xref: /openbmc/qemu/hw/net/vmxnet3.c (revision f0513d2c)
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     qemu_set_irq(d->irq[int_idx], 1);
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     qemu_set_irq(d->irq[lidx], 0);
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         assert(false);
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             assert(false);
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         assert(false);
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             assert(false);
638         }
639 
640         if (tot_len > s->mtu) {
641             stats->LROPktsRxOK++;
642             stats->LROBytesRxOK += tot_len;
643         }
644         break;
645     default:
646         assert(false);
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         assert(false);
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 void vmxnet3_activate_device(VMXNET3State *s)
1303 {
1304     int i;
1305     static const uint32_t VMXNET3_DEF_TX_THRESHOLD = 1;
1306     hwaddr qdescr_table_pa;
1307     uint64_t pa;
1308     uint32_t size;
1309 
1310     /* Verify configuration consistency */
1311     if (!vmxnet3_verify_driver_magic(s->drv_shmem)) {
1312         VMW_ERPRN("Device configuration received from driver is invalid");
1313         return;
1314     }
1315 
1316     vmxnet3_adjust_by_guest_type(s);
1317     vmxnet3_update_features(s);
1318     vmxnet3_update_pm_state(s);
1319     vmxnet3_setup_rx_filtering(s);
1320     /* Cache fields from shared memory */
1321     s->mtu = VMXNET3_READ_DRV_SHARED32(s->drv_shmem, devRead.misc.mtu);
1322     VMW_CFPRN("MTU is %u", s->mtu);
1323 
1324     s->max_rx_frags =
1325         VMXNET3_READ_DRV_SHARED16(s->drv_shmem, devRead.misc.maxNumRxSG);
1326 
1327     if (s->max_rx_frags == 0) {
1328         s->max_rx_frags = 1;
1329     }
1330 
1331     VMW_CFPRN("Max RX fragments is %u", s->max_rx_frags);
1332 
1333     s->event_int_idx =
1334         VMXNET3_READ_DRV_SHARED8(s->drv_shmem, devRead.intrConf.eventIntrIdx);
1335     VMW_CFPRN("Events interrupt line is %u", s->event_int_idx);
1336 
1337     s->auto_int_masking =
1338         VMXNET3_READ_DRV_SHARED8(s->drv_shmem, devRead.intrConf.autoMask);
1339     VMW_CFPRN("Automatic interrupt masking is %d", (int)s->auto_int_masking);
1340 
1341     s->txq_num =
1342         VMXNET3_READ_DRV_SHARED8(s->drv_shmem, devRead.misc.numTxQueues);
1343     s->rxq_num =
1344         VMXNET3_READ_DRV_SHARED8(s->drv_shmem, devRead.misc.numRxQueues);
1345 
1346     VMW_CFPRN("Number of TX/RX queues %u/%u", s->txq_num, s->rxq_num);
1347     assert(s->txq_num <= VMXNET3_DEVICE_MAX_TX_QUEUES);
1348 
1349     qdescr_table_pa =
1350         VMXNET3_READ_DRV_SHARED64(s->drv_shmem, devRead.misc.queueDescPA);
1351     VMW_CFPRN("TX queues descriptors table is at 0x%" PRIx64, qdescr_table_pa);
1352 
1353     /*
1354      * Worst-case scenario is a packet that holds all TX rings space so
1355      * we calculate total size of all TX rings for max TX fragments number
1356      */
1357     s->max_tx_frags = 0;
1358 
1359     /* TX queues */
1360     for (i = 0; i < s->txq_num; i++) {
1361         hwaddr qdescr_pa =
1362             qdescr_table_pa + i * sizeof(struct Vmxnet3_TxQueueDesc);
1363 
1364         /* Read interrupt number for this TX queue */
1365         s->txq_descr[i].intr_idx =
1366             VMXNET3_READ_TX_QUEUE_DESCR8(qdescr_pa, conf.intrIdx);
1367 
1368         VMW_CFPRN("TX Queue %d interrupt: %d", i, s->txq_descr[i].intr_idx);
1369 
1370         /* Read rings memory locations for TX queues */
1371         pa = VMXNET3_READ_TX_QUEUE_DESCR64(qdescr_pa, conf.txRingBasePA);
1372         size = VMXNET3_READ_TX_QUEUE_DESCR32(qdescr_pa, conf.txRingSize);
1373 
1374         vmxnet3_ring_init(&s->txq_descr[i].tx_ring, pa, size,
1375                           sizeof(struct Vmxnet3_TxDesc), false);
1376         VMXNET3_RING_DUMP(VMW_CFPRN, "TX", i, &s->txq_descr[i].tx_ring);
1377 
1378         s->max_tx_frags += size;
1379 
1380         /* TXC ring */
1381         pa = VMXNET3_READ_TX_QUEUE_DESCR64(qdescr_pa, conf.compRingBasePA);
1382         size = VMXNET3_READ_TX_QUEUE_DESCR32(qdescr_pa, conf.compRingSize);
1383         vmxnet3_ring_init(&s->txq_descr[i].comp_ring, pa, size,
1384                           sizeof(struct Vmxnet3_TxCompDesc), true);
1385         VMXNET3_RING_DUMP(VMW_CFPRN, "TXC", i, &s->txq_descr[i].comp_ring);
1386 
1387         s->txq_descr[i].tx_stats_pa =
1388             qdescr_pa + offsetof(struct Vmxnet3_TxQueueDesc, stats);
1389 
1390         memset(&s->txq_descr[i].txq_stats, 0,
1391                sizeof(s->txq_descr[i].txq_stats));
1392 
1393         /* Fill device-managed parameters for queues */
1394         VMXNET3_WRITE_TX_QUEUE_DESCR32(qdescr_pa,
1395                                        ctrl.txThreshold,
1396                                        VMXNET3_DEF_TX_THRESHOLD);
1397     }
1398 
1399     /* Preallocate TX packet wrapper */
1400     VMW_CFPRN("Max TX fragments is %u", s->max_tx_frags);
1401     vmxnet_tx_pkt_init(&s->tx_pkt, s->max_tx_frags, s->peer_has_vhdr);
1402     vmxnet_rx_pkt_init(&s->rx_pkt, s->peer_has_vhdr);
1403 
1404     /* Read rings memory locations for RX queues */
1405     for (i = 0; i < s->rxq_num; i++) {
1406         int j;
1407         hwaddr qd_pa =
1408             qdescr_table_pa + s->txq_num * sizeof(struct Vmxnet3_TxQueueDesc) +
1409             i * sizeof(struct Vmxnet3_RxQueueDesc);
1410 
1411         /* Read interrupt number for this RX queue */
1412         s->rxq_descr[i].intr_idx =
1413             VMXNET3_READ_TX_QUEUE_DESCR8(qd_pa, conf.intrIdx);
1414 
1415         VMW_CFPRN("RX Queue %d interrupt: %d", i, s->rxq_descr[i].intr_idx);
1416 
1417         /* Read rings memory locations */
1418         for (j = 0; j < VMXNET3_RX_RINGS_PER_QUEUE; j++) {
1419             /* RX rings */
1420             pa = VMXNET3_READ_RX_QUEUE_DESCR64(qd_pa, conf.rxRingBasePA[j]);
1421             size = VMXNET3_READ_RX_QUEUE_DESCR32(qd_pa, conf.rxRingSize[j]);
1422             vmxnet3_ring_init(&s->rxq_descr[i].rx_ring[j], pa, size,
1423                               sizeof(struct Vmxnet3_RxDesc), false);
1424             VMW_CFPRN("RX queue %d:%d: Base: %" PRIx64 ", Size: %d",
1425                       i, j, pa, size);
1426         }
1427 
1428         /* RXC ring */
1429         pa = VMXNET3_READ_RX_QUEUE_DESCR64(qd_pa, conf.compRingBasePA);
1430         size = VMXNET3_READ_RX_QUEUE_DESCR32(qd_pa, conf.compRingSize);
1431         vmxnet3_ring_init(&s->rxq_descr[i].comp_ring, pa, size,
1432                           sizeof(struct Vmxnet3_RxCompDesc), true);
1433         VMW_CFPRN("RXC queue %d: Base: %" PRIx64 ", Size: %d", i, pa, size);
1434 
1435         s->rxq_descr[i].rx_stats_pa =
1436             qd_pa + offsetof(struct Vmxnet3_RxQueueDesc, stats);
1437         memset(&s->rxq_descr[i].rxq_stats, 0,
1438                sizeof(s->rxq_descr[i].rxq_stats));
1439     }
1440 
1441     /* Make sure everything is in place before device activation */
1442     smp_wmb();
1443 
1444     vmxnet3_reset_mac(s);
1445 
1446     s->device_active = true;
1447 }
1448 
1449 static void vmxnet3_handle_command(VMXNET3State *s, uint64_t cmd)
1450 {
1451     s->last_command = cmd;
1452 
1453     switch (cmd) {
1454     case VMXNET3_CMD_GET_PERM_MAC_HI:
1455         VMW_CBPRN("Set: Get upper part of permanent MAC");
1456         break;
1457 
1458     case VMXNET3_CMD_GET_PERM_MAC_LO:
1459         VMW_CBPRN("Set: Get lower part of permanent MAC");
1460         break;
1461 
1462     case VMXNET3_CMD_GET_STATS:
1463         VMW_CBPRN("Set: Get device statistics");
1464         vmxnet3_fill_stats(s);
1465         break;
1466 
1467     case VMXNET3_CMD_ACTIVATE_DEV:
1468         VMW_CBPRN("Set: Activating vmxnet3 device");
1469         vmxnet3_activate_device(s);
1470         break;
1471 
1472     case VMXNET3_CMD_UPDATE_RX_MODE:
1473         VMW_CBPRN("Set: Update rx mode");
1474         vmxnet3_update_rx_mode(s);
1475         break;
1476 
1477     case VMXNET3_CMD_UPDATE_VLAN_FILTERS:
1478         VMW_CBPRN("Set: Update VLAN filters");
1479         vmxnet3_update_vlan_filters(s);
1480         break;
1481 
1482     case VMXNET3_CMD_UPDATE_MAC_FILTERS:
1483         VMW_CBPRN("Set: Update MAC filters");
1484         vmxnet3_update_mcast_filters(s);
1485         break;
1486 
1487     case VMXNET3_CMD_UPDATE_FEATURE:
1488         VMW_CBPRN("Set: Update features");
1489         vmxnet3_update_features(s);
1490         break;
1491 
1492     case VMXNET3_CMD_UPDATE_PMCFG:
1493         VMW_CBPRN("Set: Update power management config");
1494         vmxnet3_update_pm_state(s);
1495         break;
1496 
1497     case VMXNET3_CMD_GET_LINK:
1498         VMW_CBPRN("Set: Get link");
1499         break;
1500 
1501     case VMXNET3_CMD_RESET_DEV:
1502         VMW_CBPRN("Set: Reset device");
1503         vmxnet3_reset(s);
1504         break;
1505 
1506     case VMXNET3_CMD_QUIESCE_DEV:
1507         VMW_CBPRN("Set: VMXNET3_CMD_QUIESCE_DEV - pause the device");
1508         vmxnet3_deactivate_device(s);
1509         break;
1510 
1511     case VMXNET3_CMD_GET_CONF_INTR:
1512         VMW_CBPRN("Set: VMXNET3_CMD_GET_CONF_INTR - interrupt configuration");
1513         break;
1514 
1515     default:
1516         VMW_CBPRN("Received unknown command: %" PRIx64, cmd);
1517         break;
1518     }
1519 }
1520 
1521 static uint64_t vmxnet3_get_command_status(VMXNET3State *s)
1522 {
1523     uint64_t ret;
1524 
1525     switch (s->last_command) {
1526     case VMXNET3_CMD_ACTIVATE_DEV:
1527         ret = (s->device_active) ? 0 : -1;
1528         VMW_CFPRN("Device active: %" PRIx64, ret);
1529         break;
1530 
1531     case VMXNET3_CMD_RESET_DEV:
1532     case VMXNET3_CMD_QUIESCE_DEV:
1533     case VMXNET3_CMD_GET_QUEUE_STATUS:
1534         ret = 0;
1535         break;
1536 
1537     case VMXNET3_CMD_GET_LINK:
1538         ret = s->link_status_and_speed;
1539         VMW_CFPRN("Link and speed: %" PRIx64, ret);
1540         break;
1541 
1542     case VMXNET3_CMD_GET_PERM_MAC_LO:
1543         ret = vmxnet3_get_mac_low(&s->perm_mac);
1544         break;
1545 
1546     case VMXNET3_CMD_GET_PERM_MAC_HI:
1547         ret = vmxnet3_get_mac_high(&s->perm_mac);
1548         break;
1549 
1550     case VMXNET3_CMD_GET_CONF_INTR:
1551         ret = vmxnet3_get_interrupt_config(s);
1552         break;
1553 
1554     default:
1555         VMW_WRPRN("Received request for unknown command: %x", s->last_command);
1556         ret = -1;
1557         break;
1558     }
1559 
1560     return ret;
1561 }
1562 
1563 static void vmxnet3_set_events(VMXNET3State *s, uint32_t val)
1564 {
1565     uint32_t events;
1566 
1567     VMW_CBPRN("Setting events: 0x%x", val);
1568     events = VMXNET3_READ_DRV_SHARED32(s->drv_shmem, ecr) | val;
1569     VMXNET3_WRITE_DRV_SHARED32(s->drv_shmem, ecr, events);
1570 }
1571 
1572 static void vmxnet3_ack_events(VMXNET3State *s, uint32_t val)
1573 {
1574     uint32_t events;
1575 
1576     VMW_CBPRN("Clearing 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
1582 vmxnet3_io_bar1_write(void *opaque,
1583                       hwaddr addr,
1584                       uint64_t val,
1585                       unsigned size)
1586 {
1587     VMXNET3State *s = opaque;
1588 
1589     switch (addr) {
1590     /* Vmxnet3 Revision Report Selection */
1591     case VMXNET3_REG_VRRS:
1592         VMW_CBPRN("Write BAR1 [VMXNET3_REG_VRRS] = %" PRIx64 ", size %d",
1593                   val, size);
1594         break;
1595 
1596     /* UPT Version Report Selection */
1597     case VMXNET3_REG_UVRS:
1598         VMW_CBPRN("Write BAR1 [VMXNET3_REG_UVRS] = %" PRIx64 ", size %d",
1599                   val, size);
1600         break;
1601 
1602     /* Driver Shared Address Low */
1603     case VMXNET3_REG_DSAL:
1604         VMW_CBPRN("Write BAR1 [VMXNET3_REG_DSAL] = %" PRIx64 ", size %d",
1605                   val, size);
1606         /*
1607          * Guest driver will first write the low part of the shared
1608          * memory address. We save it to temp variable and set the
1609          * shared address only after we get the high part
1610          */
1611         if (0 == val) {
1612             s->device_active = false;
1613         }
1614         s->temp_shared_guest_driver_memory = val;
1615         s->drv_shmem = 0;
1616         break;
1617 
1618     /* Driver Shared Address High */
1619     case VMXNET3_REG_DSAH:
1620         VMW_CBPRN("Write BAR1 [VMXNET3_REG_DSAH] = %" PRIx64 ", size %d",
1621                   val, size);
1622         /*
1623          * Set the shared memory between guest driver and device.
1624          * We already should have low address part.
1625          */
1626         s->drv_shmem = s->temp_shared_guest_driver_memory | (val << 32);
1627         break;
1628 
1629     /* Command */
1630     case VMXNET3_REG_CMD:
1631         VMW_CBPRN("Write BAR1 [VMXNET3_REG_CMD] = %" PRIx64 ", size %d",
1632                   val, size);
1633         vmxnet3_handle_command(s, val);
1634         break;
1635 
1636     /* MAC Address Low */
1637     case VMXNET3_REG_MACL:
1638         VMW_CBPRN("Write BAR1 [VMXNET3_REG_MACL] = %" PRIx64 ", size %d",
1639                   val, size);
1640         s->temp_mac = val;
1641         break;
1642 
1643     /* MAC Address High */
1644     case VMXNET3_REG_MACH:
1645         VMW_CBPRN("Write BAR1 [VMXNET3_REG_MACH] = %" PRIx64 ", size %d",
1646                   val, size);
1647         vmxnet3_set_variable_mac(s, val, s->temp_mac);
1648         break;
1649 
1650     /* Interrupt Cause Register */
1651     case VMXNET3_REG_ICR:
1652         VMW_CBPRN("Write BAR1 [VMXNET3_REG_ICR] = %" PRIx64 ", size %d",
1653                   val, size);
1654         assert(false);
1655         break;
1656 
1657     /* Event Cause Register */
1658     case VMXNET3_REG_ECR:
1659         VMW_CBPRN("Write BAR1 [VMXNET3_REG_ECR] = %" PRIx64 ", size %d",
1660                   val, size);
1661         vmxnet3_ack_events(s, val);
1662         break;
1663 
1664     default:
1665         VMW_CBPRN("Unknown Write to BAR1 [%" PRIx64 "] = %" PRIx64 ", size %d",
1666                   addr, val, size);
1667         break;
1668     }
1669 }
1670 
1671 static uint64_t
1672 vmxnet3_io_bar1_read(void *opaque, hwaddr addr, unsigned size)
1673 {
1674         VMXNET3State *s = opaque;
1675         uint64_t ret = 0;
1676 
1677         switch (addr) {
1678         /* Vmxnet3 Revision Report Selection */
1679         case VMXNET3_REG_VRRS:
1680             VMW_CBPRN("Read BAR1 [VMXNET3_REG_VRRS], size %d", size);
1681             ret = VMXNET3_DEVICE_REVISION;
1682             break;
1683 
1684         /* UPT Version Report Selection */
1685         case VMXNET3_REG_UVRS:
1686             VMW_CBPRN("Read BAR1 [VMXNET3_REG_UVRS], size %d", size);
1687             ret = VMXNET3_DEVICE_VERSION;
1688             break;
1689 
1690         /* Command */
1691         case VMXNET3_REG_CMD:
1692             VMW_CBPRN("Read BAR1 [VMXNET3_REG_CMD], size %d", size);
1693             ret = vmxnet3_get_command_status(s);
1694             break;
1695 
1696         /* MAC Address Low */
1697         case VMXNET3_REG_MACL:
1698             VMW_CBPRN("Read BAR1 [VMXNET3_REG_MACL], size %d", size);
1699             ret = vmxnet3_get_mac_low(&s->conf.macaddr);
1700             break;
1701 
1702         /* MAC Address High */
1703         case VMXNET3_REG_MACH:
1704             VMW_CBPRN("Read BAR1 [VMXNET3_REG_MACH], size %d", size);
1705             ret = vmxnet3_get_mac_high(&s->conf.macaddr);
1706             break;
1707 
1708         /*
1709          * Interrupt Cause Register
1710          * Used for legacy interrupts only so interrupt index always 0
1711          */
1712         case VMXNET3_REG_ICR:
1713             VMW_CBPRN("Read BAR1 [VMXNET3_REG_ICR], size %d", size);
1714             if (vmxnet3_interrupt_asserted(s, 0)) {
1715                 vmxnet3_clear_interrupt(s, 0);
1716                 ret = true;
1717             } else {
1718                 ret = false;
1719             }
1720             break;
1721 
1722         default:
1723             VMW_CBPRN("Unknow read BAR1[%" PRIx64 "], %d bytes", addr, size);
1724             break;
1725         }
1726 
1727         return ret;
1728 }
1729 
1730 static int
1731 vmxnet3_can_receive(NetClientState *nc)
1732 {
1733     VMXNET3State *s = qemu_get_nic_opaque(nc);
1734     return s->device_active &&
1735            VMXNET_FLAG_IS_SET(s->link_status_and_speed, VMXNET3_LINK_STATUS_UP);
1736 }
1737 
1738 static inline bool
1739 vmxnet3_is_registered_vlan(VMXNET3State *s, const void *data)
1740 {
1741     uint16_t vlan_tag = eth_get_pkt_tci(data) & VLAN_VID_MASK;
1742     if (IS_SPECIAL_VLAN_ID(vlan_tag)) {
1743         return true;
1744     }
1745 
1746     return VMXNET3_VFTABLE_ENTRY_IS_SET(s->vlan_table, vlan_tag);
1747 }
1748 
1749 static bool
1750 vmxnet3_is_allowed_mcast_group(VMXNET3State *s, const uint8_t *group_mac)
1751 {
1752     int i;
1753     for (i = 0; i < s->mcast_list_len; i++) {
1754         if (!memcmp(group_mac, s->mcast_list[i].a, sizeof(s->mcast_list[i]))) {
1755             return true;
1756         }
1757     }
1758     return false;
1759 }
1760 
1761 static bool
1762 vmxnet3_rx_filter_may_indicate(VMXNET3State *s, const void *data,
1763     size_t size)
1764 {
1765     struct eth_header *ehdr = PKT_GET_ETH_HDR(data);
1766 
1767     if (VMXNET_FLAG_IS_SET(s->rx_mode, VMXNET3_RXM_PROMISC)) {
1768         return true;
1769     }
1770 
1771     if (!vmxnet3_is_registered_vlan(s, data)) {
1772         return false;
1773     }
1774 
1775     switch (vmxnet_rx_pkt_get_packet_type(s->rx_pkt)) {
1776     case ETH_PKT_UCAST:
1777         if (!VMXNET_FLAG_IS_SET(s->rx_mode, VMXNET3_RXM_UCAST)) {
1778             return false;
1779         }
1780         if (memcmp(s->conf.macaddr.a, ehdr->h_dest, ETH_ALEN)) {
1781             return false;
1782         }
1783         break;
1784 
1785     case ETH_PKT_BCAST:
1786         if (!VMXNET_FLAG_IS_SET(s->rx_mode, VMXNET3_RXM_BCAST)) {
1787             return false;
1788         }
1789         break;
1790 
1791     case ETH_PKT_MCAST:
1792         if (VMXNET_FLAG_IS_SET(s->rx_mode, VMXNET3_RXM_ALL_MULTI)) {
1793             return true;
1794         }
1795         if (!VMXNET_FLAG_IS_SET(s->rx_mode, VMXNET3_RXM_MCAST)) {
1796             return false;
1797         }
1798         if (!vmxnet3_is_allowed_mcast_group(s, ehdr->h_dest)) {
1799             return false;
1800         }
1801         break;
1802 
1803     default:
1804         assert(false);
1805     }
1806 
1807     return true;
1808 }
1809 
1810 static ssize_t
1811 vmxnet3_receive(NetClientState *nc, const uint8_t *buf, size_t size)
1812 {
1813     VMXNET3State *s = qemu_get_nic_opaque(nc);
1814     size_t bytes_indicated;
1815 
1816     if (!vmxnet3_can_receive(nc)) {
1817         VMW_PKPRN("Cannot receive now");
1818         return -1;
1819     }
1820 
1821     if (s->peer_has_vhdr) {
1822         vmxnet_rx_pkt_set_vhdr(s->rx_pkt, (struct virtio_net_hdr *)buf);
1823         buf += sizeof(struct virtio_net_hdr);
1824         size -= sizeof(struct virtio_net_hdr);
1825     }
1826 
1827     vmxnet_rx_pkt_set_packet_type(s->rx_pkt,
1828         get_eth_packet_type(PKT_GET_ETH_HDR(buf)));
1829 
1830     if (vmxnet3_rx_filter_may_indicate(s, buf, size)) {
1831         vmxnet_rx_pkt_attach_data(s->rx_pkt, buf, size, s->rx_vlan_stripping);
1832         bytes_indicated = vmxnet3_indicate_packet(s) ? size : -1;
1833         if (bytes_indicated < size) {
1834             VMW_PKPRN("RX: %lu of %lu bytes indicated", bytes_indicated, size);
1835         }
1836     } else {
1837         VMW_PKPRN("Packet dropped by RX filter");
1838         bytes_indicated = size;
1839     }
1840 
1841     assert(size > 0);
1842     assert(bytes_indicated != 0);
1843     return bytes_indicated;
1844 }
1845 
1846 static void vmxnet3_cleanup(NetClientState *nc)
1847 {
1848     VMXNET3State *s = qemu_get_nic_opaque(nc);
1849     s->nic = NULL;
1850 }
1851 
1852 static void vmxnet3_set_link_status(NetClientState *nc)
1853 {
1854     VMXNET3State *s = qemu_get_nic_opaque(nc);
1855 
1856     if (nc->link_down) {
1857         s->link_status_and_speed &= ~VMXNET3_LINK_STATUS_UP;
1858     } else {
1859         s->link_status_and_speed |= VMXNET3_LINK_STATUS_UP;
1860     }
1861 
1862     vmxnet3_set_events(s, VMXNET3_ECR_LINK);
1863     vmxnet3_trigger_interrupt(s, s->event_int_idx);
1864 }
1865 
1866 static NetClientInfo net_vmxnet3_info = {
1867         .type = NET_CLIENT_OPTIONS_KIND_NIC,
1868         .size = sizeof(NICState),
1869         .can_receive = vmxnet3_can_receive,
1870         .receive = vmxnet3_receive,
1871         .cleanup = vmxnet3_cleanup,
1872         .link_status_changed = vmxnet3_set_link_status,
1873 };
1874 
1875 static bool vmxnet3_peer_has_vnet_hdr(VMXNET3State *s)
1876 {
1877     NetClientState *peer = qemu_get_queue(s->nic)->peer;
1878 
1879     if ((NULL != peer)                              &&
1880         (peer->info->type == NET_CLIENT_OPTIONS_KIND_TAP)   &&
1881         tap_has_vnet_hdr(peer)) {
1882         return true;
1883     }
1884 
1885     VMW_WRPRN("Peer has no virtio extension. Task offloads will be emulated.");
1886     return false;
1887 }
1888 
1889 static void vmxnet3_net_uninit(VMXNET3State *s)
1890 {
1891     g_free(s->mcast_list);
1892     vmxnet_tx_pkt_reset(s->tx_pkt);
1893     vmxnet_tx_pkt_uninit(s->tx_pkt);
1894     vmxnet_rx_pkt_uninit(s->rx_pkt);
1895     qemu_del_net_client(qemu_get_queue(s->nic));
1896 }
1897 
1898 static void vmxnet3_net_init(VMXNET3State *s)
1899 {
1900     DeviceState *d = DEVICE(s);
1901 
1902     VMW_CBPRN("vmxnet3_net_init called...");
1903 
1904     qemu_macaddr_default_if_unset(&s->conf.macaddr);
1905 
1906     /* Windows guest will query the address that was set on init */
1907     memcpy(&s->perm_mac.a, &s->conf.macaddr.a, sizeof(s->perm_mac.a));
1908 
1909     s->mcast_list = NULL;
1910     s->mcast_list_len = 0;
1911 
1912     s->link_status_and_speed = VMXNET3_LINK_SPEED | VMXNET3_LINK_STATUS_UP;
1913 
1914     VMW_CFPRN("Permanent MAC: " MAC_FMT, MAC_ARG(s->perm_mac.a));
1915 
1916     s->nic = qemu_new_nic(&net_vmxnet3_info, &s->conf,
1917                           object_get_typename(OBJECT(s)),
1918                           d->id, s);
1919 
1920     s->peer_has_vhdr = vmxnet3_peer_has_vnet_hdr(s);
1921     s->tx_sop = true;
1922     s->skip_current_tx_pkt = false;
1923     s->tx_pkt = NULL;
1924     s->rx_pkt = NULL;
1925     s->rx_vlan_stripping = false;
1926     s->lro_supported = false;
1927 
1928     if (s->peer_has_vhdr) {
1929         tap_set_vnet_hdr_len(qemu_get_queue(s->nic)->peer,
1930             sizeof(struct virtio_net_hdr));
1931 
1932         tap_using_vnet_hdr(qemu_get_queue(s->nic)->peer, 1);
1933     }
1934 
1935     qemu_format_nic_info_str(qemu_get_queue(s->nic), s->conf.macaddr.a);
1936 }
1937 
1938 static void
1939 vmxnet3_unuse_msix_vectors(VMXNET3State *s, int num_vectors)
1940 {
1941     PCIDevice *d = PCI_DEVICE(s);
1942     int i;
1943     for (i = 0; i < num_vectors; i++) {
1944         msix_vector_unuse(d, i);
1945     }
1946 }
1947 
1948 static bool
1949 vmxnet3_use_msix_vectors(VMXNET3State *s, int num_vectors)
1950 {
1951     PCIDevice *d = PCI_DEVICE(s);
1952     int i;
1953     for (i = 0; i < num_vectors; i++) {
1954         int res = msix_vector_use(d, i);
1955         if (0 > res) {
1956             VMW_WRPRN("Failed to use MSI-X vector %d, error %d", i, res);
1957             vmxnet3_unuse_msix_vectors(s, i);
1958             return false;
1959         }
1960     }
1961     return true;
1962 }
1963 
1964 static bool
1965 vmxnet3_init_msix(VMXNET3State *s)
1966 {
1967     PCIDevice *d = PCI_DEVICE(s);
1968     int res = msix_init(d, VMXNET3_MAX_INTRS,
1969                         &s->msix_bar,
1970                         VMXNET3_MSIX_BAR_IDX, VMXNET3_OFF_MSIX_TABLE,
1971                         &s->msix_bar,
1972                         VMXNET3_MSIX_BAR_IDX, VMXNET3_OFF_MSIX_PBA,
1973                         0);
1974 
1975     if (0 > res) {
1976         VMW_WRPRN("Failed to initialize MSI-X, error %d", res);
1977         s->msix_used = false;
1978     } else {
1979         if (!vmxnet3_use_msix_vectors(s, VMXNET3_MAX_INTRS)) {
1980             VMW_WRPRN("Failed to use MSI-X vectors, error %d", res);
1981             msix_uninit(d, &s->msix_bar, &s->msix_bar);
1982             s->msix_used = false;
1983         } else {
1984             s->msix_used = true;
1985         }
1986     }
1987     return s->msix_used;
1988 }
1989 
1990 static void
1991 vmxnet3_cleanup_msix(VMXNET3State *s)
1992 {
1993     PCIDevice *d = PCI_DEVICE(s);
1994 
1995     if (s->msix_used) {
1996         msix_vector_unuse(d, VMXNET3_MAX_INTRS);
1997         msix_uninit(d, &s->msix_bar, &s->msix_bar);
1998     }
1999 }
2000 
2001 #define VMXNET3_MSI_NUM_VECTORS   (1)
2002 #define VMXNET3_MSI_OFFSET        (0x50)
2003 #define VMXNET3_USE_64BIT         (true)
2004 #define VMXNET3_PER_VECTOR_MASK   (false)
2005 
2006 static bool
2007 vmxnet3_init_msi(VMXNET3State *s)
2008 {
2009     PCIDevice *d = PCI_DEVICE(s);
2010     int res;
2011 
2012     res = msi_init(d, VMXNET3_MSI_OFFSET, VMXNET3_MSI_NUM_VECTORS,
2013                    VMXNET3_USE_64BIT, VMXNET3_PER_VECTOR_MASK);
2014     if (0 > res) {
2015         VMW_WRPRN("Failed to initialize MSI, error %d", res);
2016         s->msi_used = false;
2017     } else {
2018         s->msi_used = true;
2019     }
2020 
2021     return s->msi_used;
2022 }
2023 
2024 static void
2025 vmxnet3_cleanup_msi(VMXNET3State *s)
2026 {
2027     PCIDevice *d = PCI_DEVICE(s);
2028 
2029     if (s->msi_used) {
2030         msi_uninit(d);
2031     }
2032 }
2033 
2034 static void
2035 vmxnet3_msix_save(QEMUFile *f, void *opaque)
2036 {
2037     PCIDevice *d = PCI_DEVICE(opaque);
2038     msix_save(d, f);
2039 }
2040 
2041 static int
2042 vmxnet3_msix_load(QEMUFile *f, void *opaque, int version_id)
2043 {
2044     PCIDevice *d = PCI_DEVICE(opaque);
2045     msix_load(d, f);
2046     return 0;
2047 }
2048 
2049 static const MemoryRegionOps b0_ops = {
2050     .read = vmxnet3_io_bar0_read,
2051     .write = vmxnet3_io_bar0_write,
2052     .endianness = DEVICE_LITTLE_ENDIAN,
2053     .impl = {
2054             .min_access_size = 4,
2055             .max_access_size = 4,
2056     },
2057 };
2058 
2059 static const MemoryRegionOps b1_ops = {
2060     .read = vmxnet3_io_bar1_read,
2061     .write = vmxnet3_io_bar1_write,
2062     .endianness = DEVICE_LITTLE_ENDIAN,
2063     .impl = {
2064             .min_access_size = 4,
2065             .max_access_size = 4,
2066     },
2067 };
2068 
2069 static int vmxnet3_pci_init(PCIDevice *pci_dev)
2070 {
2071     DeviceState *dev = DEVICE(pci_dev);
2072     VMXNET3State *s = VMXNET3(pci_dev);
2073 
2074     VMW_CBPRN("Starting init...");
2075 
2076     memory_region_init_io(&s->bar0, &b0_ops, s,
2077                           "vmxnet3-b0", VMXNET3_PT_REG_SIZE);
2078     pci_register_bar(pci_dev, VMXNET3_BAR0_IDX,
2079                      PCI_BASE_ADDRESS_SPACE_MEMORY, &s->bar0);
2080 
2081     memory_region_init_io(&s->bar1, &b1_ops, s,
2082                           "vmxnet3-b1", VMXNET3_VD_REG_SIZE);
2083     pci_register_bar(pci_dev, VMXNET3_BAR1_IDX,
2084                      PCI_BASE_ADDRESS_SPACE_MEMORY, &s->bar1);
2085 
2086     memory_region_init(&s->msix_bar, "vmxnet3-msix-bar",
2087                        VMXNET3_MSIX_BAR_SIZE);
2088     pci_register_bar(pci_dev, VMXNET3_MSIX_BAR_IDX,
2089                      PCI_BASE_ADDRESS_SPACE_MEMORY, &s->msix_bar);
2090 
2091     vmxnet3_reset_interrupt_states(s);
2092 
2093     /* Interrupt pin A */
2094     pci_dev->config[PCI_INTERRUPT_PIN] = 0x01;
2095 
2096     if (!vmxnet3_init_msix(s)) {
2097         VMW_WRPRN("Failed to initialize MSI-X, configuration is inconsistent.");
2098     }
2099 
2100     if (!vmxnet3_init_msi(s)) {
2101         VMW_WRPRN("Failed to initialize MSI, configuration is inconsistent.");
2102     }
2103 
2104     vmxnet3_net_init(s);
2105 
2106     register_savevm(dev, "vmxnet3-msix", -1, 1,
2107                     vmxnet3_msix_save, vmxnet3_msix_load, s);
2108 
2109     add_boot_device_path(s->conf.bootindex, dev, "/ethernet-phy@0");
2110 
2111     return 0;
2112 }
2113 
2114 
2115 static void vmxnet3_pci_uninit(PCIDevice *pci_dev)
2116 {
2117     DeviceState *dev = DEVICE(pci_dev);
2118     VMXNET3State *s = VMXNET3(pci_dev);
2119 
2120     VMW_CBPRN("Starting uninit...");
2121 
2122     unregister_savevm(dev, "vmxnet3-msix", s);
2123 
2124     vmxnet3_net_uninit(s);
2125 
2126     vmxnet3_cleanup_msix(s);
2127 
2128     vmxnet3_cleanup_msi(s);
2129 
2130     memory_region_destroy(&s->bar0);
2131     memory_region_destroy(&s->bar1);
2132     memory_region_destroy(&s->msix_bar);
2133 }
2134 
2135 static void vmxnet3_qdev_reset(DeviceState *dev)
2136 {
2137     PCIDevice *d = PCI_DEVICE(dev);
2138     VMXNET3State *s = VMXNET3(d);
2139 
2140     VMW_CBPRN("Starting QDEV reset...");
2141     vmxnet3_reset(s);
2142 }
2143 
2144 static bool vmxnet3_mc_list_needed(void *opaque)
2145 {
2146     return true;
2147 }
2148 
2149 static int vmxnet3_mcast_list_pre_load(void *opaque)
2150 {
2151     VMXNET3State *s = opaque;
2152 
2153     s->mcast_list = g_malloc(s->mcast_list_buff_size);
2154 
2155     return 0;
2156 }
2157 
2158 
2159 static void vmxnet3_pre_save(void *opaque)
2160 {
2161     VMXNET3State *s = opaque;
2162 
2163     s->mcast_list_buff_size = s->mcast_list_len * sizeof(MACAddr);
2164 }
2165 
2166 static const VMStateDescription vmxstate_vmxnet3_mcast_list = {
2167     .name = "vmxnet3/mcast_list",
2168     .version_id = 1,
2169     .minimum_version_id = 1,
2170     .minimum_version_id_old = 1,
2171     .pre_load = vmxnet3_mcast_list_pre_load,
2172     .fields = (VMStateField[]) {
2173         VMSTATE_VBUFFER_UINT32(mcast_list, VMXNET3State, 0, NULL, 0,
2174             mcast_list_buff_size),
2175         VMSTATE_END_OF_LIST()
2176     }
2177 };
2178 
2179 static void vmxnet3_get_ring_from_file(QEMUFile *f, Vmxnet3Ring *r)
2180 {
2181     r->pa = qemu_get_be64(f);
2182     r->size = qemu_get_be32(f);
2183     r->cell_size = qemu_get_be32(f);
2184     r->next = qemu_get_be32(f);
2185     r->gen = qemu_get_byte(f);
2186 }
2187 
2188 static void vmxnet3_put_ring_to_file(QEMUFile *f, Vmxnet3Ring *r)
2189 {
2190     qemu_put_be64(f, r->pa);
2191     qemu_put_be32(f, r->size);
2192     qemu_put_be32(f, r->cell_size);
2193     qemu_put_be32(f, r->next);
2194     qemu_put_byte(f, r->gen);
2195 }
2196 
2197 static void vmxnet3_get_tx_stats_from_file(QEMUFile *f,
2198     struct UPT1_TxStats *tx_stat)
2199 {
2200     tx_stat->TSOPktsTxOK = qemu_get_be64(f);
2201     tx_stat->TSOBytesTxOK = qemu_get_be64(f);
2202     tx_stat->ucastPktsTxOK = qemu_get_be64(f);
2203     tx_stat->ucastBytesTxOK = qemu_get_be64(f);
2204     tx_stat->mcastPktsTxOK = qemu_get_be64(f);
2205     tx_stat->mcastBytesTxOK = qemu_get_be64(f);
2206     tx_stat->bcastPktsTxOK = qemu_get_be64(f);
2207     tx_stat->bcastBytesTxOK = qemu_get_be64(f);
2208     tx_stat->pktsTxError = qemu_get_be64(f);
2209     tx_stat->pktsTxDiscard = qemu_get_be64(f);
2210 }
2211 
2212 static void vmxnet3_put_tx_stats_to_file(QEMUFile *f,
2213     struct UPT1_TxStats *tx_stat)
2214 {
2215     qemu_put_be64(f, tx_stat->TSOPktsTxOK);
2216     qemu_put_be64(f, tx_stat->TSOBytesTxOK);
2217     qemu_put_be64(f, tx_stat->ucastPktsTxOK);
2218     qemu_put_be64(f, tx_stat->ucastBytesTxOK);
2219     qemu_put_be64(f, tx_stat->mcastPktsTxOK);
2220     qemu_put_be64(f, tx_stat->mcastBytesTxOK);
2221     qemu_put_be64(f, tx_stat->bcastPktsTxOK);
2222     qemu_put_be64(f, tx_stat->bcastBytesTxOK);
2223     qemu_put_be64(f, tx_stat->pktsTxError);
2224     qemu_put_be64(f, tx_stat->pktsTxDiscard);
2225 }
2226 
2227 static int vmxnet3_get_txq_descr(QEMUFile *f, void *pv, size_t size)
2228 {
2229     Vmxnet3TxqDescr *r = pv;
2230 
2231     vmxnet3_get_ring_from_file(f, &r->tx_ring);
2232     vmxnet3_get_ring_from_file(f, &r->comp_ring);
2233     r->intr_idx = qemu_get_byte(f);
2234     r->tx_stats_pa = qemu_get_be64(f);
2235 
2236     vmxnet3_get_tx_stats_from_file(f, &r->txq_stats);
2237 
2238     return 0;
2239 }
2240 
2241 static void vmxnet3_put_txq_descr(QEMUFile *f, void *pv, size_t size)
2242 {
2243     Vmxnet3TxqDescr *r = pv;
2244 
2245     vmxnet3_put_ring_to_file(f, &r->tx_ring);
2246     vmxnet3_put_ring_to_file(f, &r->comp_ring);
2247     qemu_put_byte(f, r->intr_idx);
2248     qemu_put_be64(f, r->tx_stats_pa);
2249     vmxnet3_put_tx_stats_to_file(f, &r->txq_stats);
2250 }
2251 
2252 const VMStateInfo txq_descr_info = {
2253     .name = "txq_descr",
2254     .get = vmxnet3_get_txq_descr,
2255     .put = vmxnet3_put_txq_descr
2256 };
2257 
2258 static void vmxnet3_get_rx_stats_from_file(QEMUFile *f,
2259     struct UPT1_RxStats *rx_stat)
2260 {
2261     rx_stat->LROPktsRxOK = qemu_get_be64(f);
2262     rx_stat->LROBytesRxOK = qemu_get_be64(f);
2263     rx_stat->ucastPktsRxOK = qemu_get_be64(f);
2264     rx_stat->ucastBytesRxOK = qemu_get_be64(f);
2265     rx_stat->mcastPktsRxOK = qemu_get_be64(f);
2266     rx_stat->mcastBytesRxOK = qemu_get_be64(f);
2267     rx_stat->bcastPktsRxOK = qemu_get_be64(f);
2268     rx_stat->bcastBytesRxOK = qemu_get_be64(f);
2269     rx_stat->pktsRxOutOfBuf = qemu_get_be64(f);
2270     rx_stat->pktsRxError = qemu_get_be64(f);
2271 }
2272 
2273 static void vmxnet3_put_rx_stats_to_file(QEMUFile *f,
2274     struct UPT1_RxStats *rx_stat)
2275 {
2276     qemu_put_be64(f, rx_stat->LROPktsRxOK);
2277     qemu_put_be64(f, rx_stat->LROBytesRxOK);
2278     qemu_put_be64(f, rx_stat->ucastPktsRxOK);
2279     qemu_put_be64(f, rx_stat->ucastBytesRxOK);
2280     qemu_put_be64(f, rx_stat->mcastPktsRxOK);
2281     qemu_put_be64(f, rx_stat->mcastBytesRxOK);
2282     qemu_put_be64(f, rx_stat->bcastPktsRxOK);
2283     qemu_put_be64(f, rx_stat->bcastBytesRxOK);
2284     qemu_put_be64(f, rx_stat->pktsRxOutOfBuf);
2285     qemu_put_be64(f, rx_stat->pktsRxError);
2286 }
2287 
2288 static int vmxnet3_get_rxq_descr(QEMUFile *f, void *pv, size_t size)
2289 {
2290     Vmxnet3RxqDescr *r = pv;
2291     int i;
2292 
2293     for (i = 0; i < VMXNET3_RX_RINGS_PER_QUEUE; i++) {
2294         vmxnet3_get_ring_from_file(f, &r->rx_ring[i]);
2295     }
2296 
2297     vmxnet3_get_ring_from_file(f, &r->comp_ring);
2298     r->intr_idx = qemu_get_byte(f);
2299     r->rx_stats_pa = qemu_get_be64(f);
2300 
2301     vmxnet3_get_rx_stats_from_file(f, &r->rxq_stats);
2302 
2303     return 0;
2304 }
2305 
2306 static void vmxnet3_put_rxq_descr(QEMUFile *f, void *pv, size_t size)
2307 {
2308     Vmxnet3RxqDescr *r = pv;
2309     int i;
2310 
2311     for (i = 0; i < VMXNET3_RX_RINGS_PER_QUEUE; i++) {
2312         vmxnet3_put_ring_to_file(f, &r->rx_ring[i]);
2313     }
2314 
2315     vmxnet3_put_ring_to_file(f, &r->comp_ring);
2316     qemu_put_byte(f, r->intr_idx);
2317     qemu_put_be64(f, r->rx_stats_pa);
2318     vmxnet3_put_rx_stats_to_file(f, &r->rxq_stats);
2319 }
2320 
2321 static int vmxnet3_post_load(void *opaque, int version_id)
2322 {
2323     VMXNET3State *s = opaque;
2324     PCIDevice *d = PCI_DEVICE(s);
2325 
2326     vmxnet_tx_pkt_init(&s->tx_pkt, s->max_tx_frags, s->peer_has_vhdr);
2327     vmxnet_rx_pkt_init(&s->rx_pkt, s->peer_has_vhdr);
2328 
2329     if (s->msix_used) {
2330         if  (!vmxnet3_use_msix_vectors(s, VMXNET3_MAX_INTRS)) {
2331             VMW_WRPRN("Failed to re-use MSI-X vectors");
2332             msix_uninit(d, &s->msix_bar, &s->msix_bar);
2333             s->msix_used = false;
2334             return -1;
2335         }
2336     }
2337 
2338     return 0;
2339 }
2340 
2341 const VMStateInfo rxq_descr_info = {
2342     .name = "rxq_descr",
2343     .get = vmxnet3_get_rxq_descr,
2344     .put = vmxnet3_put_rxq_descr
2345 };
2346 
2347 static int vmxnet3_get_int_state(QEMUFile *f, void *pv, size_t size)
2348 {
2349     Vmxnet3IntState *r = pv;
2350 
2351     r->is_masked = qemu_get_byte(f);
2352     r->is_pending = qemu_get_byte(f);
2353     r->is_asserted = qemu_get_byte(f);
2354 
2355     return 0;
2356 }
2357 
2358 static void vmxnet3_put_int_state(QEMUFile *f, void *pv, size_t size)
2359 {
2360     Vmxnet3IntState *r = pv;
2361 
2362     qemu_put_byte(f, r->is_masked);
2363     qemu_put_byte(f, r->is_pending);
2364     qemu_put_byte(f, r->is_asserted);
2365 }
2366 
2367 const VMStateInfo int_state_info = {
2368     .name = "int_state",
2369     .get = vmxnet3_get_int_state,
2370     .put = vmxnet3_put_int_state
2371 };
2372 
2373 static const VMStateDescription vmstate_vmxnet3 = {
2374     .name = "vmxnet3",
2375     .version_id = 1,
2376     .minimum_version_id = 1,
2377     .minimum_version_id_old = 1,
2378     .pre_save = vmxnet3_pre_save,
2379     .post_load = vmxnet3_post_load,
2380     .fields      = (VMStateField[]) {
2381             VMSTATE_PCI_DEVICE(parent_obj, VMXNET3State),
2382             VMSTATE_BOOL(rx_packets_compound, VMXNET3State),
2383             VMSTATE_BOOL(rx_vlan_stripping, VMXNET3State),
2384             VMSTATE_BOOL(lro_supported, VMXNET3State),
2385             VMSTATE_UINT32(rx_mode, VMXNET3State),
2386             VMSTATE_UINT32(mcast_list_len, VMXNET3State),
2387             VMSTATE_UINT32(mcast_list_buff_size, VMXNET3State),
2388             VMSTATE_UINT32_ARRAY(vlan_table, VMXNET3State, VMXNET3_VFT_SIZE),
2389             VMSTATE_UINT32(mtu, VMXNET3State),
2390             VMSTATE_UINT16(max_rx_frags, VMXNET3State),
2391             VMSTATE_UINT32(max_tx_frags, VMXNET3State),
2392             VMSTATE_UINT8(event_int_idx, VMXNET3State),
2393             VMSTATE_BOOL(auto_int_masking, VMXNET3State),
2394             VMSTATE_UINT8(txq_num, VMXNET3State),
2395             VMSTATE_UINT8(rxq_num, VMXNET3State),
2396             VMSTATE_UINT32(device_active, VMXNET3State),
2397             VMSTATE_UINT32(last_command, VMXNET3State),
2398             VMSTATE_UINT32(link_status_and_speed, VMXNET3State),
2399             VMSTATE_UINT32(temp_mac, VMXNET3State),
2400             VMSTATE_UINT64(drv_shmem, VMXNET3State),
2401             VMSTATE_UINT64(temp_shared_guest_driver_memory, VMXNET3State),
2402 
2403             VMSTATE_ARRAY(txq_descr, VMXNET3State,
2404                 VMXNET3_DEVICE_MAX_TX_QUEUES, 0, txq_descr_info,
2405                 Vmxnet3TxqDescr),
2406             VMSTATE_ARRAY(rxq_descr, VMXNET3State,
2407                 VMXNET3_DEVICE_MAX_RX_QUEUES, 0, rxq_descr_info,
2408                 Vmxnet3RxqDescr),
2409             VMSTATE_ARRAY(interrupt_states, VMXNET3State, VMXNET3_MAX_INTRS,
2410                 0, int_state_info, Vmxnet3IntState),
2411 
2412             VMSTATE_END_OF_LIST()
2413     },
2414     .subsections = (VMStateSubsection[]) {
2415         {
2416             .vmsd = &vmxstate_vmxnet3_mcast_list,
2417             .needed = vmxnet3_mc_list_needed
2418         },
2419         {
2420             /* empty element. */
2421         }
2422     }
2423 };
2424 
2425 static void
2426 vmxnet3_write_config(PCIDevice *pci_dev, uint32_t addr, uint32_t val, int len)
2427 {
2428     pci_default_write_config(pci_dev, addr, val, len);
2429     msix_write_config(pci_dev, addr, val, len);
2430     msi_write_config(pci_dev, addr, val, len);
2431 }
2432 
2433 static Property vmxnet3_properties[] = {
2434     DEFINE_NIC_PROPERTIES(VMXNET3State, conf),
2435     DEFINE_PROP_END_OF_LIST(),
2436 };
2437 
2438 static void vmxnet3_class_init(ObjectClass *class, void *data)
2439 {
2440     DeviceClass *dc = DEVICE_CLASS(class);
2441     PCIDeviceClass *c = PCI_DEVICE_CLASS(class);
2442 
2443     c->init = vmxnet3_pci_init;
2444     c->exit = vmxnet3_pci_uninit;
2445     c->vendor_id = PCI_VENDOR_ID_VMWARE;
2446     c->device_id = PCI_DEVICE_ID_VMWARE_VMXNET3;
2447     c->revision = PCI_DEVICE_ID_VMWARE_VMXNET3_REVISION;
2448     c->class_id = PCI_CLASS_NETWORK_ETHERNET;
2449     c->subsystem_vendor_id = PCI_VENDOR_ID_VMWARE;
2450     c->subsystem_id = PCI_DEVICE_ID_VMWARE_VMXNET3;
2451     c->config_write = vmxnet3_write_config,
2452     dc->desc = "VMWare Paravirtualized Ethernet v3";
2453     dc->reset = vmxnet3_qdev_reset;
2454     dc->vmsd = &vmstate_vmxnet3;
2455     dc->props = vmxnet3_properties;
2456 }
2457 
2458 static const TypeInfo vmxnet3_info = {
2459     .name          = TYPE_VMXNET3,
2460     .parent        = TYPE_PCI_DEVICE,
2461     .instance_size = sizeof(VMXNET3State),
2462     .class_init    = vmxnet3_class_init,
2463 };
2464 
2465 static void vmxnet3_register_types(void)
2466 {
2467     VMW_CBPRN("vmxnet3_register_types called...");
2468     type_register_static(&vmxnet3_info);
2469 }
2470 
2471 type_init(vmxnet3_register_types)
2472