xref: /openbmc/qemu/hw/net/vmxnet3.c (revision cde3c425)
1 /*
2  * QEMU VMWARE VMXNET3 paravirtual NIC
3  *
4  * Copyright (c) 2012 Ravello Systems LTD (http://ravellosystems.com)
5  *
6  * Developed by Daynix Computing LTD (http://www.daynix.com)
7  *
8  * Authors:
9  * Dmitry Fleytman <dmitry@daynix.com>
10  * Tamir Shomer <tamirs@daynix.com>
11  * Yan Vugenfirer <yan@daynix.com>
12  *
13  * This work is licensed under the terms of the GNU GPL, version 2.
14  * See the COPYING file in the top-level directory.
15  *
16  */
17 
18 #include "qemu/osdep.h"
19 #include "hw/hw.h"
20 #include "hw/pci/pci.h"
21 #include "hw/qdev-properties.h"
22 #include "net/tap.h"
23 #include "net/checksum.h"
24 #include "sysemu/sysemu.h"
25 #include "qemu/bswap.h"
26 #include "qemu/log.h"
27 #include "qemu/module.h"
28 #include "hw/pci/msix.h"
29 #include "hw/pci/msi.h"
30 #include "migration/register.h"
31 #include "migration/vmstate.h"
32 
33 #include "vmxnet3.h"
34 #include "vmxnet3_defs.h"
35 #include "vmxnet_debug.h"
36 #include "vmware_utils.h"
37 #include "net_tx_pkt.h"
38 #include "net_rx_pkt.h"
39 #include "qom/object.h"
40 
41 #define PCI_DEVICE_ID_VMWARE_VMXNET3_REVISION 0x1
42 #define VMXNET3_MSIX_BAR_SIZE 0x2000
43 
44 /* Compatibility flags for migration */
45 #define VMXNET3_COMPAT_FLAG_OLD_MSI_OFFSETS_BIT 0
46 #define VMXNET3_COMPAT_FLAG_OLD_MSI_OFFSETS \
47     (1 << VMXNET3_COMPAT_FLAG_OLD_MSI_OFFSETS_BIT)
48 #define VMXNET3_COMPAT_FLAG_DISABLE_PCIE_BIT 1
49 #define VMXNET3_COMPAT_FLAG_DISABLE_PCIE \
50     (1 << VMXNET3_COMPAT_FLAG_DISABLE_PCIE_BIT)
51 
52 #define VMXNET3_EXP_EP_OFFSET (0x48)
53 #define VMXNET3_MSI_OFFSET(s) \
54     ((s)->compat_flags & VMXNET3_COMPAT_FLAG_OLD_MSI_OFFSETS ? 0x50 : 0x84)
55 #define VMXNET3_MSIX_OFFSET(s) \
56     ((s)->compat_flags & VMXNET3_COMPAT_FLAG_OLD_MSI_OFFSETS ? 0 : 0x9c)
57 #define VMXNET3_DSN_OFFSET     (0x100)
58 
59 #define VMXNET3_BAR0_IDX      (0)
60 #define VMXNET3_BAR1_IDX      (1)
61 #define VMXNET3_MSIX_BAR_IDX  (2)
62 
63 #define VMXNET3_OFF_MSIX_TABLE (0x000)
64 #define VMXNET3_OFF_MSIX_PBA(s) \
65     ((s)->compat_flags & VMXNET3_COMPAT_FLAG_OLD_MSI_OFFSETS ? 0x800 : 0x1000)
66 
67 /* Link speed in Mbps should be shifted by 16 */
68 #define VMXNET3_LINK_SPEED      (1000 << 16)
69 
70 /* Link status: 1 - up, 0 - down. */
71 #define VMXNET3_LINK_STATUS_UP  0x1
72 
73 /* Least significant bit should be set for revision and version */
74 #define VMXNET3_UPT_REVISION      0x1
75 #define VMXNET3_DEVICE_REVISION   0x1
76 
77 /* Number of interrupt vectors for non-MSIx modes */
78 #define VMXNET3_MAX_NMSIX_INTRS   (1)
79 
80 /* Macros for rings descriptors access */
81 #define VMXNET3_READ_TX_QUEUE_DESCR8(_d, dpa, field) \
82     (vmw_shmem_ld8(_d, dpa + offsetof(struct Vmxnet3_TxQueueDesc, field)))
83 
84 #define VMXNET3_WRITE_TX_QUEUE_DESCR8(_d, dpa, field, value) \
85     (vmw_shmem_st8(_d, dpa + offsetof(struct Vmxnet3_TxQueueDesc, field, value)))
86 
87 #define VMXNET3_READ_TX_QUEUE_DESCR32(_d, dpa, field) \
88     (vmw_shmem_ld32(_d, dpa + offsetof(struct Vmxnet3_TxQueueDesc, field)))
89 
90 #define VMXNET3_WRITE_TX_QUEUE_DESCR32(_d, dpa, field, value) \
91     (vmw_shmem_st32(_d, dpa + offsetof(struct Vmxnet3_TxQueueDesc, field), value))
92 
93 #define VMXNET3_READ_TX_QUEUE_DESCR64(_d, dpa, field) \
94     (vmw_shmem_ld64(_d, dpa + offsetof(struct Vmxnet3_TxQueueDesc, field)))
95 
96 #define VMXNET3_WRITE_TX_QUEUE_DESCR64(_d, dpa, field, value) \
97     (vmw_shmem_st64(_d, dpa + offsetof(struct Vmxnet3_TxQueueDesc, field), value))
98 
99 #define VMXNET3_READ_RX_QUEUE_DESCR64(_d, dpa, field) \
100     (vmw_shmem_ld64(_d, dpa + offsetof(struct Vmxnet3_RxQueueDesc, field)))
101 
102 #define VMXNET3_READ_RX_QUEUE_DESCR32(_d, dpa, field) \
103     (vmw_shmem_ld32(_d, dpa + offsetof(struct Vmxnet3_RxQueueDesc, field)))
104 
105 #define VMXNET3_WRITE_RX_QUEUE_DESCR64(_d, dpa, field, value) \
106     (vmw_shmem_st64(_d, dpa + offsetof(struct Vmxnet3_RxQueueDesc, field), value))
107 
108 #define VMXNET3_WRITE_RX_QUEUE_DESCR8(_d, dpa, field, value) \
109     (vmw_shmem_st8(_d, dpa + offsetof(struct Vmxnet3_RxQueueDesc, field), value))
110 
111 /* Macros for guest driver shared area access */
112 #define VMXNET3_READ_DRV_SHARED64(_d, shpa, field) \
113     (vmw_shmem_ld64(_d, shpa + offsetof(struct Vmxnet3_DriverShared, field)))
114 
115 #define VMXNET3_READ_DRV_SHARED32(_d, shpa, field) \
116     (vmw_shmem_ld32(_d, shpa + offsetof(struct Vmxnet3_DriverShared, field)))
117 
118 #define VMXNET3_WRITE_DRV_SHARED32(_d, shpa, field, val) \
119     (vmw_shmem_st32(_d, shpa + offsetof(struct Vmxnet3_DriverShared, field), val))
120 
121 #define VMXNET3_READ_DRV_SHARED16(_d, shpa, field) \
122     (vmw_shmem_ld16(_d, shpa + offsetof(struct Vmxnet3_DriverShared, field)))
123 
124 #define VMXNET3_READ_DRV_SHARED8(_d, shpa, field) \
125     (vmw_shmem_ld8(_d, shpa + offsetof(struct Vmxnet3_DriverShared, field)))
126 
127 #define VMXNET3_READ_DRV_SHARED(_d, shpa, field, b, l) \
128     (vmw_shmem_read(_d, shpa + offsetof(struct Vmxnet3_DriverShared, field), b, l))
129 
130 #define VMXNET_FLAG_IS_SET(field, flag) (((field) & (flag)) == (flag))
131 
132 struct VMXNET3Class {
133     PCIDeviceClass parent_class;
134     DeviceRealize parent_dc_realize;
135 };
136 typedef struct VMXNET3Class VMXNET3Class;
137 
138 DECLARE_CLASS_CHECKERS(VMXNET3Class, VMXNET3_DEVICE,
139                        TYPE_VMXNET3)
140 
141 static inline void vmxnet3_ring_init(PCIDevice *d,
142                                      Vmxnet3Ring *ring,
143                                      hwaddr pa,
144                                      uint32_t size,
145                                      uint32_t cell_size,
146                                      bool zero_region)
147 {
148     ring->pa = pa;
149     ring->size = size;
150     ring->cell_size = cell_size;
151     ring->gen = VMXNET3_INIT_GEN;
152     ring->next = 0;
153 
154     if (zero_region) {
155         vmw_shmem_set(d, pa, 0, size * cell_size);
156     }
157 }
158 
159 #define VMXNET3_RING_DUMP(macro, ring_name, ridx, r)                         \
160     macro("%s#%d: base %" PRIx64 " size %u cell_size %u gen %d next %u",  \
161           (ring_name), (ridx),                                               \
162           (r)->pa, (r)->size, (r)->cell_size, (r)->gen, (r)->next)
163 
164 static inline void vmxnet3_ring_inc(Vmxnet3Ring *ring)
165 {
166     if (++ring->next >= ring->size) {
167         ring->next = 0;
168         ring->gen ^= 1;
169     }
170 }
171 
172 static inline void vmxnet3_ring_dec(Vmxnet3Ring *ring)
173 {
174     if (ring->next-- == 0) {
175         ring->next = ring->size - 1;
176         ring->gen ^= 1;
177     }
178 }
179 
180 static inline hwaddr vmxnet3_ring_curr_cell_pa(Vmxnet3Ring *ring)
181 {
182     return ring->pa + ring->next * ring->cell_size;
183 }
184 
185 static inline void vmxnet3_ring_read_curr_cell(PCIDevice *d, Vmxnet3Ring *ring,
186                                                void *buff)
187 {
188     vmw_shmem_read(d, vmxnet3_ring_curr_cell_pa(ring), buff, ring->cell_size);
189 }
190 
191 static inline void vmxnet3_ring_write_curr_cell(PCIDevice *d, Vmxnet3Ring *ring,
192                                                 void *buff)
193 {
194     vmw_shmem_write(d, vmxnet3_ring_curr_cell_pa(ring), buff, ring->cell_size);
195 }
196 
197 static inline size_t vmxnet3_ring_curr_cell_idx(Vmxnet3Ring *ring)
198 {
199     return ring->next;
200 }
201 
202 static inline uint8_t vmxnet3_ring_curr_gen(Vmxnet3Ring *ring)
203 {
204     return ring->gen;
205 }
206 
207 /* Debug trace-related functions */
208 static inline void
209 vmxnet3_dump_tx_descr(struct Vmxnet3_TxDesc *descr)
210 {
211     VMW_PKPRN("TX DESCR: "
212               "addr %" PRIx64 ", len: %d, gen: %d, rsvd: %d, "
213               "dtype: %d, ext1: %d, msscof: %d, hlen: %d, om: %d, "
214               "eop: %d, cq: %d, ext2: %d, ti: %d, tci: %d",
215               descr->addr, descr->len, descr->gen, descr->rsvd,
216               descr->dtype, descr->ext1, descr->msscof, descr->hlen, descr->om,
217               descr->eop, descr->cq, descr->ext2, descr->ti, descr->tci);
218 }
219 
220 static inline void
221 vmxnet3_dump_virt_hdr(struct virtio_net_hdr *vhdr)
222 {
223     VMW_PKPRN("VHDR: flags 0x%x, gso_type: 0x%x, hdr_len: %d, gso_size: %d, "
224               "csum_start: %d, csum_offset: %d",
225               vhdr->flags, vhdr->gso_type, vhdr->hdr_len, vhdr->gso_size,
226               vhdr->csum_start, vhdr->csum_offset);
227 }
228 
229 static inline void
230 vmxnet3_dump_rx_descr(struct Vmxnet3_RxDesc *descr)
231 {
232     VMW_PKPRN("RX DESCR: addr %" PRIx64 ", len: %d, gen: %d, rsvd: %d, "
233               "dtype: %d, ext1: %d, btype: %d",
234               descr->addr, descr->len, descr->gen,
235               descr->rsvd, descr->dtype, descr->ext1, descr->btype);
236 }
237 
238 /* Interrupt management */
239 
240 /*
241  * This function returns sign whether interrupt line is in asserted state
242  * This depends on the type of interrupt used. For INTX interrupt line will
243  * be asserted until explicit deassertion, for MSI(X) interrupt line will
244  * be deasserted automatically due to notification semantics of the MSI(X)
245  * interrupts
246  */
247 static bool _vmxnet3_assert_interrupt_line(VMXNET3State *s, uint32_t int_idx)
248 {
249     PCIDevice *d = PCI_DEVICE(s);
250 
251     if (s->msix_used && msix_enabled(d)) {
252         VMW_IRPRN("Sending MSI-X notification for vector %u", int_idx);
253         msix_notify(d, int_idx);
254         return false;
255     }
256     if (msi_enabled(d)) {
257         VMW_IRPRN("Sending MSI notification for vector %u", int_idx);
258         msi_notify(d, int_idx);
259         return false;
260     }
261 
262     VMW_IRPRN("Asserting line for interrupt %u", int_idx);
263     pci_irq_assert(d);
264     return true;
265 }
266 
267 static void _vmxnet3_deassert_interrupt_line(VMXNET3State *s, int lidx)
268 {
269     PCIDevice *d = PCI_DEVICE(s);
270 
271     /*
272      * This function should never be called for MSI(X) interrupts
273      * because deassertion never required for message interrupts
274      */
275     assert(!s->msix_used || !msix_enabled(d));
276     /*
277      * This function should never be called for MSI(X) interrupts
278      * because deassertion never required for message interrupts
279      */
280     assert(!msi_enabled(d));
281 
282     VMW_IRPRN("Deasserting line for interrupt %u", lidx);
283     pci_irq_deassert(d);
284 }
285 
286 static void vmxnet3_update_interrupt_line_state(VMXNET3State *s, int lidx)
287 {
288     if (!s->interrupt_states[lidx].is_pending &&
289        s->interrupt_states[lidx].is_asserted) {
290         VMW_IRPRN("New interrupt line state for index %d is DOWN", lidx);
291         _vmxnet3_deassert_interrupt_line(s, lidx);
292         s->interrupt_states[lidx].is_asserted = false;
293         return;
294     }
295 
296     if (s->interrupt_states[lidx].is_pending &&
297        !s->interrupt_states[lidx].is_masked &&
298        !s->interrupt_states[lidx].is_asserted) {
299         VMW_IRPRN("New interrupt line state for index %d is UP", lidx);
300         s->interrupt_states[lidx].is_asserted =
301             _vmxnet3_assert_interrupt_line(s, lidx);
302         s->interrupt_states[lidx].is_pending = false;
303         return;
304     }
305 }
306 
307 static void vmxnet3_trigger_interrupt(VMXNET3State *s, int lidx)
308 {
309     PCIDevice *d = PCI_DEVICE(s);
310     s->interrupt_states[lidx].is_pending = true;
311     vmxnet3_update_interrupt_line_state(s, lidx);
312 
313     if (s->msix_used && msix_enabled(d) && s->auto_int_masking) {
314         goto do_automask;
315     }
316 
317     if (msi_enabled(d) && s->auto_int_masking) {
318         goto do_automask;
319     }
320 
321     return;
322 
323 do_automask:
324     s->interrupt_states[lidx].is_masked = true;
325     vmxnet3_update_interrupt_line_state(s, lidx);
326 }
327 
328 static bool vmxnet3_interrupt_asserted(VMXNET3State *s, int lidx)
329 {
330     return s->interrupt_states[lidx].is_asserted;
331 }
332 
333 static void vmxnet3_clear_interrupt(VMXNET3State *s, int int_idx)
334 {
335     s->interrupt_states[int_idx].is_pending = false;
336     if (s->auto_int_masking) {
337         s->interrupt_states[int_idx].is_masked = true;
338     }
339     vmxnet3_update_interrupt_line_state(s, int_idx);
340 }
341 
342 static void
343 vmxnet3_on_interrupt_mask_changed(VMXNET3State *s, int lidx, bool is_masked)
344 {
345     s->interrupt_states[lidx].is_masked = is_masked;
346     vmxnet3_update_interrupt_line_state(s, lidx);
347 }
348 
349 static bool vmxnet3_verify_driver_magic(PCIDevice *d, hwaddr dshmem)
350 {
351     return (VMXNET3_READ_DRV_SHARED32(d, dshmem, magic) == VMXNET3_REV1_MAGIC);
352 }
353 
354 #define VMXNET3_GET_BYTE(x, byte_num) (((x) >> (byte_num)*8) & 0xFF)
355 #define VMXNET3_MAKE_BYTE(byte_num, val) \
356     (((uint32_t)((val) & 0xFF)) << (byte_num)*8)
357 
358 static void vmxnet3_set_variable_mac(VMXNET3State *s, uint32_t h, uint32_t l)
359 {
360     s->conf.macaddr.a[0] = VMXNET3_GET_BYTE(l,  0);
361     s->conf.macaddr.a[1] = VMXNET3_GET_BYTE(l,  1);
362     s->conf.macaddr.a[2] = VMXNET3_GET_BYTE(l,  2);
363     s->conf.macaddr.a[3] = VMXNET3_GET_BYTE(l,  3);
364     s->conf.macaddr.a[4] = VMXNET3_GET_BYTE(h, 0);
365     s->conf.macaddr.a[5] = VMXNET3_GET_BYTE(h, 1);
366 
367     VMW_CFPRN("Variable MAC: " MAC_FMT, MAC_ARG(s->conf.macaddr.a));
368 
369     qemu_format_nic_info_str(qemu_get_queue(s->nic), s->conf.macaddr.a);
370 }
371 
372 static uint64_t vmxnet3_get_mac_low(MACAddr *addr)
373 {
374     return VMXNET3_MAKE_BYTE(0, addr->a[0]) |
375            VMXNET3_MAKE_BYTE(1, addr->a[1]) |
376            VMXNET3_MAKE_BYTE(2, addr->a[2]) |
377            VMXNET3_MAKE_BYTE(3, addr->a[3]);
378 }
379 
380 static uint64_t vmxnet3_get_mac_high(MACAddr *addr)
381 {
382     return VMXNET3_MAKE_BYTE(0, addr->a[4]) |
383            VMXNET3_MAKE_BYTE(1, addr->a[5]);
384 }
385 
386 static void
387 vmxnet3_inc_tx_consumption_counter(VMXNET3State *s, int qidx)
388 {
389     vmxnet3_ring_inc(&s->txq_descr[qidx].tx_ring);
390 }
391 
392 static inline void
393 vmxnet3_inc_rx_consumption_counter(VMXNET3State *s, int qidx, int ridx)
394 {
395     vmxnet3_ring_inc(&s->rxq_descr[qidx].rx_ring[ridx]);
396 }
397 
398 static inline void
399 vmxnet3_inc_tx_completion_counter(VMXNET3State *s, int qidx)
400 {
401     vmxnet3_ring_inc(&s->txq_descr[qidx].comp_ring);
402 }
403 
404 static void
405 vmxnet3_inc_rx_completion_counter(VMXNET3State *s, int qidx)
406 {
407     vmxnet3_ring_inc(&s->rxq_descr[qidx].comp_ring);
408 }
409 
410 static void
411 vmxnet3_dec_rx_completion_counter(VMXNET3State *s, int qidx)
412 {
413     vmxnet3_ring_dec(&s->rxq_descr[qidx].comp_ring);
414 }
415 
416 static void vmxnet3_complete_packet(VMXNET3State *s, int qidx, uint32_t tx_ridx)
417 {
418     struct Vmxnet3_TxCompDesc txcq_descr;
419     PCIDevice *d = PCI_DEVICE(s);
420 
421     VMXNET3_RING_DUMP(VMW_RIPRN, "TXC", qidx, &s->txq_descr[qidx].comp_ring);
422 
423     memset(&txcq_descr, 0, sizeof(txcq_descr));
424     txcq_descr.txdIdx = tx_ridx;
425     txcq_descr.gen = vmxnet3_ring_curr_gen(&s->txq_descr[qidx].comp_ring);
426     txcq_descr.val1 = cpu_to_le32(txcq_descr.val1);
427     txcq_descr.val2 = cpu_to_le32(txcq_descr.val2);
428     vmxnet3_ring_write_curr_cell(d, &s->txq_descr[qidx].comp_ring, &txcq_descr);
429 
430     /* Flush changes in TX descriptor before changing the counter value */
431     smp_wmb();
432 
433     vmxnet3_inc_tx_completion_counter(s, qidx);
434     vmxnet3_trigger_interrupt(s, s->txq_descr[qidx].intr_idx);
435 }
436 
437 static bool
438 vmxnet3_setup_tx_offloads(VMXNET3State *s)
439 {
440     switch (s->offload_mode) {
441     case VMXNET3_OM_NONE:
442         return net_tx_pkt_build_vheader(s->tx_pkt, false, false, 0);
443 
444     case VMXNET3_OM_CSUM:
445         VMW_PKPRN("L4 CSO requested\n");
446         return net_tx_pkt_build_vheader(s->tx_pkt, false, true, 0);
447 
448     case VMXNET3_OM_TSO:
449         VMW_PKPRN("GSO offload requested.");
450         if (!net_tx_pkt_build_vheader(s->tx_pkt, true, true,
451             s->cso_or_gso_size)) {
452             return false;
453         }
454         net_tx_pkt_update_ip_checksums(s->tx_pkt);
455         break;
456 
457     default:
458         g_assert_not_reached();
459     }
460 
461     return true;
462 }
463 
464 static void
465 vmxnet3_tx_retrieve_metadata(VMXNET3State *s,
466                              const struct Vmxnet3_TxDesc *txd)
467 {
468     s->offload_mode = txd->om;
469     s->cso_or_gso_size = txd->msscof;
470     s->tci = txd->tci;
471     s->needs_vlan = txd->ti;
472 }
473 
474 typedef enum {
475     VMXNET3_PKT_STATUS_OK,
476     VMXNET3_PKT_STATUS_ERROR,
477     VMXNET3_PKT_STATUS_DISCARD,/* only for tx */
478     VMXNET3_PKT_STATUS_OUT_OF_BUF /* only for rx */
479 } Vmxnet3PktStatus;
480 
481 static void
482 vmxnet3_on_tx_done_update_stats(VMXNET3State *s, int qidx,
483     Vmxnet3PktStatus status)
484 {
485     size_t tot_len = net_tx_pkt_get_total_len(s->tx_pkt);
486     struct UPT1_TxStats *stats = &s->txq_descr[qidx].txq_stats;
487 
488     switch (status) {
489     case VMXNET3_PKT_STATUS_OK:
490         switch (net_tx_pkt_get_packet_type(s->tx_pkt)) {
491         case ETH_PKT_BCAST:
492             stats->bcastPktsTxOK++;
493             stats->bcastBytesTxOK += tot_len;
494             break;
495         case ETH_PKT_MCAST:
496             stats->mcastPktsTxOK++;
497             stats->mcastBytesTxOK += tot_len;
498             break;
499         case ETH_PKT_UCAST:
500             stats->ucastPktsTxOK++;
501             stats->ucastBytesTxOK += tot_len;
502             break;
503         default:
504             g_assert_not_reached();
505         }
506 
507         if (s->offload_mode == VMXNET3_OM_TSO) {
508             /*
509              * According to VMWARE headers this statistic is a number
510              * of packets after segmentation but since we don't have
511              * this information in QEMU model, the best we can do is to
512              * provide number of non-segmented packets
513              */
514             stats->TSOPktsTxOK++;
515             stats->TSOBytesTxOK += tot_len;
516         }
517         break;
518 
519     case VMXNET3_PKT_STATUS_DISCARD:
520         stats->pktsTxDiscard++;
521         break;
522 
523     case VMXNET3_PKT_STATUS_ERROR:
524         stats->pktsTxError++;
525         break;
526 
527     default:
528         g_assert_not_reached();
529     }
530 }
531 
532 static void
533 vmxnet3_on_rx_done_update_stats(VMXNET3State *s,
534                                 int qidx,
535                                 Vmxnet3PktStatus status)
536 {
537     struct UPT1_RxStats *stats = &s->rxq_descr[qidx].rxq_stats;
538     size_t tot_len = net_rx_pkt_get_total_len(s->rx_pkt);
539 
540     switch (status) {
541     case VMXNET3_PKT_STATUS_OUT_OF_BUF:
542         stats->pktsRxOutOfBuf++;
543         break;
544 
545     case VMXNET3_PKT_STATUS_ERROR:
546         stats->pktsRxError++;
547         break;
548     case VMXNET3_PKT_STATUS_OK:
549         switch (net_rx_pkt_get_packet_type(s->rx_pkt)) {
550         case ETH_PKT_BCAST:
551             stats->bcastPktsRxOK++;
552             stats->bcastBytesRxOK += tot_len;
553             break;
554         case ETH_PKT_MCAST:
555             stats->mcastPktsRxOK++;
556             stats->mcastBytesRxOK += tot_len;
557             break;
558         case ETH_PKT_UCAST:
559             stats->ucastPktsRxOK++;
560             stats->ucastBytesRxOK += tot_len;
561             break;
562         default:
563             g_assert_not_reached();
564         }
565 
566         if (tot_len > s->mtu) {
567             stats->LROPktsRxOK++;
568             stats->LROBytesRxOK += tot_len;
569         }
570         break;
571     default:
572         g_assert_not_reached();
573     }
574 }
575 
576 static inline void
577 vmxnet3_ring_read_curr_txdesc(PCIDevice *pcidev, Vmxnet3Ring *ring,
578                               struct Vmxnet3_TxDesc *txd)
579 {
580     vmxnet3_ring_read_curr_cell(pcidev, ring, txd);
581     txd->addr = le64_to_cpu(txd->addr);
582     txd->val1 = le32_to_cpu(txd->val1);
583     txd->val2 = le32_to_cpu(txd->val2);
584 }
585 
586 static inline bool
587 vmxnet3_pop_next_tx_descr(VMXNET3State *s,
588                           int qidx,
589                           struct Vmxnet3_TxDesc *txd,
590                           uint32_t *descr_idx)
591 {
592     Vmxnet3Ring *ring = &s->txq_descr[qidx].tx_ring;
593     PCIDevice *d = PCI_DEVICE(s);
594 
595     vmxnet3_ring_read_curr_txdesc(d, ring, txd);
596     if (txd->gen == vmxnet3_ring_curr_gen(ring)) {
597         /* Only read after generation field verification */
598         smp_rmb();
599         /* Re-read to be sure we got the latest version */
600         vmxnet3_ring_read_curr_txdesc(d, ring, txd);
601         VMXNET3_RING_DUMP(VMW_RIPRN, "TX", qidx, ring);
602         *descr_idx = vmxnet3_ring_curr_cell_idx(ring);
603         vmxnet3_inc_tx_consumption_counter(s, qidx);
604         return true;
605     }
606 
607     return false;
608 }
609 
610 static bool
611 vmxnet3_send_packet(VMXNET3State *s, uint32_t qidx)
612 {
613     Vmxnet3PktStatus status = VMXNET3_PKT_STATUS_OK;
614 
615     if (!vmxnet3_setup_tx_offloads(s)) {
616         status = VMXNET3_PKT_STATUS_ERROR;
617         goto func_exit;
618     }
619 
620     /* debug prints */
621     vmxnet3_dump_virt_hdr(net_tx_pkt_get_vhdr(s->tx_pkt));
622     net_tx_pkt_dump(s->tx_pkt);
623 
624     if (!net_tx_pkt_send(s->tx_pkt, qemu_get_queue(s->nic))) {
625         status = VMXNET3_PKT_STATUS_DISCARD;
626         goto func_exit;
627     }
628 
629 func_exit:
630     vmxnet3_on_tx_done_update_stats(s, qidx, status);
631     return (status == VMXNET3_PKT_STATUS_OK);
632 }
633 
634 static void vmxnet3_process_tx_queue(VMXNET3State *s, int qidx)
635 {
636     struct Vmxnet3_TxDesc txd;
637     uint32_t txd_idx;
638     uint32_t data_len;
639     hwaddr data_pa;
640 
641     for (;;) {
642         if (!vmxnet3_pop_next_tx_descr(s, qidx, &txd, &txd_idx)) {
643             break;
644         }
645 
646         vmxnet3_dump_tx_descr(&txd);
647 
648         if (!s->skip_current_tx_pkt) {
649             data_len = (txd.len > 0) ? txd.len : VMXNET3_MAX_TX_BUF_SIZE;
650             data_pa = txd.addr;
651 
652             if (!net_tx_pkt_add_raw_fragment_pci(s->tx_pkt, PCI_DEVICE(s),
653                                                  data_pa, 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                              net_tx_pkt_unmap_frag_pci, PCI_DEVICE(s));
680         }
681     }
682 
683     net_tx_pkt_reset(s->tx_pkt, net_tx_pkt_unmap_frag_pci, PCI_DEVICE(s));
684 }
685 
686 static inline void
687 vmxnet3_read_next_rx_descr(VMXNET3State *s, int qidx, int ridx,
688                            struct Vmxnet3_RxDesc *dbuf, uint32_t *didx)
689 {
690     PCIDevice *d = PCI_DEVICE(s);
691 
692     Vmxnet3Ring *ring = &s->rxq_descr[qidx].rx_ring[ridx];
693     *didx = vmxnet3_ring_curr_cell_idx(ring);
694     vmxnet3_ring_read_curr_cell(d, ring, dbuf);
695     dbuf->addr = le64_to_cpu(dbuf->addr);
696     dbuf->val1 = le32_to_cpu(dbuf->val1);
697     dbuf->ext1 = le32_to_cpu(dbuf->ext1);
698 }
699 
700 static inline uint8_t
701 vmxnet3_get_rx_ring_gen(VMXNET3State *s, int qidx, int ridx)
702 {
703     return s->rxq_descr[qidx].rx_ring[ridx].gen;
704 }
705 
706 static inline hwaddr
707 vmxnet3_pop_rxc_descr(VMXNET3State *s, int qidx, uint32_t *descr_gen)
708 {
709     uint8_t ring_gen;
710     struct Vmxnet3_RxCompDesc rxcd;
711 
712     hwaddr daddr =
713         vmxnet3_ring_curr_cell_pa(&s->rxq_descr[qidx].comp_ring);
714 
715     pci_dma_read(PCI_DEVICE(s),
716                  daddr, &rxcd, sizeof(struct Vmxnet3_RxCompDesc));
717     rxcd.val1 = le32_to_cpu(rxcd.val1);
718     rxcd.val2 = le32_to_cpu(rxcd.val2);
719     rxcd.val3 = le32_to_cpu(rxcd.val3);
720     ring_gen = vmxnet3_ring_curr_gen(&s->rxq_descr[qidx].comp_ring);
721 
722     if (rxcd.gen != ring_gen) {
723         *descr_gen = ring_gen;
724         vmxnet3_inc_rx_completion_counter(s, qidx);
725         return daddr;
726     }
727 
728     return 0;
729 }
730 
731 static inline void
732 vmxnet3_revert_rxc_descr(VMXNET3State *s, int qidx)
733 {
734     vmxnet3_dec_rx_completion_counter(s, qidx);
735 }
736 
737 #define RXQ_IDX      (0)
738 #define RX_HEAD_BODY_RING (0)
739 #define RX_BODY_ONLY_RING (1)
740 
741 static bool
742 vmxnet3_get_next_head_rx_descr(VMXNET3State *s,
743                                struct Vmxnet3_RxDesc *descr_buf,
744                                uint32_t *descr_idx,
745                                uint32_t *ridx)
746 {
747     for (;;) {
748         uint32_t ring_gen;
749         vmxnet3_read_next_rx_descr(s, RXQ_IDX, RX_HEAD_BODY_RING,
750                                    descr_buf, descr_idx);
751 
752         /* If no more free descriptors - return */
753         ring_gen = vmxnet3_get_rx_ring_gen(s, RXQ_IDX, RX_HEAD_BODY_RING);
754         if (descr_buf->gen != ring_gen) {
755             return false;
756         }
757 
758         /* Only read after generation field verification */
759         smp_rmb();
760         /* Re-read to be sure we got the latest version */
761         vmxnet3_read_next_rx_descr(s, RXQ_IDX, RX_HEAD_BODY_RING,
762                                    descr_buf, descr_idx);
763 
764         /* Mark current descriptor as used/skipped */
765         vmxnet3_inc_rx_consumption_counter(s, RXQ_IDX, RX_HEAD_BODY_RING);
766 
767         /* If this is what we are looking for - return */
768         if (descr_buf->btype == VMXNET3_RXD_BTYPE_HEAD) {
769             *ridx = RX_HEAD_BODY_RING;
770             return true;
771         }
772     }
773 }
774 
775 static bool
776 vmxnet3_get_next_body_rx_descr(VMXNET3State *s,
777                                struct Vmxnet3_RxDesc *d,
778                                uint32_t *didx,
779                                uint32_t *ridx)
780 {
781     vmxnet3_read_next_rx_descr(s, RXQ_IDX, RX_HEAD_BODY_RING, d, didx);
782 
783     /* Try to find corresponding descriptor in head/body ring */
784     if (d->gen == vmxnet3_get_rx_ring_gen(s, RXQ_IDX, RX_HEAD_BODY_RING)) {
785         /* Only read after generation field verification */
786         smp_rmb();
787         /* Re-read to be sure we got the latest version */
788         vmxnet3_read_next_rx_descr(s, RXQ_IDX, RX_HEAD_BODY_RING, d, didx);
789         if (d->btype == VMXNET3_RXD_BTYPE_BODY) {
790             vmxnet3_inc_rx_consumption_counter(s, RXQ_IDX, RX_HEAD_BODY_RING);
791             *ridx = RX_HEAD_BODY_RING;
792             return true;
793         }
794     }
795 
796     /*
797      * If there is no free descriptors on head/body ring or next free
798      * descriptor is a head descriptor switch to body only ring
799      */
800     vmxnet3_read_next_rx_descr(s, RXQ_IDX, RX_BODY_ONLY_RING, d, didx);
801 
802     /* If no more free descriptors - return */
803     if (d->gen == vmxnet3_get_rx_ring_gen(s, RXQ_IDX, RX_BODY_ONLY_RING)) {
804         /* Only read after generation field verification */
805         smp_rmb();
806         /* Re-read to be sure we got the latest version */
807         vmxnet3_read_next_rx_descr(s, RXQ_IDX, RX_BODY_ONLY_RING, d, didx);
808         assert(d->btype == VMXNET3_RXD_BTYPE_BODY);
809         *ridx = RX_BODY_ONLY_RING;
810         vmxnet3_inc_rx_consumption_counter(s, RXQ_IDX, RX_BODY_ONLY_RING);
811         return true;
812     }
813 
814     return false;
815 }
816 
817 static inline bool
818 vmxnet3_get_next_rx_descr(VMXNET3State *s, bool is_head,
819                           struct Vmxnet3_RxDesc *descr_buf,
820                           uint32_t *descr_idx,
821                           uint32_t *ridx)
822 {
823     if (is_head || !s->rx_packets_compound) {
824         return vmxnet3_get_next_head_rx_descr(s, descr_buf, descr_idx, ridx);
825     } else {
826         return vmxnet3_get_next_body_rx_descr(s, descr_buf, descr_idx, ridx);
827     }
828 }
829 
830 /* In case packet was csum offloaded (either NEEDS_CSUM or DATA_VALID),
831  * the implementation always passes an RxCompDesc with a "Checksum
832  * calculated and found correct" to the OS (cnc=0 and tuc=1, see
833  * vmxnet3_rx_update_descr). This emulates the observed ESXi behavior.
834  *
835  * Therefore, if packet has the NEEDS_CSUM set, we must calculate
836  * and place a fully computed checksum into the tcp/udp header.
837  * Otherwise, the OS driver will receive a checksum-correct indication
838  * (CHECKSUM_UNNECESSARY), but with the actual tcp/udp checksum field
839  * having just the pseudo header csum value.
840  *
841  * While this is not a problem if packet is destined for local delivery,
842  * in the case the host OS performs forwarding, it will forward an
843  * incorrectly checksummed packet.
844  */
845 static void vmxnet3_rx_need_csum_calculate(struct NetRxPkt *pkt,
846                                            const void *pkt_data,
847                                            size_t pkt_len)
848 {
849     struct virtio_net_hdr *vhdr;
850     bool hasip4, hasip6;
851     EthL4HdrProto l4hdr_proto;
852     uint8_t *data;
853     int len;
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, &hasip4, &hasip6, &l4hdr_proto);
861     if (!(hasip4 || hasip6) ||
862         (l4hdr_proto != ETH_L4_HDR_PROTO_TCP &&
863          l4hdr_proto != ETH_L4_HDR_PROTO_UDP)) {
864         return;
865     }
866 
867     vmxnet3_dump_virt_hdr(vhdr);
868 
869     /* Validate packet len: csum_start + scum_offset + length of csum field */
870     if (pkt_len < (vhdr->csum_start + vhdr->csum_offset + 2)) {
871         VMW_PKPRN("packet len:%zu < csum_start(%d) + csum_offset(%d) + 2, "
872                   "cannot calculate checksum",
873                   pkt_len, vhdr->csum_start, vhdr->csum_offset);
874         return;
875     }
876 
877     data = (uint8_t *)pkt_data + vhdr->csum_start;
878     len = pkt_len - vhdr->csum_start;
879     /* Put the checksum obtained into the packet */
880     stw_be_p(data + vhdr->csum_offset,
881              net_checksum_finish_nozero(net_checksum_add(len, data)));
882 
883     vhdr->flags &= ~VIRTIO_NET_HDR_F_NEEDS_CSUM;
884     vhdr->flags |= VIRTIO_NET_HDR_F_DATA_VALID;
885 }
886 
887 static void vmxnet3_rx_update_descr(struct NetRxPkt *pkt,
888     struct Vmxnet3_RxCompDesc *rxcd)
889 {
890     int csum_ok, is_gso;
891     bool hasip4, hasip6;
892     EthL4HdrProto l4hdr_proto;
893     struct virtio_net_hdr *vhdr;
894     uint8_t offload_type;
895 
896     if (net_rx_pkt_is_vlan_stripped(pkt)) {
897         rxcd->ts = 1;
898         rxcd->tci = net_rx_pkt_get_vlan_tag(pkt);
899     }
900 
901     vhdr = net_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     net_rx_pkt_get_protocols(pkt, &hasip4, &hasip6, &l4hdr_proto);
919     if ((l4hdr_proto != ETH_L4_HDR_PROTO_TCP &&
920          l4hdr_proto != ETH_L4_HDR_PROTO_UDP) ||
921         (!hasip4 && !hasip6)) {
922         goto nocsum;
923     }
924 
925     rxcd->cnc = 0;
926     rxcd->v4 = hasip4 ? 1 : 0;
927     rxcd->v6 = hasip6 ? 1 : 0;
928     rxcd->tcp = l4hdr_proto == ETH_L4_HDR_PROTO_TCP;
929     rxcd->udp = l4hdr_proto == ETH_L4_HDR_PROTO_UDP;
930     rxcd->fcs = rxcd->tuc = rxcd->ipc = 1;
931     return;
932 
933 nocsum:
934     rxcd->cnc = 1;
935     return;
936 }
937 
938 static void
939 vmxnet3_pci_dma_writev(PCIDevice *pci_dev,
940                        const struct iovec *iov,
941                        size_t start_iov_off,
942                        hwaddr target_addr,
943                        size_t bytes_to_copy)
944 {
945     size_t curr_off = 0;
946     size_t copied = 0;
947 
948     while (bytes_to_copy) {
949         if (start_iov_off < (curr_off + iov->iov_len)) {
950             size_t chunk_len =
951                 MIN((curr_off + iov->iov_len) - start_iov_off, bytes_to_copy);
952 
953             pci_dma_write(pci_dev, target_addr + copied,
954                           iov->iov_base + start_iov_off - curr_off,
955                           chunk_len);
956 
957             copied += chunk_len;
958             start_iov_off += chunk_len;
959             curr_off = start_iov_off;
960             bytes_to_copy -= chunk_len;
961         } else {
962             curr_off += iov->iov_len;
963         }
964         iov++;
965     }
966 }
967 
968 static void
969 vmxnet3_pci_dma_write_rxcd(PCIDevice *pcidev, dma_addr_t pa,
970                            struct Vmxnet3_RxCompDesc *rxcd)
971 {
972     rxcd->val1 = cpu_to_le32(rxcd->val1);
973     rxcd->val2 = cpu_to_le32(rxcd->val2);
974     rxcd->val3 = cpu_to_le32(rxcd->val3);
975     pci_dma_write(pcidev, pa, rxcd, sizeof(*rxcd));
976 }
977 
978 static bool
979 vmxnet3_indicate_packet(VMXNET3State *s)
980 {
981     struct Vmxnet3_RxDesc rxd;
982     PCIDevice *d = PCI_DEVICE(s);
983     bool is_head = true;
984     uint32_t rxd_idx;
985     uint32_t rx_ridx = 0;
986 
987     struct Vmxnet3_RxCompDesc rxcd;
988     uint32_t new_rxcd_gen = VMXNET3_INIT_GEN;
989     hwaddr new_rxcd_pa = 0;
990     hwaddr ready_rxcd_pa = 0;
991     struct iovec *data = net_rx_pkt_get_iovec(s->rx_pkt);
992     size_t bytes_copied = 0;
993     size_t bytes_left = net_rx_pkt_get_total_len(s->rx_pkt);
994     uint16_t num_frags = 0;
995     size_t chunk_size;
996 
997     net_rx_pkt_dump(s->rx_pkt);
998 
999     while (bytes_left > 0) {
1000 
1001         /* cannot add more frags to packet */
1002         if (num_frags == s->max_rx_frags) {
1003             break;
1004         }
1005 
1006         new_rxcd_pa = vmxnet3_pop_rxc_descr(s, RXQ_IDX, &new_rxcd_gen);
1007         if (!new_rxcd_pa) {
1008             break;
1009         }
1010 
1011         if (!vmxnet3_get_next_rx_descr(s, is_head, &rxd, &rxd_idx, &rx_ridx)) {
1012             break;
1013         }
1014 
1015         chunk_size = MIN(bytes_left, rxd.len);
1016         vmxnet3_pci_dma_writev(d, data, bytes_copied, rxd.addr, chunk_size);
1017         bytes_copied += chunk_size;
1018         bytes_left -= chunk_size;
1019 
1020         vmxnet3_dump_rx_descr(&rxd);
1021 
1022         if (ready_rxcd_pa != 0) {
1023             vmxnet3_pci_dma_write_rxcd(d, ready_rxcd_pa, &rxcd);
1024         }
1025 
1026         memset(&rxcd, 0, sizeof(struct Vmxnet3_RxCompDesc));
1027         rxcd.rxdIdx = rxd_idx;
1028         rxcd.len = chunk_size;
1029         rxcd.sop = is_head;
1030         rxcd.gen = new_rxcd_gen;
1031         rxcd.rqID = RXQ_IDX + rx_ridx * s->rxq_num;
1032 
1033         if (bytes_left == 0) {
1034             vmxnet3_rx_update_descr(s->rx_pkt, &rxcd);
1035         }
1036 
1037         VMW_RIPRN("RX Completion descriptor: rxRing: %lu rxIdx %lu len %lu "
1038                   "sop %d csum_correct %lu",
1039                   (unsigned long) rx_ridx,
1040                   (unsigned long) rxcd.rxdIdx,
1041                   (unsigned long) rxcd.len,
1042                   (int) rxcd.sop,
1043                   (unsigned long) rxcd.tuc);
1044 
1045         is_head = false;
1046         ready_rxcd_pa = new_rxcd_pa;
1047         new_rxcd_pa = 0;
1048         num_frags++;
1049     }
1050 
1051     if (ready_rxcd_pa != 0) {
1052         rxcd.eop = 1;
1053         rxcd.err = (bytes_left != 0);
1054 
1055         vmxnet3_pci_dma_write_rxcd(d, ready_rxcd_pa, &rxcd);
1056 
1057         /* Flush RX descriptor changes */
1058         smp_wmb();
1059     }
1060 
1061     if (new_rxcd_pa != 0) {
1062         vmxnet3_revert_rxc_descr(s, RXQ_IDX);
1063     }
1064 
1065     vmxnet3_trigger_interrupt(s, s->rxq_descr[RXQ_IDX].intr_idx);
1066 
1067     if (bytes_left == 0) {
1068         vmxnet3_on_rx_done_update_stats(s, RXQ_IDX, VMXNET3_PKT_STATUS_OK);
1069         return true;
1070     } else if (num_frags == s->max_rx_frags) {
1071         vmxnet3_on_rx_done_update_stats(s, RXQ_IDX, VMXNET3_PKT_STATUS_ERROR);
1072         return false;
1073     } else {
1074         vmxnet3_on_rx_done_update_stats(s, RXQ_IDX,
1075                                         VMXNET3_PKT_STATUS_OUT_OF_BUF);
1076         return false;
1077     }
1078 }
1079 
1080 static void
1081 vmxnet3_io_bar0_write(void *opaque, hwaddr addr,
1082                       uint64_t val, unsigned size)
1083 {
1084     VMXNET3State *s = opaque;
1085 
1086     if (!s->device_active) {
1087         return;
1088     }
1089 
1090     if (VMW_IS_MULTIREG_ADDR(addr, VMXNET3_REG_TXPROD,
1091                         VMXNET3_DEVICE_MAX_TX_QUEUES, VMXNET3_REG_ALIGN)) {
1092         int tx_queue_idx =
1093             VMW_MULTIREG_IDX_BY_ADDR(addr, VMXNET3_REG_TXPROD,
1094                                      VMXNET3_REG_ALIGN);
1095         if (tx_queue_idx <= s->txq_num) {
1096             vmxnet3_process_tx_queue(s, tx_queue_idx);
1097         } else {
1098             qemu_log_mask(LOG_GUEST_ERROR, "vmxnet3: Illegal TX queue %d/%d\n",
1099                           tx_queue_idx, s->txq_num);
1100         }
1101         return;
1102     }
1103 
1104     if (VMW_IS_MULTIREG_ADDR(addr, VMXNET3_REG_IMR,
1105                         VMXNET3_MAX_INTRS, VMXNET3_REG_ALIGN)) {
1106         int l = VMW_MULTIREG_IDX_BY_ADDR(addr, VMXNET3_REG_IMR,
1107                                          VMXNET3_REG_ALIGN);
1108 
1109         VMW_CBPRN("Interrupt mask for line %d written: 0x%" PRIx64, l, val);
1110 
1111         vmxnet3_on_interrupt_mask_changed(s, l, val);
1112         return;
1113     }
1114 
1115     if (VMW_IS_MULTIREG_ADDR(addr, VMXNET3_REG_RXPROD,
1116                         VMXNET3_DEVICE_MAX_RX_QUEUES, VMXNET3_REG_ALIGN) ||
1117        VMW_IS_MULTIREG_ADDR(addr, VMXNET3_REG_RXPROD2,
1118                         VMXNET3_DEVICE_MAX_RX_QUEUES, VMXNET3_REG_ALIGN)) {
1119         return;
1120     }
1121 
1122     VMW_WRPRN("BAR0 unknown write [%" PRIx64 "] = %" PRIx64 ", size %d",
1123               (uint64_t) addr, val, size);
1124 }
1125 
1126 static uint64_t
1127 vmxnet3_io_bar0_read(void *opaque, hwaddr addr, unsigned size)
1128 {
1129     VMXNET3State *s = opaque;
1130 
1131     if (VMW_IS_MULTIREG_ADDR(addr, VMXNET3_REG_IMR,
1132                         VMXNET3_MAX_INTRS, VMXNET3_REG_ALIGN)) {
1133         int l = VMW_MULTIREG_IDX_BY_ADDR(addr, VMXNET3_REG_IMR,
1134                                          VMXNET3_REG_ALIGN);
1135         return s->interrupt_states[l].is_masked;
1136     }
1137 
1138     VMW_CBPRN("BAR0 unknown read [%" PRIx64 "], size %d", addr, size);
1139     return 0;
1140 }
1141 
1142 static void vmxnet3_reset_interrupt_states(VMXNET3State *s)
1143 {
1144     int i;
1145     for (i = 0; i < ARRAY_SIZE(s->interrupt_states); i++) {
1146         s->interrupt_states[i].is_asserted = false;
1147         s->interrupt_states[i].is_pending = false;
1148         s->interrupt_states[i].is_masked = true;
1149     }
1150 }
1151 
1152 static void vmxnet3_reset_mac(VMXNET3State *s)
1153 {
1154     memcpy(&s->conf.macaddr.a, &s->perm_mac.a, sizeof(s->perm_mac.a));
1155     VMW_CFPRN("MAC address set to: " MAC_FMT, MAC_ARG(s->conf.macaddr.a));
1156 }
1157 
1158 static void vmxnet3_deactivate_device(VMXNET3State *s)
1159 {
1160     if (s->device_active) {
1161         VMW_CBPRN("Deactivating vmxnet3...");
1162         net_tx_pkt_uninit(s->tx_pkt);
1163         net_rx_pkt_uninit(s->rx_pkt);
1164         s->device_active = false;
1165     }
1166 }
1167 
1168 static void vmxnet3_reset(VMXNET3State *s)
1169 {
1170     VMW_CBPRN("Resetting vmxnet3...");
1171 
1172     vmxnet3_deactivate_device(s);
1173     vmxnet3_reset_interrupt_states(s);
1174     s->drv_shmem = 0;
1175     s->tx_sop = true;
1176     s->skip_current_tx_pkt = false;
1177 }
1178 
1179 static void vmxnet3_update_rx_mode(VMXNET3State *s)
1180 {
1181     PCIDevice *d = PCI_DEVICE(s);
1182 
1183     s->rx_mode = VMXNET3_READ_DRV_SHARED32(d, s->drv_shmem,
1184                                            devRead.rxFilterConf.rxMode);
1185     VMW_CFPRN("RX mode: 0x%08X", s->rx_mode);
1186 }
1187 
1188 static void vmxnet3_update_vlan_filters(VMXNET3State *s)
1189 {
1190     int i;
1191     PCIDevice *d = PCI_DEVICE(s);
1192 
1193     /* Copy configuration from shared memory */
1194     VMXNET3_READ_DRV_SHARED(d, s->drv_shmem,
1195                             devRead.rxFilterConf.vfTable,
1196                             s->vlan_table,
1197                             sizeof(s->vlan_table));
1198 
1199     /* Invert byte order when needed */
1200     for (i = 0; i < ARRAY_SIZE(s->vlan_table); i++) {
1201         s->vlan_table[i] = le32_to_cpu(s->vlan_table[i]);
1202     }
1203 
1204     /* Dump configuration for debugging purposes */
1205     VMW_CFPRN("Configured VLANs:");
1206     for (i = 0; i < sizeof(s->vlan_table) * 8; i++) {
1207         if (VMXNET3_VFTABLE_ENTRY_IS_SET(s->vlan_table, i)) {
1208             VMW_CFPRN("\tVLAN %d is present", i);
1209         }
1210     }
1211 }
1212 
1213 static void vmxnet3_update_mcast_filters(VMXNET3State *s)
1214 {
1215     PCIDevice *d = PCI_DEVICE(s);
1216 
1217     uint16_t list_bytes =
1218         VMXNET3_READ_DRV_SHARED16(d, s->drv_shmem,
1219                                   devRead.rxFilterConf.mfTableLen);
1220 
1221     s->mcast_list_len = list_bytes / sizeof(s->mcast_list[0]);
1222 
1223     s->mcast_list = g_realloc(s->mcast_list, list_bytes);
1224     if (!s->mcast_list) {
1225         if (s->mcast_list_len == 0) {
1226             VMW_CFPRN("Current multicast list is empty");
1227         } else {
1228             VMW_ERPRN("Failed to allocate multicast list of %d elements",
1229                       s->mcast_list_len);
1230         }
1231         s->mcast_list_len = 0;
1232     } else {
1233         int i;
1234         hwaddr mcast_list_pa =
1235             VMXNET3_READ_DRV_SHARED64(d, s->drv_shmem,
1236                                       devRead.rxFilterConf.mfTablePA);
1237 
1238         pci_dma_read(d, mcast_list_pa, s->mcast_list, list_bytes);
1239 
1240         VMW_CFPRN("Current multicast list len is %d:", s->mcast_list_len);
1241         for (i = 0; i < s->mcast_list_len; i++) {
1242             VMW_CFPRN("\t" MAC_FMT, MAC_ARG(s->mcast_list[i].a));
1243         }
1244     }
1245 }
1246 
1247 static void vmxnet3_setup_rx_filtering(VMXNET3State *s)
1248 {
1249     vmxnet3_update_rx_mode(s);
1250     vmxnet3_update_vlan_filters(s);
1251     vmxnet3_update_mcast_filters(s);
1252 }
1253 
1254 static uint32_t vmxnet3_get_interrupt_config(VMXNET3State *s)
1255 {
1256     uint32_t interrupt_mode = VMXNET3_IT_AUTO | (VMXNET3_IMM_AUTO << 2);
1257     VMW_CFPRN("Interrupt config is 0x%X", interrupt_mode);
1258     return interrupt_mode;
1259 }
1260 
1261 static void vmxnet3_fill_stats(VMXNET3State *s)
1262 {
1263     int i;
1264     PCIDevice *d = PCI_DEVICE(s);
1265 
1266     if (!s->device_active)
1267         return;
1268 
1269     for (i = 0; i < s->txq_num; i++) {
1270         pci_dma_write(d,
1271                       s->txq_descr[i].tx_stats_pa,
1272                       &s->txq_descr[i].txq_stats,
1273                       sizeof(s->txq_descr[i].txq_stats));
1274     }
1275 
1276     for (i = 0; i < s->rxq_num; i++) {
1277         pci_dma_write(d,
1278                       s->rxq_descr[i].rx_stats_pa,
1279                       &s->rxq_descr[i].rxq_stats,
1280                       sizeof(s->rxq_descr[i].rxq_stats));
1281     }
1282 }
1283 
1284 static void vmxnet3_adjust_by_guest_type(VMXNET3State *s)
1285 {
1286     struct Vmxnet3_GOSInfo gos;
1287     PCIDevice *d = PCI_DEVICE(s);
1288 
1289     VMXNET3_READ_DRV_SHARED(d, s->drv_shmem, devRead.misc.driverInfo.gos,
1290                             &gos, sizeof(gos));
1291     s->rx_packets_compound =
1292         (gos.gosType == VMXNET3_GOS_TYPE_WIN) ? false : true;
1293 
1294     VMW_CFPRN("Guest type specifics: RXCOMPOUND: %d", s->rx_packets_compound);
1295 }
1296 
1297 static void
1298 vmxnet3_dump_conf_descr(const char *name,
1299                         struct Vmxnet3_VariableLenConfDesc *pm_descr)
1300 {
1301     VMW_CFPRN("%s descriptor dump: Version %u, Length %u",
1302               name, pm_descr->confVer, pm_descr->confLen);
1303 
1304 };
1305 
1306 static void vmxnet3_update_pm_state(VMXNET3State *s)
1307 {
1308     struct Vmxnet3_VariableLenConfDesc pm_descr;
1309     PCIDevice *d = PCI_DEVICE(s);
1310 
1311     pm_descr.confLen =
1312         VMXNET3_READ_DRV_SHARED32(d, s->drv_shmem, devRead.pmConfDesc.confLen);
1313     pm_descr.confVer =
1314         VMXNET3_READ_DRV_SHARED32(d, s->drv_shmem, devRead.pmConfDesc.confVer);
1315     pm_descr.confPA =
1316         VMXNET3_READ_DRV_SHARED64(d, s->drv_shmem, devRead.pmConfDesc.confPA);
1317 
1318     vmxnet3_dump_conf_descr("PM State", &pm_descr);
1319 }
1320 
1321 static void vmxnet3_update_features(VMXNET3State *s)
1322 {
1323     uint32_t guest_features;
1324     int rxcso_supported;
1325     PCIDevice *d = PCI_DEVICE(s);
1326 
1327     guest_features = VMXNET3_READ_DRV_SHARED32(d, s->drv_shmem,
1328                                                devRead.misc.uptFeatures);
1329 
1330     rxcso_supported = VMXNET_FLAG_IS_SET(guest_features, UPT1_F_RXCSUM);
1331     s->rx_vlan_stripping = VMXNET_FLAG_IS_SET(guest_features, UPT1_F_RXVLAN);
1332     s->lro_supported = VMXNET_FLAG_IS_SET(guest_features, UPT1_F_LRO);
1333 
1334     VMW_CFPRN("Features configuration: LRO: %d, RXCSUM: %d, VLANSTRIP: %d",
1335               s->lro_supported, rxcso_supported,
1336               s->rx_vlan_stripping);
1337     if (s->peer_has_vhdr) {
1338         qemu_set_offload(qemu_get_queue(s->nic)->peer,
1339                          rxcso_supported,
1340                          s->lro_supported,
1341                          s->lro_supported,
1342                          0,
1343                          0,
1344                          0,
1345                          0);
1346     }
1347 }
1348 
1349 static bool vmxnet3_verify_intx(VMXNET3State *s, int intx)
1350 {
1351     return s->msix_used || msi_enabled(PCI_DEVICE(s))
1352         || intx == pci_get_byte(s->parent_obj.config + PCI_INTERRUPT_PIN) - 1;
1353 }
1354 
1355 static void vmxnet3_validate_interrupt_idx(bool is_msix, int idx)
1356 {
1357     int max_ints = is_msix ? VMXNET3_MAX_INTRS : VMXNET3_MAX_NMSIX_INTRS;
1358     if (idx >= max_ints) {
1359         hw_error("Bad interrupt index: %d\n", idx);
1360     }
1361 }
1362 
1363 static void vmxnet3_validate_interrupts(VMXNET3State *s)
1364 {
1365     int i;
1366 
1367     VMW_CFPRN("Verifying event interrupt index (%d)", s->event_int_idx);
1368     vmxnet3_validate_interrupt_idx(s->msix_used, s->event_int_idx);
1369 
1370     for (i = 0; i < s->txq_num; i++) {
1371         int idx = s->txq_descr[i].intr_idx;
1372         VMW_CFPRN("Verifying TX queue %d interrupt index (%d)", i, idx);
1373         vmxnet3_validate_interrupt_idx(s->msix_used, idx);
1374     }
1375 
1376     for (i = 0; i < s->rxq_num; i++) {
1377         int idx = s->rxq_descr[i].intr_idx;
1378         VMW_CFPRN("Verifying RX queue %d interrupt index (%d)", i, idx);
1379         vmxnet3_validate_interrupt_idx(s->msix_used, idx);
1380     }
1381 }
1382 
1383 static bool vmxnet3_validate_queues(VMXNET3State *s)
1384 {
1385     /*
1386     * txq_num and rxq_num are total number of queues
1387     * configured by guest. These numbers must not
1388     * exceed corresponding maximal values.
1389     */
1390 
1391     if (s->txq_num > VMXNET3_DEVICE_MAX_TX_QUEUES) {
1392         qemu_log_mask(LOG_GUEST_ERROR, "vmxnet3: Bad TX queues number: %d\n",
1393                       s->txq_num);
1394         return false;
1395     }
1396 
1397     if (s->rxq_num > VMXNET3_DEVICE_MAX_RX_QUEUES) {
1398         qemu_log_mask(LOG_GUEST_ERROR, "vmxnet3: Bad RX queues number: %d\n",
1399                       s->rxq_num);
1400         return false;
1401     }
1402 
1403     return true;
1404 }
1405 
1406 static void vmxnet3_activate_device(VMXNET3State *s)
1407 {
1408     int i;
1409     static const uint32_t VMXNET3_DEF_TX_THRESHOLD = 1;
1410     PCIDevice *d = PCI_DEVICE(s);
1411     hwaddr qdescr_table_pa;
1412     uint64_t pa;
1413     uint32_t size;
1414 
1415     /* Verify configuration consistency */
1416     if (!vmxnet3_verify_driver_magic(d, s->drv_shmem)) {
1417         VMW_ERPRN("Device configuration received from driver is invalid");
1418         return;
1419     }
1420 
1421     /* Verify if device is active */
1422     if (s->device_active) {
1423         VMW_CFPRN("Vmxnet3 device is active");
1424         return;
1425     }
1426 
1427     s->txq_num =
1428         VMXNET3_READ_DRV_SHARED8(d, s->drv_shmem, devRead.misc.numTxQueues);
1429     s->rxq_num =
1430         VMXNET3_READ_DRV_SHARED8(d, s->drv_shmem, devRead.misc.numRxQueues);
1431 
1432     VMW_CFPRN("Number of TX/RX queues %u/%u", s->txq_num, s->rxq_num);
1433     if (!vmxnet3_validate_queues(s)) {
1434         return;
1435     }
1436 
1437     vmxnet3_adjust_by_guest_type(s);
1438     vmxnet3_update_features(s);
1439     vmxnet3_update_pm_state(s);
1440     vmxnet3_setup_rx_filtering(s);
1441     /* Cache fields from shared memory */
1442     s->mtu = VMXNET3_READ_DRV_SHARED32(d, s->drv_shmem, devRead.misc.mtu);
1443     if (s->mtu < VMXNET3_MIN_MTU || s->mtu > VMXNET3_MAX_MTU) {
1444         qemu_log_mask(LOG_GUEST_ERROR, "vmxnet3: Bad MTU size: %u\n", s->mtu);
1445         return;
1446     }
1447     VMW_CFPRN("MTU is %u", s->mtu);
1448 
1449     s->max_rx_frags =
1450         VMXNET3_READ_DRV_SHARED16(d, s->drv_shmem, devRead.misc.maxNumRxSG);
1451 
1452     if (s->max_rx_frags == 0) {
1453         s->max_rx_frags = 1;
1454     }
1455 
1456     VMW_CFPRN("Max RX fragments is %u", s->max_rx_frags);
1457 
1458     s->event_int_idx =
1459         VMXNET3_READ_DRV_SHARED8(d, s->drv_shmem, devRead.intrConf.eventIntrIdx);
1460     assert(vmxnet3_verify_intx(s, s->event_int_idx));
1461     VMW_CFPRN("Events interrupt line is %u", s->event_int_idx);
1462 
1463     s->auto_int_masking =
1464         VMXNET3_READ_DRV_SHARED8(d, s->drv_shmem, devRead.intrConf.autoMask);
1465     VMW_CFPRN("Automatic interrupt masking is %d", (int)s->auto_int_masking);
1466 
1467     qdescr_table_pa =
1468         VMXNET3_READ_DRV_SHARED64(d, s->drv_shmem, devRead.misc.queueDescPA);
1469     VMW_CFPRN("TX queues descriptors table is at 0x%" PRIx64, qdescr_table_pa);
1470 
1471     /*
1472      * Worst-case scenario is a packet that holds all TX rings space so
1473      * we calculate total size of all TX rings for max TX fragments number
1474      */
1475     s->max_tx_frags = 0;
1476 
1477     /* TX queues */
1478     for (i = 0; i < s->txq_num; i++) {
1479         hwaddr qdescr_pa =
1480             qdescr_table_pa + i * sizeof(struct Vmxnet3_TxQueueDesc);
1481 
1482         /* Read interrupt number for this TX queue */
1483         s->txq_descr[i].intr_idx =
1484             VMXNET3_READ_TX_QUEUE_DESCR8(d, qdescr_pa, conf.intrIdx);
1485         assert(vmxnet3_verify_intx(s, s->txq_descr[i].intr_idx));
1486 
1487         VMW_CFPRN("TX Queue %d interrupt: %d", i, s->txq_descr[i].intr_idx);
1488 
1489         /* Read rings memory locations for TX queues */
1490         pa = VMXNET3_READ_TX_QUEUE_DESCR64(d, qdescr_pa, conf.txRingBasePA);
1491         size = VMXNET3_READ_TX_QUEUE_DESCR32(d, qdescr_pa, conf.txRingSize);
1492         if (size > VMXNET3_TX_RING_MAX_SIZE) {
1493             size = VMXNET3_TX_RING_MAX_SIZE;
1494         }
1495 
1496         vmxnet3_ring_init(d, &s->txq_descr[i].tx_ring, pa, size,
1497                           sizeof(struct Vmxnet3_TxDesc), false);
1498         VMXNET3_RING_DUMP(VMW_CFPRN, "TX", i, &s->txq_descr[i].tx_ring);
1499 
1500         s->max_tx_frags += size;
1501 
1502         /* TXC ring */
1503         pa = VMXNET3_READ_TX_QUEUE_DESCR64(d, qdescr_pa, conf.compRingBasePA);
1504         size = VMXNET3_READ_TX_QUEUE_DESCR32(d, qdescr_pa, conf.compRingSize);
1505         if (size > VMXNET3_TC_RING_MAX_SIZE) {
1506             size = VMXNET3_TC_RING_MAX_SIZE;
1507         }
1508         vmxnet3_ring_init(d, &s->txq_descr[i].comp_ring, pa, size,
1509                           sizeof(struct Vmxnet3_TxCompDesc), true);
1510         VMXNET3_RING_DUMP(VMW_CFPRN, "TXC", i, &s->txq_descr[i].comp_ring);
1511 
1512         s->txq_descr[i].tx_stats_pa =
1513             qdescr_pa + offsetof(struct Vmxnet3_TxQueueDesc, stats);
1514 
1515         memset(&s->txq_descr[i].txq_stats, 0,
1516                sizeof(s->txq_descr[i].txq_stats));
1517 
1518         /* Fill device-managed parameters for queues */
1519         VMXNET3_WRITE_TX_QUEUE_DESCR32(d, qdescr_pa,
1520                                        ctrl.txThreshold,
1521                                        VMXNET3_DEF_TX_THRESHOLD);
1522     }
1523 
1524     /* Preallocate TX packet wrapper */
1525     VMW_CFPRN("Max TX fragments is %u", s->max_tx_frags);
1526     net_tx_pkt_init(&s->tx_pkt, s->max_tx_frags);
1527     net_rx_pkt_init(&s->rx_pkt);
1528 
1529     /* Read rings memory locations for RX queues */
1530     for (i = 0; i < s->rxq_num; i++) {
1531         int j;
1532         hwaddr qd_pa =
1533             qdescr_table_pa + s->txq_num * sizeof(struct Vmxnet3_TxQueueDesc) +
1534             i * sizeof(struct Vmxnet3_RxQueueDesc);
1535 
1536         /* Read interrupt number for this RX queue */
1537         s->rxq_descr[i].intr_idx =
1538             VMXNET3_READ_TX_QUEUE_DESCR8(d, qd_pa, conf.intrIdx);
1539         assert(vmxnet3_verify_intx(s, s->rxq_descr[i].intr_idx));
1540 
1541         VMW_CFPRN("RX Queue %d interrupt: %d", i, s->rxq_descr[i].intr_idx);
1542 
1543         /* Read rings memory locations */
1544         for (j = 0; j < VMXNET3_RX_RINGS_PER_QUEUE; j++) {
1545             /* RX rings */
1546             pa = VMXNET3_READ_RX_QUEUE_DESCR64(d, qd_pa, conf.rxRingBasePA[j]);
1547             size = VMXNET3_READ_RX_QUEUE_DESCR32(d, qd_pa, conf.rxRingSize[j]);
1548             if (size > VMXNET3_RX_RING_MAX_SIZE) {
1549                 size = VMXNET3_RX_RING_MAX_SIZE;
1550             }
1551             vmxnet3_ring_init(d, &s->rxq_descr[i].rx_ring[j], pa, size,
1552                               sizeof(struct Vmxnet3_RxDesc), false);
1553             VMW_CFPRN("RX queue %d:%d: Base: %" PRIx64 ", Size: %d",
1554                       i, j, pa, size);
1555         }
1556 
1557         /* RXC ring */
1558         pa = VMXNET3_READ_RX_QUEUE_DESCR64(d, qd_pa, conf.compRingBasePA);
1559         size = VMXNET3_READ_RX_QUEUE_DESCR32(d, qd_pa, conf.compRingSize);
1560         if (size > VMXNET3_RC_RING_MAX_SIZE) {
1561             size = VMXNET3_RC_RING_MAX_SIZE;
1562         }
1563         vmxnet3_ring_init(d, &s->rxq_descr[i].comp_ring, pa, size,
1564                           sizeof(struct Vmxnet3_RxCompDesc), true);
1565         VMW_CFPRN("RXC queue %d: Base: %" PRIx64 ", Size: %d", i, pa, size);
1566 
1567         s->rxq_descr[i].rx_stats_pa =
1568             qd_pa + offsetof(struct Vmxnet3_RxQueueDesc, stats);
1569         memset(&s->rxq_descr[i].rxq_stats, 0,
1570                sizeof(s->rxq_descr[i].rxq_stats));
1571     }
1572 
1573     vmxnet3_validate_interrupts(s);
1574 
1575     /* Make sure everything is in place before device activation */
1576     smp_wmb();
1577 
1578     vmxnet3_reset_mac(s);
1579 
1580     s->device_active = true;
1581 }
1582 
1583 static void vmxnet3_handle_command(VMXNET3State *s, uint64_t cmd)
1584 {
1585     s->last_command = cmd;
1586 
1587     switch (cmd) {
1588     case VMXNET3_CMD_GET_PERM_MAC_HI:
1589         VMW_CBPRN("Set: Get upper part of permanent MAC");
1590         break;
1591 
1592     case VMXNET3_CMD_GET_PERM_MAC_LO:
1593         VMW_CBPRN("Set: Get lower part of permanent MAC");
1594         break;
1595 
1596     case VMXNET3_CMD_GET_STATS:
1597         VMW_CBPRN("Set: Get device statistics");
1598         vmxnet3_fill_stats(s);
1599         break;
1600 
1601     case VMXNET3_CMD_ACTIVATE_DEV:
1602         VMW_CBPRN("Set: Activating vmxnet3 device");
1603         vmxnet3_activate_device(s);
1604         break;
1605 
1606     case VMXNET3_CMD_UPDATE_RX_MODE:
1607         VMW_CBPRN("Set: Update rx mode");
1608         vmxnet3_update_rx_mode(s);
1609         break;
1610 
1611     case VMXNET3_CMD_UPDATE_VLAN_FILTERS:
1612         VMW_CBPRN("Set: Update VLAN filters");
1613         vmxnet3_update_vlan_filters(s);
1614         break;
1615 
1616     case VMXNET3_CMD_UPDATE_MAC_FILTERS:
1617         VMW_CBPRN("Set: Update MAC filters");
1618         vmxnet3_update_mcast_filters(s);
1619         break;
1620 
1621     case VMXNET3_CMD_UPDATE_FEATURE:
1622         VMW_CBPRN("Set: Update features");
1623         vmxnet3_update_features(s);
1624         break;
1625 
1626     case VMXNET3_CMD_UPDATE_PMCFG:
1627         VMW_CBPRN("Set: Update power management config");
1628         vmxnet3_update_pm_state(s);
1629         break;
1630 
1631     case VMXNET3_CMD_GET_LINK:
1632         VMW_CBPRN("Set: Get link");
1633         break;
1634 
1635     case VMXNET3_CMD_RESET_DEV:
1636         VMW_CBPRN("Set: Reset device");
1637         vmxnet3_reset(s);
1638         break;
1639 
1640     case VMXNET3_CMD_QUIESCE_DEV:
1641         VMW_CBPRN("Set: VMXNET3_CMD_QUIESCE_DEV - deactivate the device");
1642         vmxnet3_deactivate_device(s);
1643         break;
1644 
1645     case VMXNET3_CMD_GET_CONF_INTR:
1646         VMW_CBPRN("Set: VMXNET3_CMD_GET_CONF_INTR - interrupt configuration");
1647         break;
1648 
1649     case VMXNET3_CMD_GET_ADAPTIVE_RING_INFO:
1650         VMW_CBPRN("Set: VMXNET3_CMD_GET_ADAPTIVE_RING_INFO - "
1651                   "adaptive ring info flags");
1652         break;
1653 
1654     case VMXNET3_CMD_GET_DID_LO:
1655         VMW_CBPRN("Set: Get lower part of device ID");
1656         break;
1657 
1658     case VMXNET3_CMD_GET_DID_HI:
1659         VMW_CBPRN("Set: Get upper part of device ID");
1660         break;
1661 
1662     case VMXNET3_CMD_GET_DEV_EXTRA_INFO:
1663         VMW_CBPRN("Set: Get device extra info");
1664         break;
1665 
1666     default:
1667         VMW_CBPRN("Received unknown command: %" PRIx64, cmd);
1668         break;
1669     }
1670 }
1671 
1672 static uint64_t vmxnet3_get_command_status(VMXNET3State *s)
1673 {
1674     uint64_t ret;
1675 
1676     switch (s->last_command) {
1677     case VMXNET3_CMD_ACTIVATE_DEV:
1678         ret = (s->device_active) ? 0 : 1;
1679         VMW_CFPRN("Device active: %" PRIx64, ret);
1680         break;
1681 
1682     case VMXNET3_CMD_RESET_DEV:
1683     case VMXNET3_CMD_QUIESCE_DEV:
1684     case VMXNET3_CMD_GET_QUEUE_STATUS:
1685     case VMXNET3_CMD_GET_DEV_EXTRA_INFO:
1686         ret = 0;
1687         break;
1688 
1689     case VMXNET3_CMD_GET_LINK:
1690         ret = s->link_status_and_speed;
1691         VMW_CFPRN("Link and speed: %" PRIx64, ret);
1692         break;
1693 
1694     case VMXNET3_CMD_GET_PERM_MAC_LO:
1695         ret = vmxnet3_get_mac_low(&s->perm_mac);
1696         break;
1697 
1698     case VMXNET3_CMD_GET_PERM_MAC_HI:
1699         ret = vmxnet3_get_mac_high(&s->perm_mac);
1700         break;
1701 
1702     case VMXNET3_CMD_GET_CONF_INTR:
1703         ret = vmxnet3_get_interrupt_config(s);
1704         break;
1705 
1706     case VMXNET3_CMD_GET_ADAPTIVE_RING_INFO:
1707         ret = VMXNET3_DISABLE_ADAPTIVE_RING;
1708         break;
1709 
1710     case VMXNET3_CMD_GET_DID_LO:
1711         ret = PCI_DEVICE_ID_VMWARE_VMXNET3;
1712         break;
1713 
1714     case VMXNET3_CMD_GET_DID_HI:
1715         ret = VMXNET3_DEVICE_REVISION;
1716         break;
1717 
1718     default:
1719         VMW_WRPRN("Received request for unknown command: %x", s->last_command);
1720         ret = 0;
1721         break;
1722     }
1723 
1724     return ret;
1725 }
1726 
1727 static void vmxnet3_set_events(VMXNET3State *s, uint32_t val)
1728 {
1729     uint32_t events;
1730     PCIDevice *d = PCI_DEVICE(s);
1731 
1732     VMW_CBPRN("Setting events: 0x%x", val);
1733     events = VMXNET3_READ_DRV_SHARED32(d, s->drv_shmem, ecr) | val;
1734     VMXNET3_WRITE_DRV_SHARED32(d, s->drv_shmem, ecr, events);
1735 }
1736 
1737 static void vmxnet3_ack_events(VMXNET3State *s, uint32_t val)
1738 {
1739     PCIDevice *d = PCI_DEVICE(s);
1740     uint32_t events;
1741 
1742     VMW_CBPRN("Clearing events: 0x%x", val);
1743     events = VMXNET3_READ_DRV_SHARED32(d, s->drv_shmem, ecr) & ~val;
1744     VMXNET3_WRITE_DRV_SHARED32(d, s->drv_shmem, ecr, events);
1745 }
1746 
1747 static void
1748 vmxnet3_io_bar1_write(void *opaque,
1749                       hwaddr addr,
1750                       uint64_t val,
1751                       unsigned size)
1752 {
1753     VMXNET3State *s = opaque;
1754 
1755     switch (addr) {
1756     /* Vmxnet3 Revision Report Selection */
1757     case VMXNET3_REG_VRRS:
1758         VMW_CBPRN("Write BAR1 [VMXNET3_REG_VRRS] = %" PRIx64 ", size %d",
1759                   val, size);
1760         break;
1761 
1762     /* UPT Version Report Selection */
1763     case VMXNET3_REG_UVRS:
1764         VMW_CBPRN("Write BAR1 [VMXNET3_REG_UVRS] = %" PRIx64 ", size %d",
1765                   val, size);
1766         break;
1767 
1768     /* Driver Shared Address Low */
1769     case VMXNET3_REG_DSAL:
1770         VMW_CBPRN("Write BAR1 [VMXNET3_REG_DSAL] = %" PRIx64 ", size %d",
1771                   val, size);
1772         /*
1773          * Guest driver will first write the low part of the shared
1774          * memory address. We save it to temp variable and set the
1775          * shared address only after we get the high part
1776          */
1777         if (val == 0) {
1778             vmxnet3_deactivate_device(s);
1779         }
1780         s->temp_shared_guest_driver_memory = val;
1781         s->drv_shmem = 0;
1782         break;
1783 
1784     /* Driver Shared Address High */
1785     case VMXNET3_REG_DSAH:
1786         VMW_CBPRN("Write BAR1 [VMXNET3_REG_DSAH] = %" PRIx64 ", size %d",
1787                   val, size);
1788         /*
1789          * Set the shared memory between guest driver and device.
1790          * We already should have low address part.
1791          */
1792         s->drv_shmem = s->temp_shared_guest_driver_memory | (val << 32);
1793         break;
1794 
1795     /* Command */
1796     case VMXNET3_REG_CMD:
1797         VMW_CBPRN("Write BAR1 [VMXNET3_REG_CMD] = %" PRIx64 ", size %d",
1798                   val, size);
1799         vmxnet3_handle_command(s, val);
1800         break;
1801 
1802     /* MAC Address Low */
1803     case VMXNET3_REG_MACL:
1804         VMW_CBPRN("Write BAR1 [VMXNET3_REG_MACL] = %" PRIx64 ", size %d",
1805                   val, size);
1806         s->temp_mac = val;
1807         break;
1808 
1809     /* MAC Address High */
1810     case VMXNET3_REG_MACH:
1811         VMW_CBPRN("Write BAR1 [VMXNET3_REG_MACH] = %" PRIx64 ", size %d",
1812                   val, size);
1813         vmxnet3_set_variable_mac(s, val, s->temp_mac);
1814         break;
1815 
1816     /* Interrupt Cause Register */
1817     case VMXNET3_REG_ICR:
1818         VMW_CBPRN("Write BAR1 [VMXNET3_REG_ICR] = %" PRIx64 ", size %d",
1819                   val, size);
1820         qemu_log_mask(LOG_GUEST_ERROR,
1821                       "%s: write to read-only register VMXNET3_REG_ICR\n",
1822                       TYPE_VMXNET3);
1823         break;
1824 
1825     /* Event Cause Register */
1826     case VMXNET3_REG_ECR:
1827         VMW_CBPRN("Write BAR1 [VMXNET3_REG_ECR] = %" PRIx64 ", size %d",
1828                   val, size);
1829         vmxnet3_ack_events(s, val);
1830         break;
1831 
1832     default:
1833         VMW_CBPRN("Unknown Write to BAR1 [%" PRIx64 "] = %" PRIx64 ", size %d",
1834                   addr, val, size);
1835         break;
1836     }
1837 }
1838 
1839 static uint64_t
1840 vmxnet3_io_bar1_read(void *opaque, hwaddr addr, unsigned size)
1841 {
1842         VMXNET3State *s = opaque;
1843         uint64_t ret = 0;
1844 
1845         switch (addr) {
1846         /* Vmxnet3 Revision Report Selection */
1847         case VMXNET3_REG_VRRS:
1848             VMW_CBPRN("Read BAR1 [VMXNET3_REG_VRRS], size %d", size);
1849             ret = VMXNET3_DEVICE_REVISION;
1850             break;
1851 
1852         /* UPT Version Report Selection */
1853         case VMXNET3_REG_UVRS:
1854             VMW_CBPRN("Read BAR1 [VMXNET3_REG_UVRS], size %d", size);
1855             ret = VMXNET3_UPT_REVISION;
1856             break;
1857 
1858         /* Command */
1859         case VMXNET3_REG_CMD:
1860             VMW_CBPRN("Read BAR1 [VMXNET3_REG_CMD], size %d", size);
1861             ret = vmxnet3_get_command_status(s);
1862             break;
1863 
1864         /* MAC Address Low */
1865         case VMXNET3_REG_MACL:
1866             VMW_CBPRN("Read BAR1 [VMXNET3_REG_MACL], size %d", size);
1867             ret = vmxnet3_get_mac_low(&s->conf.macaddr);
1868             break;
1869 
1870         /* MAC Address High */
1871         case VMXNET3_REG_MACH:
1872             VMW_CBPRN("Read BAR1 [VMXNET3_REG_MACH], size %d", size);
1873             ret = vmxnet3_get_mac_high(&s->conf.macaddr);
1874             break;
1875 
1876         /*
1877          * Interrupt Cause Register
1878          * Used for legacy interrupts only so interrupt index always 0
1879          */
1880         case VMXNET3_REG_ICR:
1881             VMW_CBPRN("Read BAR1 [VMXNET3_REG_ICR], size %d", size);
1882             if (vmxnet3_interrupt_asserted(s, 0)) {
1883                 vmxnet3_clear_interrupt(s, 0);
1884                 ret = true;
1885             } else {
1886                 ret = false;
1887             }
1888             break;
1889 
1890         default:
1891             VMW_CBPRN("Unknown read BAR1[%" PRIx64 "], %d bytes", addr, size);
1892             break;
1893         }
1894 
1895         return ret;
1896 }
1897 
1898 static int
1899 vmxnet3_can_receive(NetClientState *nc)
1900 {
1901     VMXNET3State *s = qemu_get_nic_opaque(nc);
1902     return s->device_active &&
1903            VMXNET_FLAG_IS_SET(s->link_status_and_speed, VMXNET3_LINK_STATUS_UP);
1904 }
1905 
1906 static inline bool
1907 vmxnet3_is_registered_vlan(VMXNET3State *s, const void *data)
1908 {
1909     uint16_t vlan_tag = eth_get_pkt_tci(data) & VLAN_VID_MASK;
1910     if (IS_SPECIAL_VLAN_ID(vlan_tag)) {
1911         return true;
1912     }
1913 
1914     return VMXNET3_VFTABLE_ENTRY_IS_SET(s->vlan_table, vlan_tag);
1915 }
1916 
1917 static bool
1918 vmxnet3_is_allowed_mcast_group(VMXNET3State *s, const uint8_t *group_mac)
1919 {
1920     int i;
1921     for (i = 0; i < s->mcast_list_len; i++) {
1922         if (!memcmp(group_mac, s->mcast_list[i].a, sizeof(s->mcast_list[i]))) {
1923             return true;
1924         }
1925     }
1926     return false;
1927 }
1928 
1929 static bool
1930 vmxnet3_rx_filter_may_indicate(VMXNET3State *s, const void *data,
1931     size_t size)
1932 {
1933     struct eth_header *ehdr = PKT_GET_ETH_HDR(data);
1934 
1935     if (VMXNET_FLAG_IS_SET(s->rx_mode, VMXNET3_RXM_PROMISC)) {
1936         return true;
1937     }
1938 
1939     if (!vmxnet3_is_registered_vlan(s, data)) {
1940         return false;
1941     }
1942 
1943     switch (net_rx_pkt_get_packet_type(s->rx_pkt)) {
1944     case ETH_PKT_UCAST:
1945         if (!VMXNET_FLAG_IS_SET(s->rx_mode, VMXNET3_RXM_UCAST)) {
1946             return false;
1947         }
1948         if (memcmp(s->conf.macaddr.a, ehdr->h_dest, ETH_ALEN)) {
1949             return false;
1950         }
1951         break;
1952 
1953     case ETH_PKT_BCAST:
1954         if (!VMXNET_FLAG_IS_SET(s->rx_mode, VMXNET3_RXM_BCAST)) {
1955             return false;
1956         }
1957         break;
1958 
1959     case ETH_PKT_MCAST:
1960         if (VMXNET_FLAG_IS_SET(s->rx_mode, VMXNET3_RXM_ALL_MULTI)) {
1961             return true;
1962         }
1963         if (!VMXNET_FLAG_IS_SET(s->rx_mode, VMXNET3_RXM_MCAST)) {
1964             return false;
1965         }
1966         if (!vmxnet3_is_allowed_mcast_group(s, ehdr->h_dest)) {
1967             return false;
1968         }
1969         break;
1970 
1971     default:
1972         g_assert_not_reached();
1973     }
1974 
1975     return true;
1976 }
1977 
1978 static ssize_t
1979 vmxnet3_receive(NetClientState *nc, const uint8_t *buf, size_t size)
1980 {
1981     VMXNET3State *s = qemu_get_nic_opaque(nc);
1982     size_t bytes_indicated;
1983 
1984     if (!vmxnet3_can_receive(nc)) {
1985         VMW_PKPRN("Cannot receive now");
1986         return -1;
1987     }
1988 
1989     if (s->peer_has_vhdr) {
1990         net_rx_pkt_set_vhdr(s->rx_pkt, (struct virtio_net_hdr *)buf);
1991         buf += sizeof(struct virtio_net_hdr);
1992         size -= sizeof(struct virtio_net_hdr);
1993     }
1994 
1995     net_rx_pkt_set_packet_type(s->rx_pkt,
1996         get_eth_packet_type(PKT_GET_ETH_HDR(buf)));
1997 
1998     if (vmxnet3_rx_filter_may_indicate(s, buf, size)) {
1999         struct iovec iov = {
2000             .iov_base = (void *)buf,
2001             .iov_len = size
2002         };
2003 
2004         net_rx_pkt_set_protocols(s->rx_pkt, &iov, 1, 0);
2005         vmxnet3_rx_need_csum_calculate(s->rx_pkt, buf, size);
2006         net_rx_pkt_attach_data(s->rx_pkt, buf, size, s->rx_vlan_stripping);
2007         bytes_indicated = vmxnet3_indicate_packet(s) ? size : -1;
2008         if (bytes_indicated < size) {
2009             VMW_PKPRN("RX: %zu of %zu bytes indicated", bytes_indicated, size);
2010         }
2011     } else {
2012         VMW_PKPRN("Packet dropped by RX filter");
2013         bytes_indicated = size;
2014     }
2015 
2016     assert(size > 0);
2017     assert(bytes_indicated != 0);
2018     return bytes_indicated;
2019 }
2020 
2021 static void vmxnet3_set_link_status(NetClientState *nc)
2022 {
2023     VMXNET3State *s = qemu_get_nic_opaque(nc);
2024 
2025     if (nc->link_down) {
2026         s->link_status_and_speed &= ~VMXNET3_LINK_STATUS_UP;
2027     } else {
2028         s->link_status_and_speed |= VMXNET3_LINK_STATUS_UP;
2029     }
2030 
2031     vmxnet3_set_events(s, VMXNET3_ECR_LINK);
2032     vmxnet3_trigger_interrupt(s, s->event_int_idx);
2033 }
2034 
2035 static NetClientInfo net_vmxnet3_info = {
2036         .type = NET_CLIENT_DRIVER_NIC,
2037         .size = sizeof(NICState),
2038         .receive = vmxnet3_receive,
2039         .link_status_changed = vmxnet3_set_link_status,
2040 };
2041 
2042 static bool vmxnet3_peer_has_vnet_hdr(VMXNET3State *s)
2043 {
2044     NetClientState *nc = qemu_get_queue(s->nic);
2045 
2046     if (qemu_has_vnet_hdr(nc->peer)) {
2047         return true;
2048     }
2049 
2050     return false;
2051 }
2052 
2053 static void vmxnet3_net_uninit(VMXNET3State *s)
2054 {
2055     g_free(s->mcast_list);
2056     vmxnet3_deactivate_device(s);
2057     qemu_del_nic(s->nic);
2058 }
2059 
2060 static void vmxnet3_net_init(VMXNET3State *s)
2061 {
2062     DeviceState *d = DEVICE(s);
2063 
2064     VMW_CBPRN("vmxnet3_net_init called...");
2065 
2066     qemu_macaddr_default_if_unset(&s->conf.macaddr);
2067 
2068     /* Windows guest will query the address that was set on init */
2069     memcpy(&s->perm_mac.a, &s->conf.macaddr.a, sizeof(s->perm_mac.a));
2070 
2071     s->mcast_list = NULL;
2072     s->mcast_list_len = 0;
2073 
2074     s->link_status_and_speed = VMXNET3_LINK_SPEED | VMXNET3_LINK_STATUS_UP;
2075 
2076     VMW_CFPRN("Permanent MAC: " MAC_FMT, MAC_ARG(s->perm_mac.a));
2077 
2078     s->nic = qemu_new_nic(&net_vmxnet3_info, &s->conf,
2079                           object_get_typename(OBJECT(s)),
2080                           d->id, &d->mem_reentrancy_guard, s);
2081 
2082     s->peer_has_vhdr = vmxnet3_peer_has_vnet_hdr(s);
2083     s->tx_sop = true;
2084     s->skip_current_tx_pkt = false;
2085     s->tx_pkt = NULL;
2086     s->rx_pkt = NULL;
2087     s->rx_vlan_stripping = false;
2088     s->lro_supported = false;
2089 
2090     if (s->peer_has_vhdr) {
2091         qemu_set_vnet_hdr_len(qemu_get_queue(s->nic)->peer,
2092             sizeof(struct virtio_net_hdr));
2093     }
2094 
2095     qemu_format_nic_info_str(qemu_get_queue(s->nic), s->conf.macaddr.a);
2096 }
2097 
2098 static void
2099 vmxnet3_unuse_msix_vectors(VMXNET3State *s, int num_vectors)
2100 {
2101     PCIDevice *d = PCI_DEVICE(s);
2102     int i;
2103     for (i = 0; i < num_vectors; i++) {
2104         msix_vector_unuse(d, i);
2105     }
2106 }
2107 
2108 static void
2109 vmxnet3_use_msix_vectors(VMXNET3State *s, int num_vectors)
2110 {
2111     PCIDevice *d = PCI_DEVICE(s);
2112     int i;
2113     for (i = 0; i < num_vectors; i++) {
2114         msix_vector_use(d, i);
2115     }
2116 }
2117 
2118 static bool
2119 vmxnet3_init_msix(VMXNET3State *s)
2120 {
2121     PCIDevice *d = PCI_DEVICE(s);
2122     int res = msix_init(d, VMXNET3_MAX_INTRS,
2123                         &s->msix_bar,
2124                         VMXNET3_MSIX_BAR_IDX, VMXNET3_OFF_MSIX_TABLE,
2125                         &s->msix_bar,
2126                         VMXNET3_MSIX_BAR_IDX, VMXNET3_OFF_MSIX_PBA(s),
2127                         VMXNET3_MSIX_OFFSET(s), NULL);
2128 
2129     if (0 > res) {
2130         VMW_WRPRN("Failed to initialize MSI-X, error %d", res);
2131         s->msix_used = false;
2132     } else {
2133         vmxnet3_use_msix_vectors(s, VMXNET3_MAX_INTRS);
2134         s->msix_used = true;
2135     }
2136     return s->msix_used;
2137 }
2138 
2139 static void
2140 vmxnet3_cleanup_msix(VMXNET3State *s)
2141 {
2142     PCIDevice *d = PCI_DEVICE(s);
2143 
2144     if (s->msix_used) {
2145         vmxnet3_unuse_msix_vectors(s, VMXNET3_MAX_INTRS);
2146         msix_uninit(d, &s->msix_bar, &s->msix_bar);
2147     }
2148 }
2149 
2150 static void
2151 vmxnet3_cleanup_msi(VMXNET3State *s)
2152 {
2153     PCIDevice *d = PCI_DEVICE(s);
2154 
2155     msi_uninit(d);
2156 }
2157 
2158 static const MemoryRegionOps b0_ops = {
2159     .read = vmxnet3_io_bar0_read,
2160     .write = vmxnet3_io_bar0_write,
2161     .endianness = DEVICE_LITTLE_ENDIAN,
2162     .impl = {
2163             .min_access_size = 4,
2164             .max_access_size = 4,
2165     },
2166 };
2167 
2168 static const MemoryRegionOps b1_ops = {
2169     .read = vmxnet3_io_bar1_read,
2170     .write = vmxnet3_io_bar1_write,
2171     .endianness = DEVICE_LITTLE_ENDIAN,
2172     .impl = {
2173             .min_access_size = 4,
2174             .max_access_size = 4,
2175     },
2176 };
2177 
2178 static uint64_t vmxnet3_device_serial_num(VMXNET3State *s)
2179 {
2180     uint64_t dsn_payload;
2181     uint8_t *dsnp = (uint8_t *)&dsn_payload;
2182 
2183     dsnp[0] = 0xfe;
2184     dsnp[1] = s->conf.macaddr.a[3];
2185     dsnp[2] = s->conf.macaddr.a[4];
2186     dsnp[3] = s->conf.macaddr.a[5];
2187     dsnp[4] = s->conf.macaddr.a[0];
2188     dsnp[5] = s->conf.macaddr.a[1];
2189     dsnp[6] = s->conf.macaddr.a[2];
2190     dsnp[7] = 0xff;
2191     return dsn_payload;
2192 }
2193 
2194 
2195 #define VMXNET3_USE_64BIT         (true)
2196 #define VMXNET3_PER_VECTOR_MASK   (false)
2197 
2198 static void vmxnet3_pci_realize(PCIDevice *pci_dev, Error **errp)
2199 {
2200     VMXNET3State *s = VMXNET3(pci_dev);
2201     int ret;
2202 
2203     VMW_CBPRN("Starting init...");
2204 
2205     memory_region_init_io(&s->bar0, OBJECT(s), &b0_ops, s,
2206                           "vmxnet3-b0", VMXNET3_PT_REG_SIZE);
2207     pci_register_bar(pci_dev, VMXNET3_BAR0_IDX,
2208                      PCI_BASE_ADDRESS_SPACE_MEMORY, &s->bar0);
2209 
2210     memory_region_init_io(&s->bar1, OBJECT(s), &b1_ops, s,
2211                           "vmxnet3-b1", VMXNET3_VD_REG_SIZE);
2212     pci_register_bar(pci_dev, VMXNET3_BAR1_IDX,
2213                      PCI_BASE_ADDRESS_SPACE_MEMORY, &s->bar1);
2214 
2215     memory_region_init(&s->msix_bar, OBJECT(s), "vmxnet3-msix-bar",
2216                        VMXNET3_MSIX_BAR_SIZE);
2217     pci_register_bar(pci_dev, VMXNET3_MSIX_BAR_IDX,
2218                      PCI_BASE_ADDRESS_SPACE_MEMORY, &s->msix_bar);
2219 
2220     vmxnet3_reset_interrupt_states(s);
2221 
2222     /* Interrupt pin A */
2223     pci_dev->config[PCI_INTERRUPT_PIN] = 0x01;
2224 
2225     ret = msi_init(pci_dev, VMXNET3_MSI_OFFSET(s), VMXNET3_MAX_NMSIX_INTRS,
2226                    VMXNET3_USE_64BIT, VMXNET3_PER_VECTOR_MASK, NULL);
2227     /* Any error other than -ENOTSUP(board's MSI support is broken)
2228      * is a programming error. Fall back to INTx silently on -ENOTSUP */
2229     assert(!ret || ret == -ENOTSUP);
2230 
2231     if (!vmxnet3_init_msix(s)) {
2232         VMW_WRPRN("Failed to initialize MSI-X, configuration is inconsistent.");
2233     }
2234 
2235     vmxnet3_net_init(s);
2236 
2237     if (pci_is_express(pci_dev)) {
2238         if (pci_bus_is_express(pci_get_bus(pci_dev))) {
2239             pcie_endpoint_cap_init(pci_dev, VMXNET3_EXP_EP_OFFSET);
2240         }
2241 
2242         pcie_dev_ser_num_init(pci_dev, VMXNET3_DSN_OFFSET,
2243                               vmxnet3_device_serial_num(s));
2244     }
2245 }
2246 
2247 static void vmxnet3_instance_init(Object *obj)
2248 {
2249     VMXNET3State *s = VMXNET3(obj);
2250     device_add_bootindex_property(obj, &s->conf.bootindex,
2251                                   "bootindex", "/ethernet-phy@0",
2252                                   DEVICE(obj));
2253 }
2254 
2255 static void vmxnet3_pci_uninit(PCIDevice *pci_dev)
2256 {
2257     VMXNET3State *s = VMXNET3(pci_dev);
2258 
2259     VMW_CBPRN("Starting uninit...");
2260 
2261     vmxnet3_net_uninit(s);
2262 
2263     vmxnet3_cleanup_msix(s);
2264 
2265     vmxnet3_cleanup_msi(s);
2266 }
2267 
2268 static void vmxnet3_qdev_reset(DeviceState *dev)
2269 {
2270     PCIDevice *d = PCI_DEVICE(dev);
2271     VMXNET3State *s = VMXNET3(d);
2272 
2273     VMW_CBPRN("Starting QDEV reset...");
2274     vmxnet3_reset(s);
2275 }
2276 
2277 static bool vmxnet3_mc_list_needed(void *opaque)
2278 {
2279     return true;
2280 }
2281 
2282 static int vmxnet3_mcast_list_pre_load(void *opaque)
2283 {
2284     VMXNET3State *s = opaque;
2285 
2286     s->mcast_list = g_malloc(s->mcast_list_buff_size);
2287 
2288     return 0;
2289 }
2290 
2291 
2292 static int vmxnet3_pre_save(void *opaque)
2293 {
2294     VMXNET3State *s = opaque;
2295 
2296     s->mcast_list_buff_size = s->mcast_list_len * sizeof(MACAddr);
2297 
2298     return 0;
2299 }
2300 
2301 static const VMStateDescription vmxstate_vmxnet3_mcast_list = {
2302     .name = "vmxnet3/mcast_list",
2303     .version_id = 1,
2304     .minimum_version_id = 1,
2305     .pre_load = vmxnet3_mcast_list_pre_load,
2306     .needed = vmxnet3_mc_list_needed,
2307     .fields = (const VMStateField[]) {
2308         VMSTATE_VBUFFER_UINT32(mcast_list, VMXNET3State, 0, NULL,
2309             mcast_list_buff_size),
2310         VMSTATE_END_OF_LIST()
2311     }
2312 };
2313 
2314 static const VMStateDescription vmstate_vmxnet3_ring = {
2315     .name = "vmxnet3-ring",
2316     .version_id = 0,
2317     .fields = (const VMStateField[]) {
2318         VMSTATE_UINT64(pa, Vmxnet3Ring),
2319         VMSTATE_UINT32(size, Vmxnet3Ring),
2320         VMSTATE_UINT32(cell_size, Vmxnet3Ring),
2321         VMSTATE_UINT32(next, Vmxnet3Ring),
2322         VMSTATE_UINT8(gen, Vmxnet3Ring),
2323         VMSTATE_END_OF_LIST()
2324     }
2325 };
2326 
2327 static const VMStateDescription vmstate_vmxnet3_tx_stats = {
2328     .name = "vmxnet3-tx-stats",
2329     .version_id = 0,
2330     .fields = (const VMStateField[]) {
2331         VMSTATE_UINT64(TSOPktsTxOK, struct UPT1_TxStats),
2332         VMSTATE_UINT64(TSOBytesTxOK, struct UPT1_TxStats),
2333         VMSTATE_UINT64(ucastPktsTxOK, struct UPT1_TxStats),
2334         VMSTATE_UINT64(ucastBytesTxOK, struct UPT1_TxStats),
2335         VMSTATE_UINT64(mcastPktsTxOK, struct UPT1_TxStats),
2336         VMSTATE_UINT64(mcastBytesTxOK, struct UPT1_TxStats),
2337         VMSTATE_UINT64(bcastPktsTxOK, struct UPT1_TxStats),
2338         VMSTATE_UINT64(bcastBytesTxOK, struct UPT1_TxStats),
2339         VMSTATE_UINT64(pktsTxError, struct UPT1_TxStats),
2340         VMSTATE_UINT64(pktsTxDiscard, struct UPT1_TxStats),
2341         VMSTATE_END_OF_LIST()
2342     }
2343 };
2344 
2345 static const VMStateDescription vmstate_vmxnet3_txq_descr = {
2346     .name = "vmxnet3-txq-descr",
2347     .version_id = 0,
2348     .fields = (const VMStateField[]) {
2349         VMSTATE_STRUCT(tx_ring, Vmxnet3TxqDescr, 0, vmstate_vmxnet3_ring,
2350                        Vmxnet3Ring),
2351         VMSTATE_STRUCT(comp_ring, Vmxnet3TxqDescr, 0, vmstate_vmxnet3_ring,
2352                        Vmxnet3Ring),
2353         VMSTATE_UINT8(intr_idx, Vmxnet3TxqDescr),
2354         VMSTATE_UINT64(tx_stats_pa, Vmxnet3TxqDescr),
2355         VMSTATE_STRUCT(txq_stats, Vmxnet3TxqDescr, 0, vmstate_vmxnet3_tx_stats,
2356                        struct UPT1_TxStats),
2357         VMSTATE_END_OF_LIST()
2358     }
2359 };
2360 
2361 static const VMStateDescription vmstate_vmxnet3_rx_stats = {
2362     .name = "vmxnet3-rx-stats",
2363     .version_id = 0,
2364     .fields = (const VMStateField[]) {
2365         VMSTATE_UINT64(LROPktsRxOK, struct UPT1_RxStats),
2366         VMSTATE_UINT64(LROBytesRxOK, struct UPT1_RxStats),
2367         VMSTATE_UINT64(ucastPktsRxOK, struct UPT1_RxStats),
2368         VMSTATE_UINT64(ucastBytesRxOK, struct UPT1_RxStats),
2369         VMSTATE_UINT64(mcastPktsRxOK, struct UPT1_RxStats),
2370         VMSTATE_UINT64(mcastBytesRxOK, struct UPT1_RxStats),
2371         VMSTATE_UINT64(bcastPktsRxOK, struct UPT1_RxStats),
2372         VMSTATE_UINT64(bcastBytesRxOK, struct UPT1_RxStats),
2373         VMSTATE_UINT64(pktsRxOutOfBuf, struct UPT1_RxStats),
2374         VMSTATE_UINT64(pktsRxError, struct UPT1_RxStats),
2375         VMSTATE_END_OF_LIST()
2376     }
2377 };
2378 
2379 static const VMStateDescription vmstate_vmxnet3_rxq_descr = {
2380     .name = "vmxnet3-rxq-descr",
2381     .version_id = 0,
2382     .fields = (const VMStateField[]) {
2383         VMSTATE_STRUCT_ARRAY(rx_ring, Vmxnet3RxqDescr,
2384                              VMXNET3_RX_RINGS_PER_QUEUE, 0,
2385                              vmstate_vmxnet3_ring, Vmxnet3Ring),
2386         VMSTATE_STRUCT(comp_ring, Vmxnet3RxqDescr, 0, vmstate_vmxnet3_ring,
2387                        Vmxnet3Ring),
2388         VMSTATE_UINT8(intr_idx, Vmxnet3RxqDescr),
2389         VMSTATE_UINT64(rx_stats_pa, Vmxnet3RxqDescr),
2390         VMSTATE_STRUCT(rxq_stats, Vmxnet3RxqDescr, 0, vmstate_vmxnet3_rx_stats,
2391                        struct UPT1_RxStats),
2392         VMSTATE_END_OF_LIST()
2393     }
2394 };
2395 
2396 static int vmxnet3_post_load(void *opaque, int version_id)
2397 {
2398     VMXNET3State *s = opaque;
2399 
2400     net_tx_pkt_init(&s->tx_pkt, s->max_tx_frags);
2401     net_rx_pkt_init(&s->rx_pkt);
2402 
2403     if (s->msix_used) {
2404         vmxnet3_use_msix_vectors(s, VMXNET3_MAX_INTRS);
2405     }
2406 
2407     if (!vmxnet3_validate_queues(s)) {
2408         return -1;
2409     }
2410     vmxnet3_validate_interrupts(s);
2411 
2412     return 0;
2413 }
2414 
2415 static const VMStateDescription vmstate_vmxnet3_int_state = {
2416     .name = "vmxnet3-int-state",
2417     .version_id = 0,
2418     .fields = (const VMStateField[]) {
2419         VMSTATE_BOOL(is_masked, Vmxnet3IntState),
2420         VMSTATE_BOOL(is_pending, Vmxnet3IntState),
2421         VMSTATE_BOOL(is_asserted, Vmxnet3IntState),
2422         VMSTATE_END_OF_LIST()
2423     }
2424 };
2425 
2426 static const VMStateDescription vmstate_vmxnet3 = {
2427     .name = "vmxnet3",
2428     .version_id = 1,
2429     .minimum_version_id = 1,
2430     .pre_save = vmxnet3_pre_save,
2431     .post_load = vmxnet3_post_load,
2432     .fields = (const VMStateField[]) {
2433             VMSTATE_PCI_DEVICE(parent_obj, VMXNET3State),
2434             VMSTATE_MSIX(parent_obj, VMXNET3State),
2435             VMSTATE_BOOL(rx_packets_compound, VMXNET3State),
2436             VMSTATE_BOOL(rx_vlan_stripping, VMXNET3State),
2437             VMSTATE_BOOL(lro_supported, VMXNET3State),
2438             VMSTATE_UINT32(rx_mode, VMXNET3State),
2439             VMSTATE_UINT32(mcast_list_len, VMXNET3State),
2440             VMSTATE_UINT32(mcast_list_buff_size, VMXNET3State),
2441             VMSTATE_UINT32_ARRAY(vlan_table, VMXNET3State, VMXNET3_VFT_SIZE),
2442             VMSTATE_UINT32(mtu, VMXNET3State),
2443             VMSTATE_UINT16(max_rx_frags, VMXNET3State),
2444             VMSTATE_UINT32(max_tx_frags, VMXNET3State),
2445             VMSTATE_UINT8(event_int_idx, VMXNET3State),
2446             VMSTATE_BOOL(auto_int_masking, VMXNET3State),
2447             VMSTATE_UINT8(txq_num, VMXNET3State),
2448             VMSTATE_UINT8(rxq_num, VMXNET3State),
2449             VMSTATE_UINT32(device_active, VMXNET3State),
2450             VMSTATE_UINT32(last_command, VMXNET3State),
2451             VMSTATE_UINT32(link_status_and_speed, VMXNET3State),
2452             VMSTATE_UINT32(temp_mac, VMXNET3State),
2453             VMSTATE_UINT64(drv_shmem, VMXNET3State),
2454             VMSTATE_UINT64(temp_shared_guest_driver_memory, VMXNET3State),
2455 
2456             VMSTATE_STRUCT_ARRAY(txq_descr, VMXNET3State,
2457                 VMXNET3_DEVICE_MAX_TX_QUEUES, 0, vmstate_vmxnet3_txq_descr,
2458                 Vmxnet3TxqDescr),
2459             VMSTATE_STRUCT_ARRAY(rxq_descr, VMXNET3State,
2460                 VMXNET3_DEVICE_MAX_RX_QUEUES, 0, vmstate_vmxnet3_rxq_descr,
2461                 Vmxnet3RxqDescr),
2462             VMSTATE_STRUCT_ARRAY(interrupt_states, VMXNET3State,
2463                 VMXNET3_MAX_INTRS, 0, vmstate_vmxnet3_int_state,
2464                 Vmxnet3IntState),
2465 
2466             VMSTATE_END_OF_LIST()
2467     },
2468     .subsections = (const VMStateDescription * const []) {
2469         &vmxstate_vmxnet3_mcast_list,
2470         NULL
2471     }
2472 };
2473 
2474 static Property vmxnet3_properties[] = {
2475     DEFINE_NIC_PROPERTIES(VMXNET3State, conf),
2476     DEFINE_PROP_BIT("x-old-msi-offsets", VMXNET3State, compat_flags,
2477                     VMXNET3_COMPAT_FLAG_OLD_MSI_OFFSETS_BIT, false),
2478     DEFINE_PROP_BIT("x-disable-pcie", VMXNET3State, compat_flags,
2479                     VMXNET3_COMPAT_FLAG_DISABLE_PCIE_BIT, false),
2480     DEFINE_PROP_END_OF_LIST(),
2481 };
2482 
2483 static void vmxnet3_realize(DeviceState *qdev, Error **errp)
2484 {
2485     VMXNET3Class *vc = VMXNET3_DEVICE_GET_CLASS(qdev);
2486     PCIDevice *pci_dev = PCI_DEVICE(qdev);
2487     VMXNET3State *s = VMXNET3(qdev);
2488 
2489     if (!(s->compat_flags & VMXNET3_COMPAT_FLAG_DISABLE_PCIE)) {
2490         pci_dev->cap_present |= QEMU_PCI_CAP_EXPRESS;
2491     }
2492 
2493     vc->parent_dc_realize(qdev, errp);
2494 }
2495 
2496 static void vmxnet3_class_init(ObjectClass *class, void *data)
2497 {
2498     DeviceClass *dc = DEVICE_CLASS(class);
2499     PCIDeviceClass *c = PCI_DEVICE_CLASS(class);
2500     VMXNET3Class *vc = VMXNET3_DEVICE_CLASS(class);
2501 
2502     c->realize = vmxnet3_pci_realize;
2503     c->exit = vmxnet3_pci_uninit;
2504     c->vendor_id = PCI_VENDOR_ID_VMWARE;
2505     c->device_id = PCI_DEVICE_ID_VMWARE_VMXNET3;
2506     c->revision = PCI_DEVICE_ID_VMWARE_VMXNET3_REVISION;
2507     c->romfile = "efi-vmxnet3.rom";
2508     c->class_id = PCI_CLASS_NETWORK_ETHERNET;
2509     c->subsystem_vendor_id = PCI_VENDOR_ID_VMWARE;
2510     c->subsystem_id = PCI_DEVICE_ID_VMWARE_VMXNET3;
2511     device_class_set_parent_realize(dc, vmxnet3_realize,
2512                                     &vc->parent_dc_realize);
2513     dc->desc = "VMWare Paravirtualized Ethernet v3";
2514     device_class_set_legacy_reset(dc, vmxnet3_qdev_reset);
2515     dc->vmsd = &vmstate_vmxnet3;
2516     device_class_set_props(dc, vmxnet3_properties);
2517     set_bit(DEVICE_CATEGORY_NETWORK, dc->categories);
2518 }
2519 
2520 static const TypeInfo vmxnet3_info = {
2521     .name          = TYPE_VMXNET3,
2522     .parent        = TYPE_PCI_DEVICE,
2523     .class_size    = sizeof(VMXNET3Class),
2524     .instance_size = sizeof(VMXNET3State),
2525     .class_init    = vmxnet3_class_init,
2526     .instance_init = vmxnet3_instance_init,
2527     .interfaces = (InterfaceInfo[]) {
2528         { INTERFACE_PCIE_DEVICE },
2529         { INTERFACE_CONVENTIONAL_PCI_DEVICE },
2530         { }
2531     },
2532 };
2533 
2534 static void vmxnet3_register_types(void)
2535 {
2536     VMW_CBPRN("vmxnet3_register_types called...");
2537     type_register_static(&vmxnet3_info);
2538 }
2539 
2540 type_init(vmxnet3_register_types)
2541