xref: /openbmc/qemu/hw/net/vmxnet3.c (revision 37677d7d)
1 /*
2  * QEMU VMWARE VMXNET3 paravirtual NIC
3  *
4  * Copyright (c) 2012 Ravello Systems LTD (http://ravellosystems.com)
5  *
6  * Developed by Daynix Computing LTD (http://www.daynix.com)
7  *
8  * Authors:
9  * Dmitry Fleytman <dmitry@daynix.com>
10  * Tamir Shomer <tamirs@daynix.com>
11  * Yan Vugenfirer <yan@daynix.com>
12  *
13  * This work is licensed under the terms of the GNU GPL, version 2.
14  * See the COPYING file in the top-level directory.
15  *
16  */
17 
18 #include "qemu/osdep.h"
19 #include "hw/hw.h"
20 #include "hw/pci/pci.h"
21 #include "net/tap.h"
22 #include "net/checksum.h"
23 #include "sysemu/sysemu.h"
24 #include "qemu/bswap.h"
25 #include "qemu/module.h"
26 #include "hw/pci/msix.h"
27 #include "hw/pci/msi.h"
28 #include "migration/register.h"
29 
30 #include "vmxnet3.h"
31 #include "vmxnet3_defs.h"
32 #include "vmxnet_debug.h"
33 #include "vmware_utils.h"
34 #include "net_tx_pkt.h"
35 #include "net_rx_pkt.h"
36 
37 #define PCI_DEVICE_ID_VMWARE_VMXNET3_REVISION 0x1
38 #define VMXNET3_MSIX_BAR_SIZE 0x2000
39 #define MIN_BUF_SIZE 60
40 
41 /* Compatibility flags for migration */
42 #define VMXNET3_COMPAT_FLAG_OLD_MSI_OFFSETS_BIT 0
43 #define VMXNET3_COMPAT_FLAG_OLD_MSI_OFFSETS \
44     (1 << VMXNET3_COMPAT_FLAG_OLD_MSI_OFFSETS_BIT)
45 #define VMXNET3_COMPAT_FLAG_DISABLE_PCIE_BIT 1
46 #define VMXNET3_COMPAT_FLAG_DISABLE_PCIE \
47     (1 << VMXNET3_COMPAT_FLAG_DISABLE_PCIE_BIT)
48 
49 #define VMXNET3_EXP_EP_OFFSET (0x48)
50 #define VMXNET3_MSI_OFFSET(s) \
51     ((s)->compat_flags & VMXNET3_COMPAT_FLAG_OLD_MSI_OFFSETS ? 0x50 : 0x84)
52 #define VMXNET3_MSIX_OFFSET(s) \
53     ((s)->compat_flags & VMXNET3_COMPAT_FLAG_OLD_MSI_OFFSETS ? 0 : 0x9c)
54 #define VMXNET3_DSN_OFFSET     (0x100)
55 
56 #define VMXNET3_BAR0_IDX      (0)
57 #define VMXNET3_BAR1_IDX      (1)
58 #define VMXNET3_MSIX_BAR_IDX  (2)
59 
60 #define VMXNET3_OFF_MSIX_TABLE (0x000)
61 #define VMXNET3_OFF_MSIX_PBA(s) \
62     ((s)->compat_flags & VMXNET3_COMPAT_FLAG_OLD_MSI_OFFSETS ? 0x800 : 0x1000)
63 
64 /* Link speed in Mbps should be shifted by 16 */
65 #define VMXNET3_LINK_SPEED      (1000 << 16)
66 
67 /* Link status: 1 - up, 0 - down. */
68 #define VMXNET3_LINK_STATUS_UP  0x1
69 
70 /* Least significant bit should be set for revision and version */
71 #define VMXNET3_UPT_REVISION      0x1
72 #define VMXNET3_DEVICE_REVISION   0x1
73 
74 /* Number of interrupt vectors for non-MSIx modes */
75 #define VMXNET3_MAX_NMSIX_INTRS   (1)
76 
77 /* Macros for rings descriptors access */
78 #define VMXNET3_READ_TX_QUEUE_DESCR8(_d, dpa, field) \
79     (vmw_shmem_ld8(_d, dpa + offsetof(struct Vmxnet3_TxQueueDesc, field)))
80 
81 #define VMXNET3_WRITE_TX_QUEUE_DESCR8(_d, dpa, field, value) \
82     (vmw_shmem_st8(_d, dpa + offsetof(struct Vmxnet3_TxQueueDesc, field, value)))
83 
84 #define VMXNET3_READ_TX_QUEUE_DESCR32(_d, dpa, field) \
85     (vmw_shmem_ld32(_d, dpa + offsetof(struct Vmxnet3_TxQueueDesc, field)))
86 
87 #define VMXNET3_WRITE_TX_QUEUE_DESCR32(_d, dpa, field, value) \
88     (vmw_shmem_st32(_d, dpa + offsetof(struct Vmxnet3_TxQueueDesc, field), value))
89 
90 #define VMXNET3_READ_TX_QUEUE_DESCR64(_d, dpa, field) \
91     (vmw_shmem_ld64(_d, dpa + offsetof(struct Vmxnet3_TxQueueDesc, field)))
92 
93 #define VMXNET3_WRITE_TX_QUEUE_DESCR64(_d, dpa, field, value) \
94     (vmw_shmem_st64(_d, dpa + offsetof(struct Vmxnet3_TxQueueDesc, field), value))
95 
96 #define VMXNET3_READ_RX_QUEUE_DESCR64(_d, dpa, field) \
97     (vmw_shmem_ld64(_d, dpa + offsetof(struct Vmxnet3_RxQueueDesc, field)))
98 
99 #define VMXNET3_READ_RX_QUEUE_DESCR32(_d, dpa, field) \
100     (vmw_shmem_ld32(_d, dpa + offsetof(struct Vmxnet3_RxQueueDesc, field)))
101 
102 #define VMXNET3_WRITE_RX_QUEUE_DESCR64(_d, dpa, field, value) \
103     (vmw_shmem_st64(_d, dpa + offsetof(struct Vmxnet3_RxQueueDesc, field), value))
104 
105 #define VMXNET3_WRITE_RX_QUEUE_DESCR8(_d, dpa, field, value) \
106     (vmw_shmem_st8(_d, dpa + offsetof(struct Vmxnet3_RxQueueDesc, field), value))
107 
108 /* Macros for guest driver shared area access */
109 #define VMXNET3_READ_DRV_SHARED64(_d, shpa, field) \
110     (vmw_shmem_ld64(_d, shpa + offsetof(struct Vmxnet3_DriverShared, field)))
111 
112 #define VMXNET3_READ_DRV_SHARED32(_d, shpa, field) \
113     (vmw_shmem_ld32(_d, shpa + offsetof(struct Vmxnet3_DriverShared, field)))
114 
115 #define VMXNET3_WRITE_DRV_SHARED32(_d, shpa, field, val) \
116     (vmw_shmem_st32(_d, shpa + offsetof(struct Vmxnet3_DriverShared, field), val))
117 
118 #define VMXNET3_READ_DRV_SHARED16(_d, shpa, field) \
119     (vmw_shmem_ld16(_d, shpa + offsetof(struct Vmxnet3_DriverShared, field)))
120 
121 #define VMXNET3_READ_DRV_SHARED8(_d, shpa, field) \
122     (vmw_shmem_ld8(_d, shpa + offsetof(struct Vmxnet3_DriverShared, field)))
123 
124 #define VMXNET3_READ_DRV_SHARED(_d, shpa, field, b, l) \
125     (vmw_shmem_read(_d, shpa + offsetof(struct Vmxnet3_DriverShared, field), b, l))
126 
127 #define VMXNET_FLAG_IS_SET(field, flag) (((field) & (flag)) == (flag))
128 
129 typedef struct VMXNET3Class {
130     PCIDeviceClass parent_class;
131     DeviceRealize parent_dc_realize;
132 } VMXNET3Class;
133 
134 #define VMXNET3_DEVICE_CLASS(klass) \
135     OBJECT_CLASS_CHECK(VMXNET3Class, (klass), TYPE_VMXNET3)
136 #define VMXNET3_DEVICE_GET_CLASS(obj) \
137     OBJECT_GET_CLASS(VMXNET3Class, (obj), TYPE_VMXNET3)
138 
139 static inline void vmxnet3_ring_init(PCIDevice *d,
140                                      Vmxnet3Ring *ring,
141                                      hwaddr pa,
142                                      uint32_t size,
143                                      uint32_t cell_size,
144                                      bool zero_region)
145 {
146     ring->pa = pa;
147     ring->size = size;
148     ring->cell_size = cell_size;
149     ring->gen = VMXNET3_INIT_GEN;
150     ring->next = 0;
151 
152     if (zero_region) {
153         vmw_shmem_set(d, pa, 0, size * cell_size);
154     }
155 }
156 
157 #define VMXNET3_RING_DUMP(macro, ring_name, ridx, r)                         \
158     macro("%s#%d: base %" PRIx64 " size %u cell_size %u gen %d next %u",  \
159           (ring_name), (ridx),                                               \
160           (r)->pa, (r)->size, (r)->cell_size, (r)->gen, (r)->next)
161 
162 static inline void vmxnet3_ring_inc(Vmxnet3Ring *ring)
163 {
164     if (++ring->next >= ring->size) {
165         ring->next = 0;
166         ring->gen ^= 1;
167     }
168 }
169 
170 static inline void vmxnet3_ring_dec(Vmxnet3Ring *ring)
171 {
172     if (ring->next-- == 0) {
173         ring->next = ring->size - 1;
174         ring->gen ^= 1;
175     }
176 }
177 
178 static inline hwaddr vmxnet3_ring_curr_cell_pa(Vmxnet3Ring *ring)
179 {
180     return ring->pa + ring->next * ring->cell_size;
181 }
182 
183 static inline void vmxnet3_ring_read_curr_cell(PCIDevice *d, Vmxnet3Ring *ring,
184                                                void *buff)
185 {
186     vmw_shmem_read(d, vmxnet3_ring_curr_cell_pa(ring), buff, ring->cell_size);
187 }
188 
189 static inline void vmxnet3_ring_write_curr_cell(PCIDevice *d, Vmxnet3Ring *ring,
190                                                 void *buff)
191 {
192     vmw_shmem_write(d, vmxnet3_ring_curr_cell_pa(ring), buff, ring->cell_size);
193 }
194 
195 static inline size_t vmxnet3_ring_curr_cell_idx(Vmxnet3Ring *ring)
196 {
197     return ring->next;
198 }
199 
200 static inline uint8_t vmxnet3_ring_curr_gen(Vmxnet3Ring *ring)
201 {
202     return ring->gen;
203 }
204 
205 /* Debug trace-related functions */
206 static inline void
207 vmxnet3_dump_tx_descr(struct Vmxnet3_TxDesc *descr)
208 {
209     VMW_PKPRN("TX DESCR: "
210               "addr %" PRIx64 ", len: %d, gen: %d, rsvd: %d, "
211               "dtype: %d, ext1: %d, msscof: %d, hlen: %d, om: %d, "
212               "eop: %d, cq: %d, ext2: %d, ti: %d, tci: %d",
213               descr->addr, descr->len, descr->gen, descr->rsvd,
214               descr->dtype, descr->ext1, descr->msscof, descr->hlen, descr->om,
215               descr->eop, descr->cq, descr->ext2, descr->ti, descr->tci);
216 }
217 
218 static inline void
219 vmxnet3_dump_virt_hdr(struct virtio_net_hdr *vhdr)
220 {
221     VMW_PKPRN("VHDR: flags 0x%x, gso_type: 0x%x, hdr_len: %d, gso_size: %d, "
222               "csum_start: %d, csum_offset: %d",
223               vhdr->flags, vhdr->gso_type, vhdr->hdr_len, vhdr->gso_size,
224               vhdr->csum_start, vhdr->csum_offset);
225 }
226 
227 static inline void
228 vmxnet3_dump_rx_descr(struct Vmxnet3_RxDesc *descr)
229 {
230     VMW_PKPRN("RX DESCR: addr %" PRIx64 ", len: %d, gen: %d, rsvd: %d, "
231               "dtype: %d, ext1: %d, btype: %d",
232               descr->addr, descr->len, descr->gen,
233               descr->rsvd, descr->dtype, descr->ext1, descr->btype);
234 }
235 
236 /* Interrupt management */
237 
238 /*
239  * This function returns sign whether interrupt line is in asserted state
240  * This depends on the type of interrupt used. For INTX interrupt line will
241  * be asserted until explicit deassertion, for MSI(X) interrupt line will
242  * be deasserted automatically due to notification semantics of the MSI(X)
243  * interrupts
244  */
245 static bool _vmxnet3_assert_interrupt_line(VMXNET3State *s, uint32_t int_idx)
246 {
247     PCIDevice *d = PCI_DEVICE(s);
248 
249     if (s->msix_used && msix_enabled(d)) {
250         VMW_IRPRN("Sending MSI-X notification for vector %u", int_idx);
251         msix_notify(d, int_idx);
252         return false;
253     }
254     if (msi_enabled(d)) {
255         VMW_IRPRN("Sending MSI notification for vector %u", int_idx);
256         msi_notify(d, int_idx);
257         return false;
258     }
259 
260     VMW_IRPRN("Asserting line for interrupt %u", int_idx);
261     pci_irq_assert(d);
262     return true;
263 }
264 
265 static void _vmxnet3_deassert_interrupt_line(VMXNET3State *s, int lidx)
266 {
267     PCIDevice *d = PCI_DEVICE(s);
268 
269     /*
270      * This function should never be called for MSI(X) interrupts
271      * because deassertion never required for message interrupts
272      */
273     assert(!s->msix_used || !msix_enabled(d));
274     /*
275      * This function should never be called for MSI(X) interrupts
276      * because deassertion never required for message interrupts
277      */
278     assert(!msi_enabled(d));
279 
280     VMW_IRPRN("Deasserting line for interrupt %u", lidx);
281     pci_irq_deassert(d);
282 }
283 
284 static void vmxnet3_update_interrupt_line_state(VMXNET3State *s, int lidx)
285 {
286     if (!s->interrupt_states[lidx].is_pending &&
287        s->interrupt_states[lidx].is_asserted) {
288         VMW_IRPRN("New interrupt line state for index %d is DOWN", lidx);
289         _vmxnet3_deassert_interrupt_line(s, lidx);
290         s->interrupt_states[lidx].is_asserted = false;
291         return;
292     }
293 
294     if (s->interrupt_states[lidx].is_pending &&
295        !s->interrupt_states[lidx].is_masked &&
296        !s->interrupt_states[lidx].is_asserted) {
297         VMW_IRPRN("New interrupt line state for index %d is UP", lidx);
298         s->interrupt_states[lidx].is_asserted =
299             _vmxnet3_assert_interrupt_line(s, lidx);
300         s->interrupt_states[lidx].is_pending = false;
301         return;
302     }
303 }
304 
305 static void vmxnet3_trigger_interrupt(VMXNET3State *s, int lidx)
306 {
307     PCIDevice *d = PCI_DEVICE(s);
308     s->interrupt_states[lidx].is_pending = true;
309     vmxnet3_update_interrupt_line_state(s, lidx);
310 
311     if (s->msix_used && msix_enabled(d) && s->auto_int_masking) {
312         goto do_automask;
313     }
314 
315     if (msi_enabled(d) && s->auto_int_masking) {
316         goto do_automask;
317     }
318 
319     return;
320 
321 do_automask:
322     s->interrupt_states[lidx].is_masked = true;
323     vmxnet3_update_interrupt_line_state(s, lidx);
324 }
325 
326 static bool vmxnet3_interrupt_asserted(VMXNET3State *s, int lidx)
327 {
328     return s->interrupt_states[lidx].is_asserted;
329 }
330 
331 static void vmxnet3_clear_interrupt(VMXNET3State *s, int int_idx)
332 {
333     s->interrupt_states[int_idx].is_pending = false;
334     if (s->auto_int_masking) {
335         s->interrupt_states[int_idx].is_masked = true;
336     }
337     vmxnet3_update_interrupt_line_state(s, int_idx);
338 }
339 
340 static void
341 vmxnet3_on_interrupt_mask_changed(VMXNET3State *s, int lidx, bool is_masked)
342 {
343     s->interrupt_states[lidx].is_masked = is_masked;
344     vmxnet3_update_interrupt_line_state(s, lidx);
345 }
346 
347 static bool vmxnet3_verify_driver_magic(PCIDevice *d, hwaddr dshmem)
348 {
349     return (VMXNET3_READ_DRV_SHARED32(d, dshmem, magic) == VMXNET3_REV1_MAGIC);
350 }
351 
352 #define VMXNET3_GET_BYTE(x, byte_num) (((x) >> (byte_num)*8) & 0xFF)
353 #define VMXNET3_MAKE_BYTE(byte_num, val) \
354     (((uint32_t)((val) & 0xFF)) << (byte_num)*8)
355 
356 static void vmxnet3_set_variable_mac(VMXNET3State *s, uint32_t h, uint32_t l)
357 {
358     s->conf.macaddr.a[0] = VMXNET3_GET_BYTE(l,  0);
359     s->conf.macaddr.a[1] = VMXNET3_GET_BYTE(l,  1);
360     s->conf.macaddr.a[2] = VMXNET3_GET_BYTE(l,  2);
361     s->conf.macaddr.a[3] = VMXNET3_GET_BYTE(l,  3);
362     s->conf.macaddr.a[4] = VMXNET3_GET_BYTE(h, 0);
363     s->conf.macaddr.a[5] = VMXNET3_GET_BYTE(h, 1);
364 
365     VMW_CFPRN("Variable MAC: " MAC_FMT, MAC_ARG(s->conf.macaddr.a));
366 
367     qemu_format_nic_info_str(qemu_get_queue(s->nic), s->conf.macaddr.a);
368 }
369 
370 static uint64_t vmxnet3_get_mac_low(MACAddr *addr)
371 {
372     return VMXNET3_MAKE_BYTE(0, addr->a[0]) |
373            VMXNET3_MAKE_BYTE(1, addr->a[1]) |
374            VMXNET3_MAKE_BYTE(2, addr->a[2]) |
375            VMXNET3_MAKE_BYTE(3, addr->a[3]);
376 }
377 
378 static uint64_t vmxnet3_get_mac_high(MACAddr *addr)
379 {
380     return VMXNET3_MAKE_BYTE(0, addr->a[4]) |
381            VMXNET3_MAKE_BYTE(1, addr->a[5]);
382 }
383 
384 static void
385 vmxnet3_inc_tx_consumption_counter(VMXNET3State *s, int qidx)
386 {
387     vmxnet3_ring_inc(&s->txq_descr[qidx].tx_ring);
388 }
389 
390 static inline void
391 vmxnet3_inc_rx_consumption_counter(VMXNET3State *s, int qidx, int ridx)
392 {
393     vmxnet3_ring_inc(&s->rxq_descr[qidx].rx_ring[ridx]);
394 }
395 
396 static inline void
397 vmxnet3_inc_tx_completion_counter(VMXNET3State *s, int qidx)
398 {
399     vmxnet3_ring_inc(&s->txq_descr[qidx].comp_ring);
400 }
401 
402 static void
403 vmxnet3_inc_rx_completion_counter(VMXNET3State *s, int qidx)
404 {
405     vmxnet3_ring_inc(&s->rxq_descr[qidx].comp_ring);
406 }
407 
408 static void
409 vmxnet3_dec_rx_completion_counter(VMXNET3State *s, int qidx)
410 {
411     vmxnet3_ring_dec(&s->rxq_descr[qidx].comp_ring);
412 }
413 
414 static void vmxnet3_complete_packet(VMXNET3State *s, int qidx, uint32_t tx_ridx)
415 {
416     struct Vmxnet3_TxCompDesc txcq_descr;
417     PCIDevice *d = PCI_DEVICE(s);
418 
419     VMXNET3_RING_DUMP(VMW_RIPRN, "TXC", qidx, &s->txq_descr[qidx].comp_ring);
420 
421     memset(&txcq_descr, 0, sizeof(txcq_descr));
422     txcq_descr.txdIdx = tx_ridx;
423     txcq_descr.gen = vmxnet3_ring_curr_gen(&s->txq_descr[qidx].comp_ring);
424     txcq_descr.val1 = cpu_to_le32(txcq_descr.val1);
425     txcq_descr.val2 = cpu_to_le32(txcq_descr.val2);
426     vmxnet3_ring_write_curr_cell(d, &s->txq_descr[qidx].comp_ring, &txcq_descr);
427 
428     /* Flush changes in TX descriptor before changing the counter value */
429     smp_wmb();
430 
431     vmxnet3_inc_tx_completion_counter(s, qidx);
432     vmxnet3_trigger_interrupt(s, s->txq_descr[qidx].intr_idx);
433 }
434 
435 static bool
436 vmxnet3_setup_tx_offloads(VMXNET3State *s)
437 {
438     switch (s->offload_mode) {
439     case VMXNET3_OM_NONE:
440         net_tx_pkt_build_vheader(s->tx_pkt, false, false, 0);
441         break;
442 
443     case VMXNET3_OM_CSUM:
444         net_tx_pkt_build_vheader(s->tx_pkt, false, true, 0);
445         VMW_PKPRN("L4 CSO requested\n");
446         break;
447 
448     case VMXNET3_OM_TSO:
449         net_tx_pkt_build_vheader(s->tx_pkt, true, true,
450             s->cso_or_gso_size);
451         net_tx_pkt_update_ip_checksums(s->tx_pkt);
452         VMW_PKPRN("GSO offload requested.");
453         break;
454 
455     default:
456         g_assert_not_reached();
457         return false;
458     }
459 
460     return true;
461 }
462 
463 static void
464 vmxnet3_tx_retrieve_metadata(VMXNET3State *s,
465                              const struct Vmxnet3_TxDesc *txd)
466 {
467     s->offload_mode = txd->om;
468     s->cso_or_gso_size = txd->msscof;
469     s->tci = txd->tci;
470     s->needs_vlan = txd->ti;
471 }
472 
473 typedef enum {
474     VMXNET3_PKT_STATUS_OK,
475     VMXNET3_PKT_STATUS_ERROR,
476     VMXNET3_PKT_STATUS_DISCARD,/* only for tx */
477     VMXNET3_PKT_STATUS_OUT_OF_BUF /* only for rx */
478 } Vmxnet3PktStatus;
479 
480 static void
481 vmxnet3_on_tx_done_update_stats(VMXNET3State *s, int qidx,
482     Vmxnet3PktStatus status)
483 {
484     size_t tot_len = net_tx_pkt_get_total_len(s->tx_pkt);
485     struct UPT1_TxStats *stats = &s->txq_descr[qidx].txq_stats;
486 
487     switch (status) {
488     case VMXNET3_PKT_STATUS_OK:
489         switch (net_tx_pkt_get_packet_type(s->tx_pkt)) {
490         case ETH_PKT_BCAST:
491             stats->bcastPktsTxOK++;
492             stats->bcastBytesTxOK += tot_len;
493             break;
494         case ETH_PKT_MCAST:
495             stats->mcastPktsTxOK++;
496             stats->mcastBytesTxOK += tot_len;
497             break;
498         case ETH_PKT_UCAST:
499             stats->ucastPktsTxOK++;
500             stats->ucastBytesTxOK += tot_len;
501             break;
502         default:
503             g_assert_not_reached();
504         }
505 
506         if (s->offload_mode == VMXNET3_OM_TSO) {
507             /*
508              * According to VMWARE headers this statistic is a number
509              * of packets after segmentation but since we don't have
510              * this information in QEMU model, the best we can do is to
511              * provide number of non-segmented packets
512              */
513             stats->TSOPktsTxOK++;
514             stats->TSOBytesTxOK += tot_len;
515         }
516         break;
517 
518     case VMXNET3_PKT_STATUS_DISCARD:
519         stats->pktsTxDiscard++;
520         break;
521 
522     case VMXNET3_PKT_STATUS_ERROR:
523         stats->pktsTxError++;
524         break;
525 
526     default:
527         g_assert_not_reached();
528     }
529 }
530 
531 static void
532 vmxnet3_on_rx_done_update_stats(VMXNET3State *s,
533                                 int qidx,
534                                 Vmxnet3PktStatus status)
535 {
536     struct UPT1_RxStats *stats = &s->rxq_descr[qidx].rxq_stats;
537     size_t tot_len = net_rx_pkt_get_total_len(s->rx_pkt);
538 
539     switch (status) {
540     case VMXNET3_PKT_STATUS_OUT_OF_BUF:
541         stats->pktsRxOutOfBuf++;
542         break;
543 
544     case VMXNET3_PKT_STATUS_ERROR:
545         stats->pktsRxError++;
546         break;
547     case VMXNET3_PKT_STATUS_OK:
548         switch (net_rx_pkt_get_packet_type(s->rx_pkt)) {
549         case ETH_PKT_BCAST:
550             stats->bcastPktsRxOK++;
551             stats->bcastBytesRxOK += tot_len;
552             break;
553         case ETH_PKT_MCAST:
554             stats->mcastPktsRxOK++;
555             stats->mcastBytesRxOK += tot_len;
556             break;
557         case ETH_PKT_UCAST:
558             stats->ucastPktsRxOK++;
559             stats->ucastBytesRxOK += tot_len;
560             break;
561         default:
562             g_assert_not_reached();
563         }
564 
565         if (tot_len > s->mtu) {
566             stats->LROPktsRxOK++;
567             stats->LROBytesRxOK += tot_len;
568         }
569         break;
570     default:
571         g_assert_not_reached();
572     }
573 }
574 
575 static inline void
576 vmxnet3_ring_read_curr_txdesc(PCIDevice *pcidev, Vmxnet3Ring *ring,
577                               struct Vmxnet3_TxDesc *txd)
578 {
579     vmxnet3_ring_read_curr_cell(pcidev, ring, txd);
580     txd->addr = le64_to_cpu(txd->addr);
581     txd->val1 = le32_to_cpu(txd->val1);
582     txd->val2 = le32_to_cpu(txd->val2);
583 }
584 
585 static inline bool
586 vmxnet3_pop_next_tx_descr(VMXNET3State *s,
587                           int qidx,
588                           struct Vmxnet3_TxDesc *txd,
589                           uint32_t *descr_idx)
590 {
591     Vmxnet3Ring *ring = &s->txq_descr[qidx].tx_ring;
592     PCIDevice *d = PCI_DEVICE(s);
593 
594     vmxnet3_ring_read_curr_txdesc(d, ring, txd);
595     if (txd->gen == vmxnet3_ring_curr_gen(ring)) {
596         /* Only read after generation field verification */
597         smp_rmb();
598         /* Re-read to be sure we got the latest version */
599         vmxnet3_ring_read_curr_txdesc(d, ring, txd);
600         VMXNET3_RING_DUMP(VMW_RIPRN, "TX", qidx, ring);
601         *descr_idx = vmxnet3_ring_curr_cell_idx(ring);
602         vmxnet3_inc_tx_consumption_counter(s, qidx);
603         return true;
604     }
605 
606     return false;
607 }
608 
609 static bool
610 vmxnet3_send_packet(VMXNET3State *s, uint32_t qidx)
611 {
612     Vmxnet3PktStatus status = VMXNET3_PKT_STATUS_OK;
613 
614     if (!vmxnet3_setup_tx_offloads(s)) {
615         status = VMXNET3_PKT_STATUS_ERROR;
616         goto func_exit;
617     }
618 
619     /* debug prints */
620     vmxnet3_dump_virt_hdr(net_tx_pkt_get_vhdr(s->tx_pkt));
621     net_tx_pkt_dump(s->tx_pkt);
622 
623     if (!net_tx_pkt_send(s->tx_pkt, qemu_get_queue(s->nic))) {
624         status = VMXNET3_PKT_STATUS_DISCARD;
625         goto func_exit;
626     }
627 
628 func_exit:
629     vmxnet3_on_tx_done_update_stats(s, qidx, status);
630     return (status == VMXNET3_PKT_STATUS_OK);
631 }
632 
633 static void vmxnet3_process_tx_queue(VMXNET3State *s, int qidx)
634 {
635     struct Vmxnet3_TxDesc txd;
636     uint32_t txd_idx;
637     uint32_t data_len;
638     hwaddr data_pa;
639 
640     for (;;) {
641         if (!vmxnet3_pop_next_tx_descr(s, qidx, &txd, &txd_idx)) {
642             break;
643         }
644 
645         vmxnet3_dump_tx_descr(&txd);
646 
647         if (!s->skip_current_tx_pkt) {
648             data_len = (txd.len > 0) ? txd.len : VMXNET3_MAX_TX_BUF_SIZE;
649             data_pa = txd.addr;
650 
651             if (!net_tx_pkt_add_raw_fragment(s->tx_pkt,
652                                                 data_pa,
653                                                 data_len)) {
654                 s->skip_current_tx_pkt = true;
655             }
656         }
657 
658         if (s->tx_sop) {
659             vmxnet3_tx_retrieve_metadata(s, &txd);
660             s->tx_sop = false;
661         }
662 
663         if (txd.eop) {
664             if (!s->skip_current_tx_pkt && net_tx_pkt_parse(s->tx_pkt)) {
665                 if (s->needs_vlan) {
666                     net_tx_pkt_setup_vlan_header(s->tx_pkt, s->tci);
667                 }
668 
669                 vmxnet3_send_packet(s, qidx);
670             } else {
671                 vmxnet3_on_tx_done_update_stats(s, qidx,
672                                                 VMXNET3_PKT_STATUS_ERROR);
673             }
674 
675             vmxnet3_complete_packet(s, qidx, txd_idx);
676             s->tx_sop = true;
677             s->skip_current_tx_pkt = false;
678             net_tx_pkt_reset(s->tx_pkt);
679         }
680     }
681 }
682 
683 static inline void
684 vmxnet3_read_next_rx_descr(VMXNET3State *s, int qidx, int ridx,
685                            struct Vmxnet3_RxDesc *dbuf, uint32_t *didx)
686 {
687     PCIDevice *d = PCI_DEVICE(s);
688 
689     Vmxnet3Ring *ring = &s->rxq_descr[qidx].rx_ring[ridx];
690     *didx = vmxnet3_ring_curr_cell_idx(ring);
691     vmxnet3_ring_read_curr_cell(d, ring, dbuf);
692     dbuf->addr = le64_to_cpu(dbuf->addr);
693     dbuf->val1 = le32_to_cpu(dbuf->val1);
694     dbuf->ext1 = le32_to_cpu(dbuf->ext1);
695 }
696 
697 static inline uint8_t
698 vmxnet3_get_rx_ring_gen(VMXNET3State *s, int qidx, int ridx)
699 {
700     return s->rxq_descr[qidx].rx_ring[ridx].gen;
701 }
702 
703 static inline hwaddr
704 vmxnet3_pop_rxc_descr(VMXNET3State *s, int qidx, uint32_t *descr_gen)
705 {
706     uint8_t ring_gen;
707     struct Vmxnet3_RxCompDesc rxcd;
708 
709     hwaddr daddr =
710         vmxnet3_ring_curr_cell_pa(&s->rxq_descr[qidx].comp_ring);
711 
712     pci_dma_read(PCI_DEVICE(s),
713                  daddr, &rxcd, sizeof(struct Vmxnet3_RxCompDesc));
714     rxcd.val1 = le32_to_cpu(rxcd.val1);
715     rxcd.val2 = le32_to_cpu(rxcd.val2);
716     rxcd.val3 = le32_to_cpu(rxcd.val3);
717     ring_gen = vmxnet3_ring_curr_gen(&s->rxq_descr[qidx].comp_ring);
718 
719     if (rxcd.gen != ring_gen) {
720         *descr_gen = ring_gen;
721         vmxnet3_inc_rx_completion_counter(s, qidx);
722         return daddr;
723     }
724 
725     return 0;
726 }
727 
728 static inline void
729 vmxnet3_revert_rxc_descr(VMXNET3State *s, int qidx)
730 {
731     vmxnet3_dec_rx_completion_counter(s, qidx);
732 }
733 
734 #define RXQ_IDX      (0)
735 #define RX_HEAD_BODY_RING (0)
736 #define RX_BODY_ONLY_RING (1)
737 
738 static bool
739 vmxnet3_get_next_head_rx_descr(VMXNET3State *s,
740                                struct Vmxnet3_RxDesc *descr_buf,
741                                uint32_t *descr_idx,
742                                uint32_t *ridx)
743 {
744     for (;;) {
745         uint32_t ring_gen;
746         vmxnet3_read_next_rx_descr(s, RXQ_IDX, RX_HEAD_BODY_RING,
747                                    descr_buf, descr_idx);
748 
749         /* If no more free descriptors - return */
750         ring_gen = vmxnet3_get_rx_ring_gen(s, RXQ_IDX, RX_HEAD_BODY_RING);
751         if (descr_buf->gen != ring_gen) {
752             return false;
753         }
754 
755         /* Only read after generation field verification */
756         smp_rmb();
757         /* Re-read to be sure we got the latest version */
758         vmxnet3_read_next_rx_descr(s, RXQ_IDX, RX_HEAD_BODY_RING,
759                                    descr_buf, descr_idx);
760 
761         /* Mark current descriptor as used/skipped */
762         vmxnet3_inc_rx_consumption_counter(s, RXQ_IDX, RX_HEAD_BODY_RING);
763 
764         /* If this is what we are looking for - return */
765         if (descr_buf->btype == VMXNET3_RXD_BTYPE_HEAD) {
766             *ridx = RX_HEAD_BODY_RING;
767             return true;
768         }
769     }
770 }
771 
772 static bool
773 vmxnet3_get_next_body_rx_descr(VMXNET3State *s,
774                                struct Vmxnet3_RxDesc *d,
775                                uint32_t *didx,
776                                uint32_t *ridx)
777 {
778     vmxnet3_read_next_rx_descr(s, RXQ_IDX, RX_HEAD_BODY_RING, d, didx);
779 
780     /* Try to find corresponding descriptor in head/body ring */
781     if (d->gen == vmxnet3_get_rx_ring_gen(s, RXQ_IDX, RX_HEAD_BODY_RING)) {
782         /* Only read after generation field verification */
783         smp_rmb();
784         /* Re-read to be sure we got the latest version */
785         vmxnet3_read_next_rx_descr(s, RXQ_IDX, RX_HEAD_BODY_RING, d, didx);
786         if (d->btype == VMXNET3_RXD_BTYPE_BODY) {
787             vmxnet3_inc_rx_consumption_counter(s, RXQ_IDX, RX_HEAD_BODY_RING);
788             *ridx = RX_HEAD_BODY_RING;
789             return true;
790         }
791     }
792 
793     /*
794      * If there is no free descriptors on head/body ring or next free
795      * descriptor is a head descriptor switch to body only ring
796      */
797     vmxnet3_read_next_rx_descr(s, RXQ_IDX, RX_BODY_ONLY_RING, d, didx);
798 
799     /* If no more free descriptors - return */
800     if (d->gen == vmxnet3_get_rx_ring_gen(s, RXQ_IDX, RX_BODY_ONLY_RING)) {
801         /* Only read after generation field verification */
802         smp_rmb();
803         /* Re-read to be sure we got the latest version */
804         vmxnet3_read_next_rx_descr(s, RXQ_IDX, RX_BODY_ONLY_RING, d, didx);
805         assert(d->btype == VMXNET3_RXD_BTYPE_BODY);
806         *ridx = RX_BODY_ONLY_RING;
807         vmxnet3_inc_rx_consumption_counter(s, RXQ_IDX, RX_BODY_ONLY_RING);
808         return true;
809     }
810 
811     return false;
812 }
813 
814 static inline bool
815 vmxnet3_get_next_rx_descr(VMXNET3State *s, bool is_head,
816                           struct Vmxnet3_RxDesc *descr_buf,
817                           uint32_t *descr_idx,
818                           uint32_t *ridx)
819 {
820     if (is_head || !s->rx_packets_compound) {
821         return vmxnet3_get_next_head_rx_descr(s, descr_buf, descr_idx, ridx);
822     } else {
823         return vmxnet3_get_next_body_rx_descr(s, descr_buf, descr_idx, ridx);
824     }
825 }
826 
827 /* In case packet was csum offloaded (either NEEDS_CSUM or DATA_VALID),
828  * the implementation always passes an RxCompDesc with a "Checksum
829  * calculated and found correct" to the OS (cnc=0 and tuc=1, see
830  * vmxnet3_rx_update_descr). This emulates the observed ESXi behavior.
831  *
832  * Therefore, if packet has the NEEDS_CSUM set, we must calculate
833  * and place a fully computed checksum into the tcp/udp header.
834  * Otherwise, the OS driver will receive a checksum-correct indication
835  * (CHECKSUM_UNNECESSARY), but with the actual tcp/udp checksum field
836  * having just the pseudo header csum value.
837  *
838  * While this is not a problem if packet is destined for local delivery,
839  * in the case the host OS performs forwarding, it will forward an
840  * incorrectly checksummed packet.
841  */
842 static void vmxnet3_rx_need_csum_calculate(struct NetRxPkt *pkt,
843                                            const void *pkt_data,
844                                            size_t pkt_len)
845 {
846     struct virtio_net_hdr *vhdr;
847     bool isip4, isip6, istcp, isudp;
848     uint8_t *data;
849     int len;
850 
851     if (!net_rx_pkt_has_virt_hdr(pkt)) {
852         return;
853     }
854 
855     vhdr = net_rx_pkt_get_vhdr(pkt);
856     if (!VMXNET_FLAG_IS_SET(vhdr->flags, VIRTIO_NET_HDR_F_NEEDS_CSUM)) {
857         return;
858     }
859 
860     net_rx_pkt_get_protocols(pkt, &isip4, &isip6, &isudp, &istcp);
861     if (!(isip4 || isip6) || !(istcp || isudp)) {
862         return;
863     }
864 
865     vmxnet3_dump_virt_hdr(vhdr);
866 
867     /* Validate packet len: csum_start + scum_offset + length of csum field */
868     if (pkt_len < (vhdr->csum_start + vhdr->csum_offset + 2)) {
869         VMW_PKPRN("packet len:%zu < csum_start(%d) + csum_offset(%d) + 2, "
870                   "cannot calculate checksum",
871                   pkt_len, vhdr->csum_start, vhdr->csum_offset);
872         return;
873     }
874 
875     data = (uint8_t *)pkt_data + vhdr->csum_start;
876     len = pkt_len - vhdr->csum_start;
877     /* Put the checksum obtained into the packet */
878     stw_be_p(data + vhdr->csum_offset,
879              net_checksum_finish_nozero(net_checksum_add(len, data)));
880 
881     vhdr->flags &= ~VIRTIO_NET_HDR_F_NEEDS_CSUM;
882     vhdr->flags |= VIRTIO_NET_HDR_F_DATA_VALID;
883 }
884 
885 static void vmxnet3_rx_update_descr(struct NetRxPkt *pkt,
886     struct Vmxnet3_RxCompDesc *rxcd)
887 {
888     int csum_ok, is_gso;
889     bool isip4, isip6, istcp, isudp;
890     struct virtio_net_hdr *vhdr;
891     uint8_t offload_type;
892 
893     if (net_rx_pkt_is_vlan_stripped(pkt)) {
894         rxcd->ts = 1;
895         rxcd->tci = net_rx_pkt_get_vlan_tag(pkt);
896     }
897 
898     if (!net_rx_pkt_has_virt_hdr(pkt)) {
899         goto nocsum;
900     }
901 
902     vhdr = net_rx_pkt_get_vhdr(pkt);
903     /*
904      * Checksum is valid when lower level tell so or when lower level
905      * requires checksum offload telling that packet produced/bridged
906      * locally and did travel over network after last checksum calculation
907      * or production
908      */
909     csum_ok = VMXNET_FLAG_IS_SET(vhdr->flags, VIRTIO_NET_HDR_F_DATA_VALID) ||
910               VMXNET_FLAG_IS_SET(vhdr->flags, VIRTIO_NET_HDR_F_NEEDS_CSUM);
911 
912     offload_type = vhdr->gso_type & ~VIRTIO_NET_HDR_GSO_ECN;
913     is_gso = (offload_type != VIRTIO_NET_HDR_GSO_NONE) ? 1 : 0;
914 
915     if (!csum_ok && !is_gso) {
916         goto nocsum;
917     }
918 
919     net_rx_pkt_get_protocols(pkt, &isip4, &isip6, &isudp, &istcp);
920     if ((!istcp && !isudp) || (!isip4 && !isip6)) {
921         goto nocsum;
922     }
923 
924     rxcd->cnc = 0;
925     rxcd->v4 = isip4 ? 1 : 0;
926     rxcd->v6 = isip6 ? 1 : 0;
927     rxcd->tcp = istcp ? 1 : 0;
928     rxcd->udp = isudp ? 1 : 0;
929     rxcd->fcs = rxcd->tuc = rxcd->ipc = 1;
930     return;
931 
932 nocsum:
933     rxcd->cnc = 1;
934     return;
935 }
936 
937 static void
938 vmxnet3_pci_dma_writev(PCIDevice *pci_dev,
939                        const struct iovec *iov,
940                        size_t start_iov_off,
941                        hwaddr target_addr,
942                        size_t bytes_to_copy)
943 {
944     size_t curr_off = 0;
945     size_t copied = 0;
946 
947     while (bytes_to_copy) {
948         if (start_iov_off < (curr_off + iov->iov_len)) {
949             size_t chunk_len =
950                 MIN((curr_off + iov->iov_len) - start_iov_off, bytes_to_copy);
951 
952             pci_dma_write(pci_dev, target_addr + copied,
953                           iov->iov_base + start_iov_off - curr_off,
954                           chunk_len);
955 
956             copied += chunk_len;
957             start_iov_off += chunk_len;
958             curr_off = start_iov_off;
959             bytes_to_copy -= chunk_len;
960         } else {
961             curr_off += iov->iov_len;
962         }
963         iov++;
964     }
965 }
966 
967 static void
968 vmxnet3_pci_dma_write_rxcd(PCIDevice *pcidev, dma_addr_t pa,
969                            struct Vmxnet3_RxCompDesc *rxcd)
970 {
971     rxcd->val1 = cpu_to_le32(rxcd->val1);
972     rxcd->val2 = cpu_to_le32(rxcd->val2);
973     rxcd->val3 = cpu_to_le32(rxcd->val3);
974     pci_dma_write(pcidev, pa, rxcd, sizeof(*rxcd));
975 }
976 
977 static bool
978 vmxnet3_indicate_packet(VMXNET3State *s)
979 {
980     struct Vmxnet3_RxDesc rxd;
981     PCIDevice *d = PCI_DEVICE(s);
982     bool is_head = true;
983     uint32_t rxd_idx;
984     uint32_t rx_ridx = 0;
985 
986     struct Vmxnet3_RxCompDesc rxcd;
987     uint32_t new_rxcd_gen = VMXNET3_INIT_GEN;
988     hwaddr new_rxcd_pa = 0;
989     hwaddr ready_rxcd_pa = 0;
990     struct iovec *data = net_rx_pkt_get_iovec(s->rx_pkt);
991     size_t bytes_copied = 0;
992     size_t bytes_left = net_rx_pkt_get_total_len(s->rx_pkt);
993     uint16_t num_frags = 0;
994     size_t chunk_size;
995 
996     net_rx_pkt_dump(s->rx_pkt);
997 
998     while (bytes_left > 0) {
999 
1000         /* cannot add more frags to packet */
1001         if (num_frags == s->max_rx_frags) {
1002             break;
1003         }
1004 
1005         new_rxcd_pa = vmxnet3_pop_rxc_descr(s, RXQ_IDX, &new_rxcd_gen);
1006         if (!new_rxcd_pa) {
1007             break;
1008         }
1009 
1010         if (!vmxnet3_get_next_rx_descr(s, is_head, &rxd, &rxd_idx, &rx_ridx)) {
1011             break;
1012         }
1013 
1014         chunk_size = MIN(bytes_left, rxd.len);
1015         vmxnet3_pci_dma_writev(d, data, bytes_copied, rxd.addr, chunk_size);
1016         bytes_copied += chunk_size;
1017         bytes_left -= chunk_size;
1018 
1019         vmxnet3_dump_rx_descr(&rxd);
1020 
1021         if (ready_rxcd_pa != 0) {
1022             vmxnet3_pci_dma_write_rxcd(d, ready_rxcd_pa, &rxcd);
1023         }
1024 
1025         memset(&rxcd, 0, sizeof(struct Vmxnet3_RxCompDesc));
1026         rxcd.rxdIdx = rxd_idx;
1027         rxcd.len = chunk_size;
1028         rxcd.sop = is_head;
1029         rxcd.gen = new_rxcd_gen;
1030         rxcd.rqID = RXQ_IDX + rx_ridx * s->rxq_num;
1031 
1032         if (bytes_left == 0) {
1033             vmxnet3_rx_update_descr(s->rx_pkt, &rxcd);
1034         }
1035 
1036         VMW_RIPRN("RX Completion descriptor: rxRing: %lu rxIdx %lu len %lu "
1037                   "sop %d csum_correct %lu",
1038                   (unsigned long) rx_ridx,
1039                   (unsigned long) rxcd.rxdIdx,
1040                   (unsigned long) rxcd.len,
1041                   (int) rxcd.sop,
1042                   (unsigned long) rxcd.tuc);
1043 
1044         is_head = false;
1045         ready_rxcd_pa = new_rxcd_pa;
1046         new_rxcd_pa = 0;
1047         num_frags++;
1048     }
1049 
1050     if (ready_rxcd_pa != 0) {
1051         rxcd.eop = 1;
1052         rxcd.err = (bytes_left != 0);
1053 
1054         vmxnet3_pci_dma_write_rxcd(d, ready_rxcd_pa, &rxcd);
1055 
1056         /* Flush RX descriptor changes */
1057         smp_wmb();
1058     }
1059 
1060     if (new_rxcd_pa != 0) {
1061         vmxnet3_revert_rxc_descr(s, RXQ_IDX);
1062     }
1063 
1064     vmxnet3_trigger_interrupt(s, s->rxq_descr[RXQ_IDX].intr_idx);
1065 
1066     if (bytes_left == 0) {
1067         vmxnet3_on_rx_done_update_stats(s, RXQ_IDX, VMXNET3_PKT_STATUS_OK);
1068         return true;
1069     } else if (num_frags == s->max_rx_frags) {
1070         vmxnet3_on_rx_done_update_stats(s, RXQ_IDX, VMXNET3_PKT_STATUS_ERROR);
1071         return false;
1072     } else {
1073         vmxnet3_on_rx_done_update_stats(s, RXQ_IDX,
1074                                         VMXNET3_PKT_STATUS_OUT_OF_BUF);
1075         return false;
1076     }
1077 }
1078 
1079 static void
1080 vmxnet3_io_bar0_write(void *opaque, hwaddr addr,
1081                       uint64_t val, unsigned size)
1082 {
1083     VMXNET3State *s = opaque;
1084 
1085     if (!s->device_active) {
1086         return;
1087     }
1088 
1089     if (VMW_IS_MULTIREG_ADDR(addr, VMXNET3_REG_TXPROD,
1090                         VMXNET3_DEVICE_MAX_TX_QUEUES, VMXNET3_REG_ALIGN)) {
1091         int tx_queue_idx =
1092             VMW_MULTIREG_IDX_BY_ADDR(addr, VMXNET3_REG_TXPROD,
1093                                      VMXNET3_REG_ALIGN);
1094         assert(tx_queue_idx <= s->txq_num);
1095         vmxnet3_process_tx_queue(s, tx_queue_idx);
1096         return;
1097     }
1098 
1099     if (VMW_IS_MULTIREG_ADDR(addr, VMXNET3_REG_IMR,
1100                         VMXNET3_MAX_INTRS, VMXNET3_REG_ALIGN)) {
1101         int l = VMW_MULTIREG_IDX_BY_ADDR(addr, VMXNET3_REG_IMR,
1102                                          VMXNET3_REG_ALIGN);
1103 
1104         VMW_CBPRN("Interrupt mask for line %d written: 0x%" PRIx64, l, val);
1105 
1106         vmxnet3_on_interrupt_mask_changed(s, l, val);
1107         return;
1108     }
1109 
1110     if (VMW_IS_MULTIREG_ADDR(addr, VMXNET3_REG_RXPROD,
1111                         VMXNET3_DEVICE_MAX_RX_QUEUES, VMXNET3_REG_ALIGN) ||
1112        VMW_IS_MULTIREG_ADDR(addr, VMXNET3_REG_RXPROD2,
1113                         VMXNET3_DEVICE_MAX_RX_QUEUES, VMXNET3_REG_ALIGN)) {
1114         return;
1115     }
1116 
1117     VMW_WRPRN("BAR0 unknown write [%" PRIx64 "] = %" PRIx64 ", size %d",
1118               (uint64_t) addr, val, size);
1119 }
1120 
1121 static uint64_t
1122 vmxnet3_io_bar0_read(void *opaque, hwaddr addr, unsigned size)
1123 {
1124     VMXNET3State *s = opaque;
1125 
1126     if (VMW_IS_MULTIREG_ADDR(addr, VMXNET3_REG_IMR,
1127                         VMXNET3_MAX_INTRS, VMXNET3_REG_ALIGN)) {
1128         int l = VMW_MULTIREG_IDX_BY_ADDR(addr, VMXNET3_REG_IMR,
1129                                          VMXNET3_REG_ALIGN);
1130         return s->interrupt_states[l].is_masked;
1131     }
1132 
1133     VMW_CBPRN("BAR0 unknown read [%" PRIx64 "], size %d", addr, size);
1134     return 0;
1135 }
1136 
1137 static void vmxnet3_reset_interrupt_states(VMXNET3State *s)
1138 {
1139     int i;
1140     for (i = 0; i < ARRAY_SIZE(s->interrupt_states); i++) {
1141         s->interrupt_states[i].is_asserted = false;
1142         s->interrupt_states[i].is_pending = false;
1143         s->interrupt_states[i].is_masked = true;
1144     }
1145 }
1146 
1147 static void vmxnet3_reset_mac(VMXNET3State *s)
1148 {
1149     memcpy(&s->conf.macaddr.a, &s->perm_mac.a, sizeof(s->perm_mac.a));
1150     VMW_CFPRN("MAC address set to: " MAC_FMT, MAC_ARG(s->conf.macaddr.a));
1151 }
1152 
1153 static void vmxnet3_deactivate_device(VMXNET3State *s)
1154 {
1155     if (s->device_active) {
1156         VMW_CBPRN("Deactivating vmxnet3...");
1157         net_tx_pkt_reset(s->tx_pkt);
1158         net_tx_pkt_uninit(s->tx_pkt);
1159         net_rx_pkt_uninit(s->rx_pkt);
1160         s->device_active = false;
1161     }
1162 }
1163 
1164 static void vmxnet3_reset(VMXNET3State *s)
1165 {
1166     VMW_CBPRN("Resetting vmxnet3...");
1167 
1168     vmxnet3_deactivate_device(s);
1169     vmxnet3_reset_interrupt_states(s);
1170     s->drv_shmem = 0;
1171     s->tx_sop = true;
1172     s->skip_current_tx_pkt = false;
1173 }
1174 
1175 static void vmxnet3_update_rx_mode(VMXNET3State *s)
1176 {
1177     PCIDevice *d = PCI_DEVICE(s);
1178 
1179     s->rx_mode = VMXNET3_READ_DRV_SHARED32(d, s->drv_shmem,
1180                                            devRead.rxFilterConf.rxMode);
1181     VMW_CFPRN("RX mode: 0x%08X", s->rx_mode);
1182 }
1183 
1184 static void vmxnet3_update_vlan_filters(VMXNET3State *s)
1185 {
1186     int i;
1187     PCIDevice *d = PCI_DEVICE(s);
1188 
1189     /* Copy configuration from shared memory */
1190     VMXNET3_READ_DRV_SHARED(d, s->drv_shmem,
1191                             devRead.rxFilterConf.vfTable,
1192                             s->vlan_table,
1193                             sizeof(s->vlan_table));
1194 
1195     /* Invert byte order when needed */
1196     for (i = 0; i < ARRAY_SIZE(s->vlan_table); i++) {
1197         s->vlan_table[i] = le32_to_cpu(s->vlan_table[i]);
1198     }
1199 
1200     /* Dump configuration for debugging purposes */
1201     VMW_CFPRN("Configured VLANs:");
1202     for (i = 0; i < sizeof(s->vlan_table) * 8; i++) {
1203         if (VMXNET3_VFTABLE_ENTRY_IS_SET(s->vlan_table, i)) {
1204             VMW_CFPRN("\tVLAN %d is present", i);
1205         }
1206     }
1207 }
1208 
1209 static void vmxnet3_update_mcast_filters(VMXNET3State *s)
1210 {
1211     PCIDevice *d = PCI_DEVICE(s);
1212 
1213     uint16_t list_bytes =
1214         VMXNET3_READ_DRV_SHARED16(d, s->drv_shmem,
1215                                   devRead.rxFilterConf.mfTableLen);
1216 
1217     s->mcast_list_len = list_bytes / sizeof(s->mcast_list[0]);
1218 
1219     s->mcast_list = g_realloc(s->mcast_list, list_bytes);
1220     if (!s->mcast_list) {
1221         if (s->mcast_list_len == 0) {
1222             VMW_CFPRN("Current multicast list is empty");
1223         } else {
1224             VMW_ERPRN("Failed to allocate multicast list of %d elements",
1225                       s->mcast_list_len);
1226         }
1227         s->mcast_list_len = 0;
1228     } else {
1229         int i;
1230         hwaddr mcast_list_pa =
1231             VMXNET3_READ_DRV_SHARED64(d, s->drv_shmem,
1232                                       devRead.rxFilterConf.mfTablePA);
1233 
1234         pci_dma_read(d, mcast_list_pa, s->mcast_list, list_bytes);
1235 
1236         VMW_CFPRN("Current multicast list len is %d:", s->mcast_list_len);
1237         for (i = 0; i < s->mcast_list_len; i++) {
1238             VMW_CFPRN("\t" MAC_FMT, MAC_ARG(s->mcast_list[i].a));
1239         }
1240     }
1241 }
1242 
1243 static void vmxnet3_setup_rx_filtering(VMXNET3State *s)
1244 {
1245     vmxnet3_update_rx_mode(s);
1246     vmxnet3_update_vlan_filters(s);
1247     vmxnet3_update_mcast_filters(s);
1248 }
1249 
1250 static uint32_t vmxnet3_get_interrupt_config(VMXNET3State *s)
1251 {
1252     uint32_t interrupt_mode = VMXNET3_IT_AUTO | (VMXNET3_IMM_AUTO << 2);
1253     VMW_CFPRN("Interrupt config is 0x%X", interrupt_mode);
1254     return interrupt_mode;
1255 }
1256 
1257 static void vmxnet3_fill_stats(VMXNET3State *s)
1258 {
1259     int i;
1260     PCIDevice *d = PCI_DEVICE(s);
1261 
1262     if (!s->device_active)
1263         return;
1264 
1265     for (i = 0; i < s->txq_num; i++) {
1266         pci_dma_write(d,
1267                       s->txq_descr[i].tx_stats_pa,
1268                       &s->txq_descr[i].txq_stats,
1269                       sizeof(s->txq_descr[i].txq_stats));
1270     }
1271 
1272     for (i = 0; i < s->rxq_num; i++) {
1273         pci_dma_write(d,
1274                       s->rxq_descr[i].rx_stats_pa,
1275                       &s->rxq_descr[i].rxq_stats,
1276                       sizeof(s->rxq_descr[i].rxq_stats));
1277     }
1278 }
1279 
1280 static void vmxnet3_adjust_by_guest_type(VMXNET3State *s)
1281 {
1282     struct Vmxnet3_GOSInfo gos;
1283     PCIDevice *d = PCI_DEVICE(s);
1284 
1285     VMXNET3_READ_DRV_SHARED(d, s->drv_shmem, devRead.misc.driverInfo.gos,
1286                             &gos, sizeof(gos));
1287     s->rx_packets_compound =
1288         (gos.gosType == VMXNET3_GOS_TYPE_WIN) ? false : true;
1289 
1290     VMW_CFPRN("Guest type specifics: RXCOMPOUND: %d", s->rx_packets_compound);
1291 }
1292 
1293 static void
1294 vmxnet3_dump_conf_descr(const char *name,
1295                         struct Vmxnet3_VariableLenConfDesc *pm_descr)
1296 {
1297     VMW_CFPRN("%s descriptor dump: Version %u, Length %u",
1298               name, pm_descr->confVer, pm_descr->confLen);
1299 
1300 };
1301 
1302 static void vmxnet3_update_pm_state(VMXNET3State *s)
1303 {
1304     struct Vmxnet3_VariableLenConfDesc pm_descr;
1305     PCIDevice *d = PCI_DEVICE(s);
1306 
1307     pm_descr.confLen =
1308         VMXNET3_READ_DRV_SHARED32(d, s->drv_shmem, devRead.pmConfDesc.confLen);
1309     pm_descr.confVer =
1310         VMXNET3_READ_DRV_SHARED32(d, s->drv_shmem, devRead.pmConfDesc.confVer);
1311     pm_descr.confPA =
1312         VMXNET3_READ_DRV_SHARED64(d, s->drv_shmem, devRead.pmConfDesc.confPA);
1313 
1314     vmxnet3_dump_conf_descr("PM State", &pm_descr);
1315 }
1316 
1317 static void vmxnet3_update_features(VMXNET3State *s)
1318 {
1319     uint32_t guest_features;
1320     int rxcso_supported;
1321     PCIDevice *d = PCI_DEVICE(s);
1322 
1323     guest_features = VMXNET3_READ_DRV_SHARED32(d, s->drv_shmem,
1324                                                devRead.misc.uptFeatures);
1325 
1326     rxcso_supported = VMXNET_FLAG_IS_SET(guest_features, UPT1_F_RXCSUM);
1327     s->rx_vlan_stripping = VMXNET_FLAG_IS_SET(guest_features, UPT1_F_RXVLAN);
1328     s->lro_supported = VMXNET_FLAG_IS_SET(guest_features, UPT1_F_LRO);
1329 
1330     VMW_CFPRN("Features configuration: LRO: %d, RXCSUM: %d, VLANSTRIP: %d",
1331               s->lro_supported, rxcso_supported,
1332               s->rx_vlan_stripping);
1333     if (s->peer_has_vhdr) {
1334         qemu_set_offload(qemu_get_queue(s->nic)->peer,
1335                          rxcso_supported,
1336                          s->lro_supported,
1337                          s->lro_supported,
1338                          0,
1339                          0);
1340     }
1341 }
1342 
1343 static bool vmxnet3_verify_intx(VMXNET3State *s, int intx)
1344 {
1345     return s->msix_used || msi_enabled(PCI_DEVICE(s))
1346         || intx == pci_get_byte(s->parent_obj.config + PCI_INTERRUPT_PIN) - 1;
1347 }
1348 
1349 static void vmxnet3_validate_interrupt_idx(bool is_msix, int idx)
1350 {
1351     int max_ints = is_msix ? VMXNET3_MAX_INTRS : VMXNET3_MAX_NMSIX_INTRS;
1352     if (idx >= max_ints) {
1353         hw_error("Bad interrupt index: %d\n", idx);
1354     }
1355 }
1356 
1357 static void vmxnet3_validate_interrupts(VMXNET3State *s)
1358 {
1359     int i;
1360 
1361     VMW_CFPRN("Verifying event interrupt index (%d)", s->event_int_idx);
1362     vmxnet3_validate_interrupt_idx(s->msix_used, s->event_int_idx);
1363 
1364     for (i = 0; i < s->txq_num; i++) {
1365         int idx = s->txq_descr[i].intr_idx;
1366         VMW_CFPRN("Verifying TX queue %d interrupt index (%d)", i, idx);
1367         vmxnet3_validate_interrupt_idx(s->msix_used, idx);
1368     }
1369 
1370     for (i = 0; i < s->rxq_num; i++) {
1371         int idx = s->rxq_descr[i].intr_idx;
1372         VMW_CFPRN("Verifying RX queue %d interrupt index (%d)", i, idx);
1373         vmxnet3_validate_interrupt_idx(s->msix_used, idx);
1374     }
1375 }
1376 
1377 static void vmxnet3_validate_queues(VMXNET3State *s)
1378 {
1379     /*
1380     * txq_num and rxq_num are total number of queues
1381     * configured by guest. These numbers must not
1382     * exceed corresponding maximal values.
1383     */
1384 
1385     if (s->txq_num > VMXNET3_DEVICE_MAX_TX_QUEUES) {
1386         hw_error("Bad TX queues number: %d\n", s->txq_num);
1387     }
1388 
1389     if (s->rxq_num > VMXNET3_DEVICE_MAX_RX_QUEUES) {
1390         hw_error("Bad RX queues number: %d\n", s->rxq_num);
1391     }
1392 }
1393 
1394 static void vmxnet3_activate_device(VMXNET3State *s)
1395 {
1396     int i;
1397     static const uint32_t VMXNET3_DEF_TX_THRESHOLD = 1;
1398     PCIDevice *d = PCI_DEVICE(s);
1399     hwaddr qdescr_table_pa;
1400     uint64_t pa;
1401     uint32_t size;
1402 
1403     /* Verify configuration consistency */
1404     if (!vmxnet3_verify_driver_magic(d, s->drv_shmem)) {
1405         VMW_ERPRN("Device configuration received from driver is invalid");
1406         return;
1407     }
1408 
1409     /* Verify if device is active */
1410     if (s->device_active) {
1411         VMW_CFPRN("Vmxnet3 device is active");
1412         return;
1413     }
1414 
1415     vmxnet3_adjust_by_guest_type(s);
1416     vmxnet3_update_features(s);
1417     vmxnet3_update_pm_state(s);
1418     vmxnet3_setup_rx_filtering(s);
1419     /* Cache fields from shared memory */
1420     s->mtu = VMXNET3_READ_DRV_SHARED32(d, s->drv_shmem, devRead.misc.mtu);
1421     VMW_CFPRN("MTU is %u", s->mtu);
1422 
1423     s->max_rx_frags =
1424         VMXNET3_READ_DRV_SHARED16(d, s->drv_shmem, devRead.misc.maxNumRxSG);
1425 
1426     if (s->max_rx_frags == 0) {
1427         s->max_rx_frags = 1;
1428     }
1429 
1430     VMW_CFPRN("Max RX fragments is %u", s->max_rx_frags);
1431 
1432     s->event_int_idx =
1433         VMXNET3_READ_DRV_SHARED8(d, s->drv_shmem, devRead.intrConf.eventIntrIdx);
1434     assert(vmxnet3_verify_intx(s, s->event_int_idx));
1435     VMW_CFPRN("Events interrupt line is %u", s->event_int_idx);
1436 
1437     s->auto_int_masking =
1438         VMXNET3_READ_DRV_SHARED8(d, s->drv_shmem, devRead.intrConf.autoMask);
1439     VMW_CFPRN("Automatic interrupt masking is %d", (int)s->auto_int_masking);
1440 
1441     s->txq_num =
1442         VMXNET3_READ_DRV_SHARED8(d, s->drv_shmem, devRead.misc.numTxQueues);
1443     s->rxq_num =
1444         VMXNET3_READ_DRV_SHARED8(d, s->drv_shmem, devRead.misc.numRxQueues);
1445 
1446     VMW_CFPRN("Number of TX/RX queues %u/%u", s->txq_num, s->rxq_num);
1447     vmxnet3_validate_queues(s);
1448 
1449     qdescr_table_pa =
1450         VMXNET3_READ_DRV_SHARED64(d, s->drv_shmem, devRead.misc.queueDescPA);
1451     VMW_CFPRN("TX queues descriptors table is at 0x%" PRIx64, qdescr_table_pa);
1452 
1453     /*
1454      * Worst-case scenario is a packet that holds all TX rings space so
1455      * we calculate total size of all TX rings for max TX fragments number
1456      */
1457     s->max_tx_frags = 0;
1458 
1459     /* TX queues */
1460     for (i = 0; i < s->txq_num; i++) {
1461         hwaddr qdescr_pa =
1462             qdescr_table_pa + i * sizeof(struct Vmxnet3_TxQueueDesc);
1463 
1464         /* Read interrupt number for this TX queue */
1465         s->txq_descr[i].intr_idx =
1466             VMXNET3_READ_TX_QUEUE_DESCR8(d, qdescr_pa, conf.intrIdx);
1467         assert(vmxnet3_verify_intx(s, s->txq_descr[i].intr_idx));
1468 
1469         VMW_CFPRN("TX Queue %d interrupt: %d", i, s->txq_descr[i].intr_idx);
1470 
1471         /* Read rings memory locations for TX queues */
1472         pa = VMXNET3_READ_TX_QUEUE_DESCR64(d, qdescr_pa, conf.txRingBasePA);
1473         size = VMXNET3_READ_TX_QUEUE_DESCR32(d, qdescr_pa, conf.txRingSize);
1474 
1475         vmxnet3_ring_init(d, &s->txq_descr[i].tx_ring, pa, size,
1476                           sizeof(struct Vmxnet3_TxDesc), false);
1477         VMXNET3_RING_DUMP(VMW_CFPRN, "TX", i, &s->txq_descr[i].tx_ring);
1478 
1479         s->max_tx_frags += size;
1480 
1481         /* TXC ring */
1482         pa = VMXNET3_READ_TX_QUEUE_DESCR64(d, qdescr_pa, conf.compRingBasePA);
1483         size = VMXNET3_READ_TX_QUEUE_DESCR32(d, qdescr_pa, conf.compRingSize);
1484         vmxnet3_ring_init(d, &s->txq_descr[i].comp_ring, pa, size,
1485                           sizeof(struct Vmxnet3_TxCompDesc), true);
1486         VMXNET3_RING_DUMP(VMW_CFPRN, "TXC", i, &s->txq_descr[i].comp_ring);
1487 
1488         s->txq_descr[i].tx_stats_pa =
1489             qdescr_pa + offsetof(struct Vmxnet3_TxQueueDesc, stats);
1490 
1491         memset(&s->txq_descr[i].txq_stats, 0,
1492                sizeof(s->txq_descr[i].txq_stats));
1493 
1494         /* Fill device-managed parameters for queues */
1495         VMXNET3_WRITE_TX_QUEUE_DESCR32(d, qdescr_pa,
1496                                        ctrl.txThreshold,
1497                                        VMXNET3_DEF_TX_THRESHOLD);
1498     }
1499 
1500     /* Preallocate TX packet wrapper */
1501     VMW_CFPRN("Max TX fragments is %u", s->max_tx_frags);
1502     net_tx_pkt_init(&s->tx_pkt, PCI_DEVICE(s),
1503                     s->max_tx_frags, s->peer_has_vhdr);
1504     net_rx_pkt_init(&s->rx_pkt, s->peer_has_vhdr);
1505 
1506     /* Read rings memory locations for RX queues */
1507     for (i = 0; i < s->rxq_num; i++) {
1508         int j;
1509         hwaddr qd_pa =
1510             qdescr_table_pa + s->txq_num * sizeof(struct Vmxnet3_TxQueueDesc) +
1511             i * sizeof(struct Vmxnet3_RxQueueDesc);
1512 
1513         /* Read interrupt number for this RX queue */
1514         s->rxq_descr[i].intr_idx =
1515             VMXNET3_READ_TX_QUEUE_DESCR8(d, qd_pa, conf.intrIdx);
1516         assert(vmxnet3_verify_intx(s, s->rxq_descr[i].intr_idx));
1517 
1518         VMW_CFPRN("RX Queue %d interrupt: %d", i, s->rxq_descr[i].intr_idx);
1519 
1520         /* Read rings memory locations */
1521         for (j = 0; j < VMXNET3_RX_RINGS_PER_QUEUE; j++) {
1522             /* RX rings */
1523             pa = VMXNET3_READ_RX_QUEUE_DESCR64(d, qd_pa, conf.rxRingBasePA[j]);
1524             size = VMXNET3_READ_RX_QUEUE_DESCR32(d, qd_pa, conf.rxRingSize[j]);
1525             vmxnet3_ring_init(d, &s->rxq_descr[i].rx_ring[j], pa, size,
1526                               sizeof(struct Vmxnet3_RxDesc), false);
1527             VMW_CFPRN("RX queue %d:%d: Base: %" PRIx64 ", Size: %d",
1528                       i, j, pa, size);
1529         }
1530 
1531         /* RXC ring */
1532         pa = VMXNET3_READ_RX_QUEUE_DESCR64(d, qd_pa, conf.compRingBasePA);
1533         size = VMXNET3_READ_RX_QUEUE_DESCR32(d, qd_pa, conf.compRingSize);
1534         vmxnet3_ring_init(d, &s->rxq_descr[i].comp_ring, pa, size,
1535                           sizeof(struct Vmxnet3_RxCompDesc), true);
1536         VMW_CFPRN("RXC queue %d: Base: %" PRIx64 ", Size: %d", i, pa, size);
1537 
1538         s->rxq_descr[i].rx_stats_pa =
1539             qd_pa + offsetof(struct Vmxnet3_RxQueueDesc, stats);
1540         memset(&s->rxq_descr[i].rxq_stats, 0,
1541                sizeof(s->rxq_descr[i].rxq_stats));
1542     }
1543 
1544     vmxnet3_validate_interrupts(s);
1545 
1546     /* Make sure everything is in place before device activation */
1547     smp_wmb();
1548 
1549     vmxnet3_reset_mac(s);
1550 
1551     s->device_active = true;
1552 }
1553 
1554 static void vmxnet3_handle_command(VMXNET3State *s, uint64_t cmd)
1555 {
1556     s->last_command = cmd;
1557 
1558     switch (cmd) {
1559     case VMXNET3_CMD_GET_PERM_MAC_HI:
1560         VMW_CBPRN("Set: Get upper part of permanent MAC");
1561         break;
1562 
1563     case VMXNET3_CMD_GET_PERM_MAC_LO:
1564         VMW_CBPRN("Set: Get lower part of permanent MAC");
1565         break;
1566 
1567     case VMXNET3_CMD_GET_STATS:
1568         VMW_CBPRN("Set: Get device statistics");
1569         vmxnet3_fill_stats(s);
1570         break;
1571 
1572     case VMXNET3_CMD_ACTIVATE_DEV:
1573         VMW_CBPRN("Set: Activating vmxnet3 device");
1574         vmxnet3_activate_device(s);
1575         break;
1576 
1577     case VMXNET3_CMD_UPDATE_RX_MODE:
1578         VMW_CBPRN("Set: Update rx mode");
1579         vmxnet3_update_rx_mode(s);
1580         break;
1581 
1582     case VMXNET3_CMD_UPDATE_VLAN_FILTERS:
1583         VMW_CBPRN("Set: Update VLAN filters");
1584         vmxnet3_update_vlan_filters(s);
1585         break;
1586 
1587     case VMXNET3_CMD_UPDATE_MAC_FILTERS:
1588         VMW_CBPRN("Set: Update MAC filters");
1589         vmxnet3_update_mcast_filters(s);
1590         break;
1591 
1592     case VMXNET3_CMD_UPDATE_FEATURE:
1593         VMW_CBPRN("Set: Update features");
1594         vmxnet3_update_features(s);
1595         break;
1596 
1597     case VMXNET3_CMD_UPDATE_PMCFG:
1598         VMW_CBPRN("Set: Update power management config");
1599         vmxnet3_update_pm_state(s);
1600         break;
1601 
1602     case VMXNET3_CMD_GET_LINK:
1603         VMW_CBPRN("Set: Get link");
1604         break;
1605 
1606     case VMXNET3_CMD_RESET_DEV:
1607         VMW_CBPRN("Set: Reset device");
1608         vmxnet3_reset(s);
1609         break;
1610 
1611     case VMXNET3_CMD_QUIESCE_DEV:
1612         VMW_CBPRN("Set: VMXNET3_CMD_QUIESCE_DEV - deactivate the device");
1613         vmxnet3_deactivate_device(s);
1614         break;
1615 
1616     case VMXNET3_CMD_GET_CONF_INTR:
1617         VMW_CBPRN("Set: VMXNET3_CMD_GET_CONF_INTR - interrupt configuration");
1618         break;
1619 
1620     case VMXNET3_CMD_GET_ADAPTIVE_RING_INFO:
1621         VMW_CBPRN("Set: VMXNET3_CMD_GET_ADAPTIVE_RING_INFO - "
1622                   "adaptive ring info flags");
1623         break;
1624 
1625     case VMXNET3_CMD_GET_DID_LO:
1626         VMW_CBPRN("Set: Get lower part of device ID");
1627         break;
1628 
1629     case VMXNET3_CMD_GET_DID_HI:
1630         VMW_CBPRN("Set: Get upper part of device ID");
1631         break;
1632 
1633     case VMXNET3_CMD_GET_DEV_EXTRA_INFO:
1634         VMW_CBPRN("Set: Get device extra info");
1635         break;
1636 
1637     default:
1638         VMW_CBPRN("Received unknown command: %" PRIx64, cmd);
1639         break;
1640     }
1641 }
1642 
1643 static uint64_t vmxnet3_get_command_status(VMXNET3State *s)
1644 {
1645     uint64_t ret;
1646 
1647     switch (s->last_command) {
1648     case VMXNET3_CMD_ACTIVATE_DEV:
1649         ret = (s->device_active) ? 0 : 1;
1650         VMW_CFPRN("Device active: %" PRIx64, ret);
1651         break;
1652 
1653     case VMXNET3_CMD_RESET_DEV:
1654     case VMXNET3_CMD_QUIESCE_DEV:
1655     case VMXNET3_CMD_GET_QUEUE_STATUS:
1656     case VMXNET3_CMD_GET_DEV_EXTRA_INFO:
1657         ret = 0;
1658         break;
1659 
1660     case VMXNET3_CMD_GET_LINK:
1661         ret = s->link_status_and_speed;
1662         VMW_CFPRN("Link and speed: %" PRIx64, ret);
1663         break;
1664 
1665     case VMXNET3_CMD_GET_PERM_MAC_LO:
1666         ret = vmxnet3_get_mac_low(&s->perm_mac);
1667         break;
1668 
1669     case VMXNET3_CMD_GET_PERM_MAC_HI:
1670         ret = vmxnet3_get_mac_high(&s->perm_mac);
1671         break;
1672 
1673     case VMXNET3_CMD_GET_CONF_INTR:
1674         ret = vmxnet3_get_interrupt_config(s);
1675         break;
1676 
1677     case VMXNET3_CMD_GET_ADAPTIVE_RING_INFO:
1678         ret = VMXNET3_DISABLE_ADAPTIVE_RING;
1679         break;
1680 
1681     case VMXNET3_CMD_GET_DID_LO:
1682         ret = PCI_DEVICE_ID_VMWARE_VMXNET3;
1683         break;
1684 
1685     case VMXNET3_CMD_GET_DID_HI:
1686         ret = VMXNET3_DEVICE_REVISION;
1687         break;
1688 
1689     default:
1690         VMW_WRPRN("Received request for unknown command: %x", s->last_command);
1691         ret = 0;
1692         break;
1693     }
1694 
1695     return ret;
1696 }
1697 
1698 static void vmxnet3_set_events(VMXNET3State *s, uint32_t val)
1699 {
1700     uint32_t events;
1701     PCIDevice *d = PCI_DEVICE(s);
1702 
1703     VMW_CBPRN("Setting events: 0x%x", val);
1704     events = VMXNET3_READ_DRV_SHARED32(d, s->drv_shmem, ecr) | val;
1705     VMXNET3_WRITE_DRV_SHARED32(d, s->drv_shmem, ecr, events);
1706 }
1707 
1708 static void vmxnet3_ack_events(VMXNET3State *s, uint32_t val)
1709 {
1710     PCIDevice *d = PCI_DEVICE(s);
1711     uint32_t events;
1712 
1713     VMW_CBPRN("Clearing events: 0x%x", val);
1714     events = VMXNET3_READ_DRV_SHARED32(d, s->drv_shmem, ecr) & ~val;
1715     VMXNET3_WRITE_DRV_SHARED32(d, s->drv_shmem, ecr, events);
1716 }
1717 
1718 static void
1719 vmxnet3_io_bar1_write(void *opaque,
1720                       hwaddr addr,
1721                       uint64_t val,
1722                       unsigned size)
1723 {
1724     VMXNET3State *s = opaque;
1725 
1726     switch (addr) {
1727     /* Vmxnet3 Revision Report Selection */
1728     case VMXNET3_REG_VRRS:
1729         VMW_CBPRN("Write BAR1 [VMXNET3_REG_VRRS] = %" PRIx64 ", size %d",
1730                   val, size);
1731         break;
1732 
1733     /* UPT Version Report Selection */
1734     case VMXNET3_REG_UVRS:
1735         VMW_CBPRN("Write BAR1 [VMXNET3_REG_UVRS] = %" PRIx64 ", size %d",
1736                   val, size);
1737         break;
1738 
1739     /* Driver Shared Address Low */
1740     case VMXNET3_REG_DSAL:
1741         VMW_CBPRN("Write BAR1 [VMXNET3_REG_DSAL] = %" PRIx64 ", size %d",
1742                   val, size);
1743         /*
1744          * Guest driver will first write the low part of the shared
1745          * memory address. We save it to temp variable and set the
1746          * shared address only after we get the high part
1747          */
1748         if (val == 0) {
1749             vmxnet3_deactivate_device(s);
1750         }
1751         s->temp_shared_guest_driver_memory = val;
1752         s->drv_shmem = 0;
1753         break;
1754 
1755     /* Driver Shared Address High */
1756     case VMXNET3_REG_DSAH:
1757         VMW_CBPRN("Write BAR1 [VMXNET3_REG_DSAH] = %" PRIx64 ", size %d",
1758                   val, size);
1759         /*
1760          * Set the shared memory between guest driver and device.
1761          * We already should have low address part.
1762          */
1763         s->drv_shmem = s->temp_shared_guest_driver_memory | (val << 32);
1764         break;
1765 
1766     /* Command */
1767     case VMXNET3_REG_CMD:
1768         VMW_CBPRN("Write BAR1 [VMXNET3_REG_CMD] = %" PRIx64 ", size %d",
1769                   val, size);
1770         vmxnet3_handle_command(s, val);
1771         break;
1772 
1773     /* MAC Address Low */
1774     case VMXNET3_REG_MACL:
1775         VMW_CBPRN("Write BAR1 [VMXNET3_REG_MACL] = %" PRIx64 ", size %d",
1776                   val, size);
1777         s->temp_mac = val;
1778         break;
1779 
1780     /* MAC Address High */
1781     case VMXNET3_REG_MACH:
1782         VMW_CBPRN("Write BAR1 [VMXNET3_REG_MACH] = %" PRIx64 ", size %d",
1783                   val, size);
1784         vmxnet3_set_variable_mac(s, val, s->temp_mac);
1785         break;
1786 
1787     /* Interrupt Cause Register */
1788     case VMXNET3_REG_ICR:
1789         VMW_CBPRN("Write BAR1 [VMXNET3_REG_ICR] = %" PRIx64 ", size %d",
1790                   val, size);
1791         g_assert_not_reached();
1792         break;
1793 
1794     /* Event Cause Register */
1795     case VMXNET3_REG_ECR:
1796         VMW_CBPRN("Write BAR1 [VMXNET3_REG_ECR] = %" PRIx64 ", size %d",
1797                   val, size);
1798         vmxnet3_ack_events(s, val);
1799         break;
1800 
1801     default:
1802         VMW_CBPRN("Unknown Write to BAR1 [%" PRIx64 "] = %" PRIx64 ", size %d",
1803                   addr, val, size);
1804         break;
1805     }
1806 }
1807 
1808 static uint64_t
1809 vmxnet3_io_bar1_read(void *opaque, hwaddr addr, unsigned size)
1810 {
1811         VMXNET3State *s = opaque;
1812         uint64_t ret = 0;
1813 
1814         switch (addr) {
1815         /* Vmxnet3 Revision Report Selection */
1816         case VMXNET3_REG_VRRS:
1817             VMW_CBPRN("Read BAR1 [VMXNET3_REG_VRRS], size %d", size);
1818             ret = VMXNET3_DEVICE_REVISION;
1819             break;
1820 
1821         /* UPT Version Report Selection */
1822         case VMXNET3_REG_UVRS:
1823             VMW_CBPRN("Read BAR1 [VMXNET3_REG_UVRS], size %d", size);
1824             ret = VMXNET3_UPT_REVISION;
1825             break;
1826 
1827         /* Command */
1828         case VMXNET3_REG_CMD:
1829             VMW_CBPRN("Read BAR1 [VMXNET3_REG_CMD], size %d", size);
1830             ret = vmxnet3_get_command_status(s);
1831             break;
1832 
1833         /* MAC Address Low */
1834         case VMXNET3_REG_MACL:
1835             VMW_CBPRN("Read BAR1 [VMXNET3_REG_MACL], size %d", size);
1836             ret = vmxnet3_get_mac_low(&s->conf.macaddr);
1837             break;
1838 
1839         /* MAC Address High */
1840         case VMXNET3_REG_MACH:
1841             VMW_CBPRN("Read BAR1 [VMXNET3_REG_MACH], size %d", size);
1842             ret = vmxnet3_get_mac_high(&s->conf.macaddr);
1843             break;
1844 
1845         /*
1846          * Interrupt Cause Register
1847          * Used for legacy interrupts only so interrupt index always 0
1848          */
1849         case VMXNET3_REG_ICR:
1850             VMW_CBPRN("Read BAR1 [VMXNET3_REG_ICR], size %d", size);
1851             if (vmxnet3_interrupt_asserted(s, 0)) {
1852                 vmxnet3_clear_interrupt(s, 0);
1853                 ret = true;
1854             } else {
1855                 ret = false;
1856             }
1857             break;
1858 
1859         default:
1860             VMW_CBPRN("Unknow read BAR1[%" PRIx64 "], %d bytes", addr, size);
1861             break;
1862         }
1863 
1864         return ret;
1865 }
1866 
1867 static int
1868 vmxnet3_can_receive(NetClientState *nc)
1869 {
1870     VMXNET3State *s = qemu_get_nic_opaque(nc);
1871     return s->device_active &&
1872            VMXNET_FLAG_IS_SET(s->link_status_and_speed, VMXNET3_LINK_STATUS_UP);
1873 }
1874 
1875 static inline bool
1876 vmxnet3_is_registered_vlan(VMXNET3State *s, const void *data)
1877 {
1878     uint16_t vlan_tag = eth_get_pkt_tci(data) & VLAN_VID_MASK;
1879     if (IS_SPECIAL_VLAN_ID(vlan_tag)) {
1880         return true;
1881     }
1882 
1883     return VMXNET3_VFTABLE_ENTRY_IS_SET(s->vlan_table, vlan_tag);
1884 }
1885 
1886 static bool
1887 vmxnet3_is_allowed_mcast_group(VMXNET3State *s, const uint8_t *group_mac)
1888 {
1889     int i;
1890     for (i = 0; i < s->mcast_list_len; i++) {
1891         if (!memcmp(group_mac, s->mcast_list[i].a, sizeof(s->mcast_list[i]))) {
1892             return true;
1893         }
1894     }
1895     return false;
1896 }
1897 
1898 static bool
1899 vmxnet3_rx_filter_may_indicate(VMXNET3State *s, const void *data,
1900     size_t size)
1901 {
1902     struct eth_header *ehdr = PKT_GET_ETH_HDR(data);
1903 
1904     if (VMXNET_FLAG_IS_SET(s->rx_mode, VMXNET3_RXM_PROMISC)) {
1905         return true;
1906     }
1907 
1908     if (!vmxnet3_is_registered_vlan(s, data)) {
1909         return false;
1910     }
1911 
1912     switch (net_rx_pkt_get_packet_type(s->rx_pkt)) {
1913     case ETH_PKT_UCAST:
1914         if (!VMXNET_FLAG_IS_SET(s->rx_mode, VMXNET3_RXM_UCAST)) {
1915             return false;
1916         }
1917         if (memcmp(s->conf.macaddr.a, ehdr->h_dest, ETH_ALEN)) {
1918             return false;
1919         }
1920         break;
1921 
1922     case ETH_PKT_BCAST:
1923         if (!VMXNET_FLAG_IS_SET(s->rx_mode, VMXNET3_RXM_BCAST)) {
1924             return false;
1925         }
1926         break;
1927 
1928     case ETH_PKT_MCAST:
1929         if (VMXNET_FLAG_IS_SET(s->rx_mode, VMXNET3_RXM_ALL_MULTI)) {
1930             return true;
1931         }
1932         if (!VMXNET_FLAG_IS_SET(s->rx_mode, VMXNET3_RXM_MCAST)) {
1933             return false;
1934         }
1935         if (!vmxnet3_is_allowed_mcast_group(s, ehdr->h_dest)) {
1936             return false;
1937         }
1938         break;
1939 
1940     default:
1941         g_assert_not_reached();
1942     }
1943 
1944     return true;
1945 }
1946 
1947 static ssize_t
1948 vmxnet3_receive(NetClientState *nc, const uint8_t *buf, size_t size)
1949 {
1950     VMXNET3State *s = qemu_get_nic_opaque(nc);
1951     size_t bytes_indicated;
1952     uint8_t min_buf[MIN_BUF_SIZE];
1953 
1954     if (!vmxnet3_can_receive(nc)) {
1955         VMW_PKPRN("Cannot receive now");
1956         return -1;
1957     }
1958 
1959     if (s->peer_has_vhdr) {
1960         net_rx_pkt_set_vhdr(s->rx_pkt, (struct virtio_net_hdr *)buf);
1961         buf += sizeof(struct virtio_net_hdr);
1962         size -= sizeof(struct virtio_net_hdr);
1963     }
1964 
1965     /* Pad to minimum Ethernet frame length */
1966     if (size < sizeof(min_buf)) {
1967         memcpy(min_buf, buf, size);
1968         memset(&min_buf[size], 0, sizeof(min_buf) - size);
1969         buf = min_buf;
1970         size = sizeof(min_buf);
1971     }
1972 
1973     net_rx_pkt_set_packet_type(s->rx_pkt,
1974         get_eth_packet_type(PKT_GET_ETH_HDR(buf)));
1975 
1976     if (vmxnet3_rx_filter_may_indicate(s, buf, size)) {
1977         net_rx_pkt_set_protocols(s->rx_pkt, buf, size);
1978         vmxnet3_rx_need_csum_calculate(s->rx_pkt, buf, size);
1979         net_rx_pkt_attach_data(s->rx_pkt, buf, size, s->rx_vlan_stripping);
1980         bytes_indicated = vmxnet3_indicate_packet(s) ? size : -1;
1981         if (bytes_indicated < size) {
1982             VMW_PKPRN("RX: %zu of %zu bytes indicated", bytes_indicated, size);
1983         }
1984     } else {
1985         VMW_PKPRN("Packet dropped by RX filter");
1986         bytes_indicated = size;
1987     }
1988 
1989     assert(size > 0);
1990     assert(bytes_indicated != 0);
1991     return bytes_indicated;
1992 }
1993 
1994 static void vmxnet3_set_link_status(NetClientState *nc)
1995 {
1996     VMXNET3State *s = qemu_get_nic_opaque(nc);
1997 
1998     if (nc->link_down) {
1999         s->link_status_and_speed &= ~VMXNET3_LINK_STATUS_UP;
2000     } else {
2001         s->link_status_and_speed |= VMXNET3_LINK_STATUS_UP;
2002     }
2003 
2004     vmxnet3_set_events(s, VMXNET3_ECR_LINK);
2005     vmxnet3_trigger_interrupt(s, s->event_int_idx);
2006 }
2007 
2008 static NetClientInfo net_vmxnet3_info = {
2009         .type = NET_CLIENT_DRIVER_NIC,
2010         .size = sizeof(NICState),
2011         .receive = vmxnet3_receive,
2012         .link_status_changed = vmxnet3_set_link_status,
2013 };
2014 
2015 static bool vmxnet3_peer_has_vnet_hdr(VMXNET3State *s)
2016 {
2017     NetClientState *nc = qemu_get_queue(s->nic);
2018 
2019     if (qemu_has_vnet_hdr(nc->peer)) {
2020         return true;
2021     }
2022 
2023     return false;
2024 }
2025 
2026 static void vmxnet3_net_uninit(VMXNET3State *s)
2027 {
2028     g_free(s->mcast_list);
2029     vmxnet3_deactivate_device(s);
2030     qemu_del_nic(s->nic);
2031 }
2032 
2033 static void vmxnet3_net_init(VMXNET3State *s)
2034 {
2035     DeviceState *d = DEVICE(s);
2036 
2037     VMW_CBPRN("vmxnet3_net_init called...");
2038 
2039     qemu_macaddr_default_if_unset(&s->conf.macaddr);
2040 
2041     /* Windows guest will query the address that was set on init */
2042     memcpy(&s->perm_mac.a, &s->conf.macaddr.a, sizeof(s->perm_mac.a));
2043 
2044     s->mcast_list = NULL;
2045     s->mcast_list_len = 0;
2046 
2047     s->link_status_and_speed = VMXNET3_LINK_SPEED | VMXNET3_LINK_STATUS_UP;
2048 
2049     VMW_CFPRN("Permanent MAC: " MAC_FMT, MAC_ARG(s->perm_mac.a));
2050 
2051     s->nic = qemu_new_nic(&net_vmxnet3_info, &s->conf,
2052                           object_get_typename(OBJECT(s)),
2053                           d->id, s);
2054 
2055     s->peer_has_vhdr = vmxnet3_peer_has_vnet_hdr(s);
2056     s->tx_sop = true;
2057     s->skip_current_tx_pkt = false;
2058     s->tx_pkt = NULL;
2059     s->rx_pkt = NULL;
2060     s->rx_vlan_stripping = false;
2061     s->lro_supported = false;
2062 
2063     if (s->peer_has_vhdr) {
2064         qemu_set_vnet_hdr_len(qemu_get_queue(s->nic)->peer,
2065             sizeof(struct virtio_net_hdr));
2066 
2067         qemu_using_vnet_hdr(qemu_get_queue(s->nic)->peer, 1);
2068     }
2069 
2070     qemu_format_nic_info_str(qemu_get_queue(s->nic), s->conf.macaddr.a);
2071 }
2072 
2073 static void
2074 vmxnet3_unuse_msix_vectors(VMXNET3State *s, int num_vectors)
2075 {
2076     PCIDevice *d = PCI_DEVICE(s);
2077     int i;
2078     for (i = 0; i < num_vectors; i++) {
2079         msix_vector_unuse(d, i);
2080     }
2081 }
2082 
2083 static bool
2084 vmxnet3_use_msix_vectors(VMXNET3State *s, int num_vectors)
2085 {
2086     PCIDevice *d = PCI_DEVICE(s);
2087     int i;
2088     for (i = 0; i < num_vectors; i++) {
2089         int res = msix_vector_use(d, i);
2090         if (0 > res) {
2091             VMW_WRPRN("Failed to use MSI-X vector %d, error %d", i, res);
2092             vmxnet3_unuse_msix_vectors(s, i);
2093             return false;
2094         }
2095     }
2096     return true;
2097 }
2098 
2099 static bool
2100 vmxnet3_init_msix(VMXNET3State *s)
2101 {
2102     PCIDevice *d = PCI_DEVICE(s);
2103     int res = msix_init(d, VMXNET3_MAX_INTRS,
2104                         &s->msix_bar,
2105                         VMXNET3_MSIX_BAR_IDX, VMXNET3_OFF_MSIX_TABLE,
2106                         &s->msix_bar,
2107                         VMXNET3_MSIX_BAR_IDX, VMXNET3_OFF_MSIX_PBA(s),
2108                         VMXNET3_MSIX_OFFSET(s), NULL);
2109 
2110     if (0 > res) {
2111         VMW_WRPRN("Failed to initialize MSI-X, error %d", res);
2112         s->msix_used = false;
2113     } else {
2114         if (!vmxnet3_use_msix_vectors(s, VMXNET3_MAX_INTRS)) {
2115             VMW_WRPRN("Failed to use MSI-X vectors, error %d", res);
2116             msix_uninit(d, &s->msix_bar, &s->msix_bar);
2117             s->msix_used = false;
2118         } else {
2119             s->msix_used = true;
2120         }
2121     }
2122     return s->msix_used;
2123 }
2124 
2125 static void
2126 vmxnet3_cleanup_msix(VMXNET3State *s)
2127 {
2128     PCIDevice *d = PCI_DEVICE(s);
2129 
2130     if (s->msix_used) {
2131         vmxnet3_unuse_msix_vectors(s, VMXNET3_MAX_INTRS);
2132         msix_uninit(d, &s->msix_bar, &s->msix_bar);
2133     }
2134 }
2135 
2136 static void
2137 vmxnet3_cleanup_msi(VMXNET3State *s)
2138 {
2139     PCIDevice *d = PCI_DEVICE(s);
2140 
2141     msi_uninit(d);
2142 }
2143 
2144 static void
2145 vmxnet3_msix_save(QEMUFile *f, void *opaque)
2146 {
2147     PCIDevice *d = PCI_DEVICE(opaque);
2148     msix_save(d, f);
2149 }
2150 
2151 static int
2152 vmxnet3_msix_load(QEMUFile *f, void *opaque, int version_id)
2153 {
2154     PCIDevice *d = PCI_DEVICE(opaque);
2155     msix_load(d, f);
2156     return 0;
2157 }
2158 
2159 static const MemoryRegionOps b0_ops = {
2160     .read = vmxnet3_io_bar0_read,
2161     .write = vmxnet3_io_bar0_write,
2162     .endianness = DEVICE_LITTLE_ENDIAN,
2163     .impl = {
2164             .min_access_size = 4,
2165             .max_access_size = 4,
2166     },
2167 };
2168 
2169 static const MemoryRegionOps b1_ops = {
2170     .read = vmxnet3_io_bar1_read,
2171     .write = vmxnet3_io_bar1_write,
2172     .endianness = DEVICE_LITTLE_ENDIAN,
2173     .impl = {
2174             .min_access_size = 4,
2175             .max_access_size = 4,
2176     },
2177 };
2178 
2179 static SaveVMHandlers savevm_vmxnet3_msix = {
2180     .save_state = vmxnet3_msix_save,
2181     .load_state = vmxnet3_msix_load,
2182 };
2183 
2184 static uint64_t vmxnet3_device_serial_num(VMXNET3State *s)
2185 {
2186     uint64_t dsn_payload;
2187     uint8_t *dsnp = (uint8_t *)&dsn_payload;
2188 
2189     dsnp[0] = 0xfe;
2190     dsnp[1] = s->conf.macaddr.a[3];
2191     dsnp[2] = s->conf.macaddr.a[4];
2192     dsnp[3] = s->conf.macaddr.a[5];
2193     dsnp[4] = s->conf.macaddr.a[0];
2194     dsnp[5] = s->conf.macaddr.a[1];
2195     dsnp[6] = s->conf.macaddr.a[2];
2196     dsnp[7] = 0xff;
2197     return dsn_payload;
2198 }
2199 
2200 
2201 #define VMXNET3_USE_64BIT         (true)
2202 #define VMXNET3_PER_VECTOR_MASK   (false)
2203 
2204 static void vmxnet3_pci_realize(PCIDevice *pci_dev, Error **errp)
2205 {
2206     DeviceState *dev = DEVICE(pci_dev);
2207     VMXNET3State *s = VMXNET3(pci_dev);
2208     int ret;
2209 
2210     VMW_CBPRN("Starting init...");
2211 
2212     memory_region_init_io(&s->bar0, OBJECT(s), &b0_ops, s,
2213                           "vmxnet3-b0", VMXNET3_PT_REG_SIZE);
2214     pci_register_bar(pci_dev, VMXNET3_BAR0_IDX,
2215                      PCI_BASE_ADDRESS_SPACE_MEMORY, &s->bar0);
2216 
2217     memory_region_init_io(&s->bar1, OBJECT(s), &b1_ops, s,
2218                           "vmxnet3-b1", VMXNET3_VD_REG_SIZE);
2219     pci_register_bar(pci_dev, VMXNET3_BAR1_IDX,
2220                      PCI_BASE_ADDRESS_SPACE_MEMORY, &s->bar1);
2221 
2222     memory_region_init(&s->msix_bar, OBJECT(s), "vmxnet3-msix-bar",
2223                        VMXNET3_MSIX_BAR_SIZE);
2224     pci_register_bar(pci_dev, VMXNET3_MSIX_BAR_IDX,
2225                      PCI_BASE_ADDRESS_SPACE_MEMORY, &s->msix_bar);
2226 
2227     vmxnet3_reset_interrupt_states(s);
2228 
2229     /* Interrupt pin A */
2230     pci_dev->config[PCI_INTERRUPT_PIN] = 0x01;
2231 
2232     ret = msi_init(pci_dev, VMXNET3_MSI_OFFSET(s), VMXNET3_MAX_NMSIX_INTRS,
2233                    VMXNET3_USE_64BIT, VMXNET3_PER_VECTOR_MASK, NULL);
2234     /* Any error other than -ENOTSUP(board's MSI support is broken)
2235      * is a programming error. Fall back to INTx silently on -ENOTSUP */
2236     assert(!ret || ret == -ENOTSUP);
2237 
2238     if (!vmxnet3_init_msix(s)) {
2239         VMW_WRPRN("Failed to initialize MSI-X, configuration is inconsistent.");
2240     }
2241 
2242     vmxnet3_net_init(s);
2243 
2244     if (pci_is_express(pci_dev)) {
2245         if (pci_bus_is_express(pci_get_bus(pci_dev))) {
2246             pcie_endpoint_cap_init(pci_dev, VMXNET3_EXP_EP_OFFSET);
2247         }
2248 
2249         pcie_dev_ser_num_init(pci_dev, VMXNET3_DSN_OFFSET,
2250                               vmxnet3_device_serial_num(s));
2251     }
2252 
2253     register_savevm_live(dev, "vmxnet3-msix", -1, 1, &savevm_vmxnet3_msix, s);
2254 }
2255 
2256 static void vmxnet3_instance_init(Object *obj)
2257 {
2258     VMXNET3State *s = VMXNET3(obj);
2259     device_add_bootindex_property(obj, &s->conf.bootindex,
2260                                   "bootindex", "/ethernet-phy@0",
2261                                   DEVICE(obj), NULL);
2262 }
2263 
2264 static void vmxnet3_pci_uninit(PCIDevice *pci_dev)
2265 {
2266     DeviceState *dev = DEVICE(pci_dev);
2267     VMXNET3State *s = VMXNET3(pci_dev);
2268 
2269     VMW_CBPRN("Starting uninit...");
2270 
2271     unregister_savevm(dev, "vmxnet3-msix", s);
2272 
2273     vmxnet3_net_uninit(s);
2274 
2275     vmxnet3_cleanup_msix(s);
2276 
2277     vmxnet3_cleanup_msi(s);
2278 }
2279 
2280 static void vmxnet3_qdev_reset(DeviceState *dev)
2281 {
2282     PCIDevice *d = PCI_DEVICE(dev);
2283     VMXNET3State *s = VMXNET3(d);
2284 
2285     VMW_CBPRN("Starting QDEV reset...");
2286     vmxnet3_reset(s);
2287 }
2288 
2289 static bool vmxnet3_mc_list_needed(void *opaque)
2290 {
2291     return true;
2292 }
2293 
2294 static int vmxnet3_mcast_list_pre_load(void *opaque)
2295 {
2296     VMXNET3State *s = opaque;
2297 
2298     s->mcast_list = g_malloc(s->mcast_list_buff_size);
2299 
2300     return 0;
2301 }
2302 
2303 
2304 static int vmxnet3_pre_save(void *opaque)
2305 {
2306     VMXNET3State *s = opaque;
2307 
2308     s->mcast_list_buff_size = s->mcast_list_len * sizeof(MACAddr);
2309 
2310     return 0;
2311 }
2312 
2313 static const VMStateDescription vmxstate_vmxnet3_mcast_list = {
2314     .name = "vmxnet3/mcast_list",
2315     .version_id = 1,
2316     .minimum_version_id = 1,
2317     .pre_load = vmxnet3_mcast_list_pre_load,
2318     .needed = vmxnet3_mc_list_needed,
2319     .fields = (VMStateField[]) {
2320         VMSTATE_VBUFFER_UINT32(mcast_list, VMXNET3State, 0, NULL,
2321             mcast_list_buff_size),
2322         VMSTATE_END_OF_LIST()
2323     }
2324 };
2325 
2326 static const VMStateDescription vmstate_vmxnet3_ring = {
2327     .name = "vmxnet3-ring",
2328     .version_id = 0,
2329     .fields = (VMStateField[]) {
2330         VMSTATE_UINT64(pa, Vmxnet3Ring),
2331         VMSTATE_UINT32(size, Vmxnet3Ring),
2332         VMSTATE_UINT32(cell_size, Vmxnet3Ring),
2333         VMSTATE_UINT32(next, Vmxnet3Ring),
2334         VMSTATE_UINT8(gen, Vmxnet3Ring),
2335         VMSTATE_END_OF_LIST()
2336     }
2337 };
2338 
2339 static const VMStateDescription vmstate_vmxnet3_tx_stats = {
2340     .name = "vmxnet3-tx-stats",
2341     .version_id = 0,
2342     .fields = (VMStateField[]) {
2343         VMSTATE_UINT64(TSOPktsTxOK, struct UPT1_TxStats),
2344         VMSTATE_UINT64(TSOBytesTxOK, struct UPT1_TxStats),
2345         VMSTATE_UINT64(ucastPktsTxOK, struct UPT1_TxStats),
2346         VMSTATE_UINT64(ucastBytesTxOK, struct UPT1_TxStats),
2347         VMSTATE_UINT64(mcastPktsTxOK, struct UPT1_TxStats),
2348         VMSTATE_UINT64(mcastBytesTxOK, struct UPT1_TxStats),
2349         VMSTATE_UINT64(bcastPktsTxOK, struct UPT1_TxStats),
2350         VMSTATE_UINT64(bcastBytesTxOK, struct UPT1_TxStats),
2351         VMSTATE_UINT64(pktsTxError, struct UPT1_TxStats),
2352         VMSTATE_UINT64(pktsTxDiscard, struct UPT1_TxStats),
2353         VMSTATE_END_OF_LIST()
2354     }
2355 };
2356 
2357 static const VMStateDescription vmstate_vmxnet3_txq_descr = {
2358     .name = "vmxnet3-txq-descr",
2359     .version_id = 0,
2360     .fields = (VMStateField[]) {
2361         VMSTATE_STRUCT(tx_ring, Vmxnet3TxqDescr, 0, vmstate_vmxnet3_ring,
2362                        Vmxnet3Ring),
2363         VMSTATE_STRUCT(comp_ring, Vmxnet3TxqDescr, 0, vmstate_vmxnet3_ring,
2364                        Vmxnet3Ring),
2365         VMSTATE_UINT8(intr_idx, Vmxnet3TxqDescr),
2366         VMSTATE_UINT64(tx_stats_pa, Vmxnet3TxqDescr),
2367         VMSTATE_STRUCT(txq_stats, Vmxnet3TxqDescr, 0, vmstate_vmxnet3_tx_stats,
2368                        struct UPT1_TxStats),
2369         VMSTATE_END_OF_LIST()
2370     }
2371 };
2372 
2373 static const VMStateDescription vmstate_vmxnet3_rx_stats = {
2374     .name = "vmxnet3-rx-stats",
2375     .version_id = 0,
2376     .fields = (VMStateField[]) {
2377         VMSTATE_UINT64(LROPktsRxOK, struct UPT1_RxStats),
2378         VMSTATE_UINT64(LROBytesRxOK, struct UPT1_RxStats),
2379         VMSTATE_UINT64(ucastPktsRxOK, struct UPT1_RxStats),
2380         VMSTATE_UINT64(ucastBytesRxOK, struct UPT1_RxStats),
2381         VMSTATE_UINT64(mcastPktsRxOK, struct UPT1_RxStats),
2382         VMSTATE_UINT64(mcastBytesRxOK, struct UPT1_RxStats),
2383         VMSTATE_UINT64(bcastPktsRxOK, struct UPT1_RxStats),
2384         VMSTATE_UINT64(bcastBytesRxOK, struct UPT1_RxStats),
2385         VMSTATE_UINT64(pktsRxOutOfBuf, struct UPT1_RxStats),
2386         VMSTATE_UINT64(pktsRxError, struct UPT1_RxStats),
2387         VMSTATE_END_OF_LIST()
2388     }
2389 };
2390 
2391 static const VMStateDescription vmstate_vmxnet3_rxq_descr = {
2392     .name = "vmxnet3-rxq-descr",
2393     .version_id = 0,
2394     .fields = (VMStateField[]) {
2395         VMSTATE_STRUCT_ARRAY(rx_ring, Vmxnet3RxqDescr,
2396                              VMXNET3_RX_RINGS_PER_QUEUE, 0,
2397                              vmstate_vmxnet3_ring, Vmxnet3Ring),
2398         VMSTATE_STRUCT(comp_ring, Vmxnet3RxqDescr, 0, vmstate_vmxnet3_ring,
2399                        Vmxnet3Ring),
2400         VMSTATE_UINT8(intr_idx, Vmxnet3RxqDescr),
2401         VMSTATE_UINT64(rx_stats_pa, Vmxnet3RxqDescr),
2402         VMSTATE_STRUCT(rxq_stats, Vmxnet3RxqDescr, 0, vmstate_vmxnet3_rx_stats,
2403                        struct UPT1_RxStats),
2404         VMSTATE_END_OF_LIST()
2405     }
2406 };
2407 
2408 static int vmxnet3_post_load(void *opaque, int version_id)
2409 {
2410     VMXNET3State *s = opaque;
2411     PCIDevice *d = PCI_DEVICE(s);
2412 
2413     net_tx_pkt_init(&s->tx_pkt, PCI_DEVICE(s),
2414                     s->max_tx_frags, s->peer_has_vhdr);
2415     net_rx_pkt_init(&s->rx_pkt, s->peer_has_vhdr);
2416 
2417     if (s->msix_used) {
2418         if  (!vmxnet3_use_msix_vectors(s, VMXNET3_MAX_INTRS)) {
2419             VMW_WRPRN("Failed to re-use MSI-X vectors");
2420             msix_uninit(d, &s->msix_bar, &s->msix_bar);
2421             s->msix_used = false;
2422             return -1;
2423         }
2424     }
2425 
2426     vmxnet3_validate_queues(s);
2427     vmxnet3_validate_interrupts(s);
2428 
2429     return 0;
2430 }
2431 
2432 static const VMStateDescription vmstate_vmxnet3_int_state = {
2433     .name = "vmxnet3-int-state",
2434     .version_id = 0,
2435     .fields = (VMStateField[]) {
2436         VMSTATE_BOOL(is_masked, Vmxnet3IntState),
2437         VMSTATE_BOOL(is_pending, Vmxnet3IntState),
2438         VMSTATE_BOOL(is_asserted, Vmxnet3IntState),
2439         VMSTATE_END_OF_LIST()
2440     }
2441 };
2442 
2443 static bool vmxnet3_vmstate_need_pcie_device(void *opaque)
2444 {
2445     VMXNET3State *s = VMXNET3(opaque);
2446 
2447     return !(s->compat_flags & VMXNET3_COMPAT_FLAG_DISABLE_PCIE);
2448 }
2449 
2450 static bool vmxnet3_vmstate_test_pci_device(void *opaque, int version_id)
2451 {
2452     return !vmxnet3_vmstate_need_pcie_device(opaque);
2453 }
2454 
2455 static const VMStateDescription vmstate_vmxnet3_pcie_device = {
2456     .name = "vmxnet3/pcie",
2457     .version_id = 1,
2458     .minimum_version_id = 1,
2459     .needed = vmxnet3_vmstate_need_pcie_device,
2460     .fields = (VMStateField[]) {
2461         VMSTATE_PCI_DEVICE(parent_obj, VMXNET3State),
2462         VMSTATE_END_OF_LIST()
2463     }
2464 };
2465 
2466 static const VMStateDescription vmstate_vmxnet3 = {
2467     .name = "vmxnet3",
2468     .version_id = 1,
2469     .minimum_version_id = 1,
2470     .pre_save = vmxnet3_pre_save,
2471     .post_load = vmxnet3_post_load,
2472     .fields = (VMStateField[]) {
2473             VMSTATE_STRUCT_TEST(parent_obj, VMXNET3State,
2474                                 vmxnet3_vmstate_test_pci_device, 0,
2475                                 vmstate_pci_device, PCIDevice),
2476             VMSTATE_BOOL(rx_packets_compound, VMXNET3State),
2477             VMSTATE_BOOL(rx_vlan_stripping, VMXNET3State),
2478             VMSTATE_BOOL(lro_supported, VMXNET3State),
2479             VMSTATE_UINT32(rx_mode, VMXNET3State),
2480             VMSTATE_UINT32(mcast_list_len, VMXNET3State),
2481             VMSTATE_UINT32(mcast_list_buff_size, VMXNET3State),
2482             VMSTATE_UINT32_ARRAY(vlan_table, VMXNET3State, VMXNET3_VFT_SIZE),
2483             VMSTATE_UINT32(mtu, VMXNET3State),
2484             VMSTATE_UINT16(max_rx_frags, VMXNET3State),
2485             VMSTATE_UINT32(max_tx_frags, VMXNET3State),
2486             VMSTATE_UINT8(event_int_idx, VMXNET3State),
2487             VMSTATE_BOOL(auto_int_masking, VMXNET3State),
2488             VMSTATE_UINT8(txq_num, VMXNET3State),
2489             VMSTATE_UINT8(rxq_num, VMXNET3State),
2490             VMSTATE_UINT32(device_active, VMXNET3State),
2491             VMSTATE_UINT32(last_command, VMXNET3State),
2492             VMSTATE_UINT32(link_status_and_speed, VMXNET3State),
2493             VMSTATE_UINT32(temp_mac, VMXNET3State),
2494             VMSTATE_UINT64(drv_shmem, VMXNET3State),
2495             VMSTATE_UINT64(temp_shared_guest_driver_memory, VMXNET3State),
2496 
2497             VMSTATE_STRUCT_ARRAY(txq_descr, VMXNET3State,
2498                 VMXNET3_DEVICE_MAX_TX_QUEUES, 0, vmstate_vmxnet3_txq_descr,
2499                 Vmxnet3TxqDescr),
2500             VMSTATE_STRUCT_ARRAY(rxq_descr, VMXNET3State,
2501                 VMXNET3_DEVICE_MAX_RX_QUEUES, 0, vmstate_vmxnet3_rxq_descr,
2502                 Vmxnet3RxqDescr),
2503             VMSTATE_STRUCT_ARRAY(interrupt_states, VMXNET3State,
2504                 VMXNET3_MAX_INTRS, 0, vmstate_vmxnet3_int_state,
2505                 Vmxnet3IntState),
2506 
2507             VMSTATE_END_OF_LIST()
2508     },
2509     .subsections = (const VMStateDescription*[]) {
2510         &vmxstate_vmxnet3_mcast_list,
2511         &vmstate_vmxnet3_pcie_device,
2512         NULL
2513     }
2514 };
2515 
2516 static Property vmxnet3_properties[] = {
2517     DEFINE_NIC_PROPERTIES(VMXNET3State, conf),
2518     DEFINE_PROP_BIT("x-old-msi-offsets", VMXNET3State, compat_flags,
2519                     VMXNET3_COMPAT_FLAG_OLD_MSI_OFFSETS_BIT, false),
2520     DEFINE_PROP_BIT("x-disable-pcie", VMXNET3State, compat_flags,
2521                     VMXNET3_COMPAT_FLAG_DISABLE_PCIE_BIT, false),
2522     DEFINE_PROP_END_OF_LIST(),
2523 };
2524 
2525 static void vmxnet3_realize(DeviceState *qdev, Error **errp)
2526 {
2527     VMXNET3Class *vc = VMXNET3_DEVICE_GET_CLASS(qdev);
2528     PCIDevice *pci_dev = PCI_DEVICE(qdev);
2529     VMXNET3State *s = VMXNET3(qdev);
2530 
2531     if (!(s->compat_flags & VMXNET3_COMPAT_FLAG_DISABLE_PCIE)) {
2532         pci_dev->cap_present |= QEMU_PCI_CAP_EXPRESS;
2533     }
2534 
2535     vc->parent_dc_realize(qdev, errp);
2536 }
2537 
2538 static void vmxnet3_class_init(ObjectClass *class, void *data)
2539 {
2540     DeviceClass *dc = DEVICE_CLASS(class);
2541     PCIDeviceClass *c = PCI_DEVICE_CLASS(class);
2542     VMXNET3Class *vc = VMXNET3_DEVICE_CLASS(class);
2543 
2544     c->realize = vmxnet3_pci_realize;
2545     c->exit = vmxnet3_pci_uninit;
2546     c->vendor_id = PCI_VENDOR_ID_VMWARE;
2547     c->device_id = PCI_DEVICE_ID_VMWARE_VMXNET3;
2548     c->revision = PCI_DEVICE_ID_VMWARE_VMXNET3_REVISION;
2549     c->romfile = "efi-vmxnet3.rom";
2550     c->class_id = PCI_CLASS_NETWORK_ETHERNET;
2551     c->subsystem_vendor_id = PCI_VENDOR_ID_VMWARE;
2552     c->subsystem_id = PCI_DEVICE_ID_VMWARE_VMXNET3;
2553     device_class_set_parent_realize(dc, vmxnet3_realize,
2554                                     &vc->parent_dc_realize);
2555     dc->desc = "VMWare Paravirtualized Ethernet v3";
2556     dc->reset = vmxnet3_qdev_reset;
2557     dc->vmsd = &vmstate_vmxnet3;
2558     dc->props = vmxnet3_properties;
2559     set_bit(DEVICE_CATEGORY_NETWORK, dc->categories);
2560 }
2561 
2562 static const TypeInfo vmxnet3_info = {
2563     .name          = TYPE_VMXNET3,
2564     .parent        = TYPE_PCI_DEVICE,
2565     .class_size    = sizeof(VMXNET3Class),
2566     .instance_size = sizeof(VMXNET3State),
2567     .class_init    = vmxnet3_class_init,
2568     .instance_init = vmxnet3_instance_init,
2569     .interfaces = (InterfaceInfo[]) {
2570         { INTERFACE_PCIE_DEVICE },
2571         { INTERFACE_CONVENTIONAL_PCI_DEVICE },
2572         { }
2573     },
2574 };
2575 
2576 static void vmxnet3_register_types(void)
2577 {
2578     VMW_CBPRN("vmxnet3_register_types called...");
2579     type_register_static(&vmxnet3_info);
2580 }
2581 
2582 type_init(vmxnet3_register_types)
2583