mcastPktsTxOK = qemu_get_be64(f);
2214     tx_stat->mcastBytesTxOK = qemu_get_be64(f);
2215     tx_stat->bcastPktsTxOK = qemu_get_be64(f);
2216     tx_stat->bcastBytesTxOK = qemu_get_be64(f);
2217     tx_stat->pktsTxError = qemu_get_be64(f);
2218     tx_stat->pktsTxDiscard = qemu_get_be64(f);
2219 }
2220 
2221 static void vmxnet3_put_tx_stats_to_file(QEMUFile *f,
2222     struct UPT1_TxStats *tx_stat)
2223 {
2224     qemu_put_be64(f, tx_stat->TSOPktsTxOK);
2225     qemu_put_be64(f, tx_stat->TSOBytesTxOK);
2226     qemu_put_be64(f, tx_stat->ucastPktsTxOK);
2227     qemu_put_be64(f, tx_stat->ucastBytesTxOK);
2228     qemu_put_be64(f, tx_stat->mcastPktsTxOK);
2229     qemu_put_be64(f, tx_stat->mcastBytesTxOK);
2230     qemu_put_be64(f, tx_stat->bcastPktsTxOK);
2231     qemu_put_be64(f, tx_stat->bcastBytesTxOK);
2232     qemu_put_be64(f, tx_stat->pktsTxError);
2233     qemu_put_be64(f, tx_stat->pktsTxDiscard);
2234 }
2235 
2236 static int vmxnet3_get_txq_descr(QEMUFile *f, void *pv, size_t size)
2237 {
2238     Vmxnet3TxqDescr *r = pv;
2239 
2240     vmxnet3_get_ring_from_file(f, &r->tx_ring);
2241     vmxnet3_get_ring_from_file(f, &r->comp_ring);
2242     r->intr_idx = qemu_get_byte(f);
2243     r->tx_stats_pa = qemu_get_be64(f);
2244 
2245     vmxnet3_get_tx_stats_from_file(f, &r->txq_stats);
2246 
2247     return 0;
2248 }
2249 
2250 static void vmxnet3_put_txq_descr(QEMUFile *f, void *pv, size_t size)
2251 {
2252     Vmxnet3TxqDescr *r = pv;
2253 
2254     vmxnet3_put_ring_to_file(f, &r->tx_ring);
2255     vmxnet3_put_ring_to_file(f, &r->comp_ring);
2256     qemu_put_byte(f, r->intr_idx);
2257     qemu_put_be64(f, r->tx_stats_pa);
2258     vmxnet3_put_tx_stats_to_file(f, &r->txq_stats);
2259 }
2260 
2261 const VMStateInfo txq_descr_info = {
2262     .name = "txq_descr",
2263     .get = vmxnet3_get_txq_descr,
2264     .put = vmxnet3_put_txq_descr
2265 };
2266 
2267 static void vmxnet3_get_rx_stats_from_file(QEMUFile *f,
2268     struct UPT1_RxStats *rx_stat)
2269 {
2270     rx_stat->LROPktsRxOK = qemu_get_be64(f);
2271     rx_stat->LROBytesRxOK = qemu_get_be64(f);
2272     rx_stat->ucastPktsRxOK = qemu_get_be64(f);
2273     rx_stat->ucastBytesRxOK = qemu_get_be64(f);
2274     rx_stat->mcastPktsRxOK = qemu_get_be64(f);
2275     rx_stat->mcastBytesRxOK = qemu_get_be64(f);
2276     rx_stat->bcastPktsRxOK = qemu_get_be64(f);
2277     rx_stat->bcastBytesRxOK = qemu_get_be64(f);
2278     rx_stat->pktsRxOutOfBuf = qemu_get_be64(f);
2279     rx_stat->pktsRxError = qemu_get_be64(f);
2280 }
2281 
2282 static void vmxnet3_put_rx_stats_to_file(QEMUFile *f,
2283     struct UPT1_RxStats *rx_stat)
2284 {
2285     qemu_put_be64(f, rx_stat->LROPktsRxOK);
2286     qemu_put_be64(f, rx_stat->LROBytesRxOK);
2287     qemu_put_be64(f, rx_stat->ucastPktsRxOK);
2288     qemu_put_be64(f, rx_stat->ucastBytesRxOK);
2289     qemu_put_be64(f, rx_stat->mcastPktsRxOK);
2290     qemu_put_be64(f, rx_stat->mcastBytesRxOK);
2291     qemu_put_be64(f, rx_stat->bcastPktsRxOK);
2292     qemu_put_be64(f, rx_stat->bcastBytesRxOK);
2293     qemu_put_be64(f, rx_stat->pktsRxOutOfBuf);
2294     qemu_put_be64(f, rx_stat->pktsRxError);
2295 }
2296 
2297 static int vmxnet3_get_rxq_descr(QEMUFile *f, void *pv, size_t size)
2298 {
2299     Vmxnet3RxqDescr *r = pv;
2300     int i;
2301 
2302     for (i = 0; i < VMXNET3_RX_RINGS_PER_QUEUE; i++) {
2303         vmxnet3_get_ring_from_file(f, &r->rx_ring[i]);
2304     }
2305 
2306     vmxnet3_get_ring_from_file(f, &r->comp_ring);
2307     r->intr_idx = qemu_get_byte(f);
2308     r->rx_stats_pa = qemu_get_be64(f);
2309 
2310     vmxnet3_get_rx_stats_from_file(f, &r->rxq_stats);
2311 
2312     return 0;
2313 }
2314 
2315 static void vmxnet3_put_rxq_descr(QEMUFile *f, void *pv, size_t size)
2316 {
2317     Vmxnet3RxqDescr *r = pv;
2318     int i;
2319 
2320     for (i = 0; i < VMXNET3_RX_RINGS_PER_QUEUE; i++) {
2321         vmxnet3_put_ring_to_file(f, &r->rx_ring[i]);
2322     }
2323 
2324     vmxnet3_put_ring_to_file(f, &r->comp_ring);
2325     qemu_put_byte(f, r->intr_idx);
2326     qemu_put_be64(f, r->rx_stats_pa);
2327     vmxnet3_put_rx_stats_to_file(f, &r->rxq_stats);
2328 }
2329 
2330 static int vmxnet3_post_load(void *opaque, int version_id)
2331 {
2332     VMXNET3State *s = opaque;
2333     PCIDevice *d = PCI_DEVICE(s);
2334 
2335     vmxnet_tx_pkt_init(&s->tx_pkt, s->max_tx_frags, s->peer_has_vhdr);
2336     vmxnet_rx_pkt_init(&s->rx_pkt, s->peer_has_vhdr);
2337 
2338     if (s->msix_used) {
2339         if  (!vmxnet3_use_msix_vectors(s, VMXNET3_MAX_INTRS)) {
2340             VMW_WRPRN("Failed to re-use MSI-X vectors");
2341             msix_uninit(d, &s->msix_bar, &s->msix_bar);
2342             s->msix_used = false;
2343             return -1;
2344         }
2345     }
2346 
2347     return 0;
2348 }
2349 
2350 const VMStateInfo rxq_descr_info = {
2351     .name = "rxq_descr",
2352     .get = vmxnet3_get_rxq_descr,
2353     .put = vmxnet3_put_rxq_descr
2354 };
2355 
2356 static int vmxnet3_get_int_state(QEMUFile *f, void *pv, size_t size)
2357 {
2358     Vmxnet3IntState *r = pv;
2359 
2360     r->is_masked = qemu_get_byte(f);
2361     r->is_pending = qemu_get_byte(f);
2362     r->is_asserted = qemu_get_byte(f);
2363 
2364     return 0;
2365 }
2366 
2367 static void vmxnet3_put_int_state(QEMUFile *f, void *pv, size_t size)
2368 {
2369     Vmxnet3IntState *r = pv;
2370 
2371     qemu_put_byte(f, r->is_masked);
2372     qemu_put_byte(f, r->is_pending);
2373     qemu_put_byte(f, r->is_asserted);
2374 }
2375 
2376 const VMStateInfo int_state_info = {
2377     .name = "int_state",
2378     .get = vmxnet3_get_int_state,
2379     .put = vmxnet3_put_int_state
2380 };
2381 
2382 static const VMStateDescription vmstate_vmxnet3 = {
2383     .name = "vmxnet3",
2384     .version_id = 1,
2385     .minimum_version_id = 1,
2386     .minimum_version_id_old = 1,
2387     .pre_save = vmxnet3_pre_save,
2388     .post_load = vmxnet3_post_load,
2389     .fields      = (VMStateField[]) {
2390             VMSTATE_PCI_DEVICE(parent_obj, VMXNET3State),
2391             VMSTATE_BOOL(rx_packets_compound, VMXNET3State),
2392             VMSTATE_BOOL(rx_vlan_stripping, VMXNET3State),
2393             VMSTATE_BOOL(lro_supported, VMXNET3State),
2394             VMSTATE_UINT32(rx_mode, VMXNET3State),
2395             VMSTATE_UINT32(mcast_list_len, VMXNET3State),
2396             VMSTATE_UINT32(mcast_list_buff_size, VMXNET3State),
2397             VMSTATE_UINT32_ARRAY(vlan_table, VMXNET3State, VMXNET3_VFT_SIZE),
2398             VMSTATE_UINT32(mtu, VMXNET3State),
2399             VMSTATE_UINT16(max_rx_frags, VMXNET3State),
2400             VMSTATE_UINT32(max_tx_frags, VMXNET3State),
2401             VMSTATE_UINT8(event_int_idx, VMXNET3State),
2402             VMSTATE_BOOL(auto_int_masking, VMXNET3State),
2403             VMSTATE_UINT8(txq_num, VMXNET3State),
2404             VMSTATE_UINT8(rxq_num, VMXNET3State),
2405             VMSTATE_UINT32(device_active, VMXNET3State),
2406             VMSTATE_UINT32(last_command, VMXNET3State),
2407             VMSTATE_UINT32(link_status_and_speed, VMXNET3State),
2408             VMSTATE_UINT32(temp_mac, VMXNET3State),
2409             VMSTATE_UINT64(drv_shmem, VMXNET3State),
2410             VMSTATE_UINT64(temp_shared_guest_driver_memory, VMXNET3State),
2411 
2412             VMSTATE_ARRAY(txq_descr, VMXNET3State,
2413                 VMXNET3_DEVICE_MAX_TX_QUEUES, 0, txq_descr_info,
2414                 Vmxnet3TxqDescr),
2415             VMSTATE_ARRAY(rxq_descr, VMXNET3State,
2416                 VMXNET3_DEVICE_MAX_RX_QUEUES, 0, rxq_descr_info,
2417                 Vmxnet3RxqDescr),
2418             VMSTATE_ARRAY(interrupt_states, VMXNET3State, VMXNET3_MAX_INTRS,
2419                 0, int_state_info, Vmxnet3IntState),
2420 
2421             VMSTATE_END_OF_LIST()
2422     },
2423     .subsections = (VMStateSubsection[]) {
2424         {
2425             .vmsd = &vmxstate_vmxnet3_mcast_list,
2426             .needed = vmxnet3_mc_list_needed
2427         },
2428         {
2429             /* empty element. */
2430         }
2431     }
2432 };
2433 
2434 static void
2435 vmxnet3_write_config(PCIDevice *pci_dev, uint32_t addr, uint32_t val, int len)
2436 {
2437     pci_default_write_config(pci_dev, addr, val, len);
2438     msix_write_config(pci_dev, addr, val, len);
2439     msi_write_config(pci_dev, addr, val, len);
2440 }
2441 
2442 static Property vmxnet3_properties[] = {
2443     DEFINE_NIC_PROPERTIES(VMXNET3State, conf),
2444     DEFINE_PROP_END_OF_LIST(),
2445 };
2446 
2447 static void vmxnet3_class_init(ObjectClass *class, void *data)
2448 {
2449     DeviceClass *dc = DEVICE_CLASS(class);
2450     PCIDeviceClass *c = PCI_DEVICE_CLASS(class);
2451 
2452     c->init = vmxnet3_pci_init;
2453     c->exit = vmxnet3_pci_uninit;
2454     c->vendor_id = PCI_VENDOR_ID_VMWARE;
2455     c->device_id = PCI_DEVICE_ID_VMWARE_VMXNET3;
2456     c->revision = PCI_DEVICE_ID_VMWARE_VMXNET3_REVISION;
2457     c->class_id = PCI_CLASS_NETWORK_ETHERNET;
2458     c->subsystem_vendor_id = PCI_VENDOR_ID_VMWARE;
2459     c->subsystem_id = PCI_DEVICE_ID_VMWARE_VMXNET3;
2460     c->config_write = vmxnet3_write_config,
2461     dc->desc = "VMWare Paravirtualized Ethernet v3";
2462     dc->reset = vmxnet3_qdev_reset;
2463     dc->vmsd = &vmstate_vmxnet3;
2464     dc->props = vmxnet3_properties;
2465     set_bit(DEVICE_CATEGORY_NETWORK, dc->categories);
2466 }
2467 
2468 static const TypeInfo vmxnet3_info = {
2469     .name          = TYPE_VMXNET3,
2470     .parent        = TYPE_PCI_DEVICE,
2471     .instance_size = sizeof(VMXNET3State),
2472     .class_init    = vmxnet3_class_init,
2473 };
2474 
2475 static void vmxnet3_register_types(void)
2476 {
2477     VMW_CBPRN("vmxnet3_register_types called...");
2478     type_register_static(&vmxnet3_info);
2479 }
2480 
2481 type_init(vmxnet3_register_types)
2482