xref: /openbmc/qemu/hw/net/vmxnet3.c (revision 834e8bf1)
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/module.h"
27 #include "hw/pci/msix.h"
28 #include "hw/pci/msi.h"
29 #include "migration/register.h"
30 #include "migration/vmstate.h"
31 
32 #include "vmxnet3.h"
33 #include "vmxnet3_defs.h"
34 #include "vmxnet_debug.h"
35 #include "vmware_utils.h"
36 #include "net_tx_pkt.h"
37 #include "net_rx_pkt.h"
38 
39 #define PCI_DEVICE_ID_VMWARE_VMXNET3_REVISION 0x1
40 #define VMXNET3_MSIX_BAR_SIZE 0x2000
41 #define MIN_BUF_SIZE 60
42 
43 /* Compatibility flags for migration */
44 #define VMXNET3_COMPAT_FLAG_OLD_MSI_OFFSETS_BIT 0
45 #define VMXNET3_COMPAT_FLAG_OLD_MSI_OFFSETS \
46     (1 << VMXNET3_COMPAT_FLAG_OLD_MSI_OFFSETS_BIT)
47 #define VMXNET3_COMPAT_FLAG_DISABLE_PCIE_BIT 1
48 #define VMXNET3_COMPAT_FLAG_DISABLE_PCIE \
49     (1 << VMXNET3_COMPAT_FLAG_DISABLE_PCIE_BIT)
50 
51 #define VMXNET3_EXP_EP_OFFSET (0x48)
52 #define VMXNET3_MSI_OFFSET(s) \
53     ((s)->compat_flags & VMXNET3_COMPAT_FLAG_OLD_MSI_OFFSETS ? 0x50 : 0x84)
54 #define VMXNET3_MSIX_OFFSET(s) \
55     ((s)->compat_flags & VMXNET3_COMPAT_FLAG_OLD_MSI_OFFSETS ? 0 : 0x9c)
56 #define VMXNET3_DSN_OFFSET     (0x100)
57 
58 #define VMXNET3_BAR0_IDX      (0)
59 #define VMXNET3_BAR1_IDX      (1)
60 #define VMXNET3_MSIX_BAR_IDX  (2)
61 
62 #define VMXNET3_OFF_MSIX_TABLE (0x000)
63 #define VMXNET3_OFF_MSIX_PBA(s) \
64     ((s)->compat_flags & VMXNET3_COMPAT_FLAG_OLD_MSI_OFFSETS ? 0x800 : 0x1000)
65 
66 /* Link speed in Mbps should be shifted by 16 */
67 #define VMXNET3_LINK_SPEED      (1000 << 16)
68 
69 /* Link status: 1 - up, 0 - down. */
70 #define VMXNET3_LINK_STATUS_UP  0x1
71 
72 /* Least significant bit should be set for revision and version */
73 #define VMXNET3_UPT_REVISION      0x1
74 #define VMXNET3_DEVICE_REVISION   0x1
75 
76 /* Number of interrupt vectors for non-MSIx modes */
77 #define VMXNET3_MAX_NMSIX_INTRS   (1)
78 
79 /* Macros for rings descriptors access */
80 #define VMXNET3_READ_TX_QUEUE_DESCR8(_d, dpa, field) \
81     (vmw_shmem_ld8(_d, dpa + offsetof(struct Vmxnet3_TxQueueDesc, field)))
82 
83 #define VMXNET3_WRITE_TX_QUEUE_DESCR8(_d, dpa, field, value) \
84     (vmw_shmem_st8(_d, dpa + offsetof(struct Vmxnet3_TxQueueDesc, field, value)))
85 
86 #define VMXNET3_READ_TX_QUEUE_DESCR32(_d, dpa, field) \
87     (vmw_shmem_ld32(_d, dpa + offsetof(struct Vmxnet3_TxQueueDesc, field)))
88 
89 #define VMXNET3_WRITE_TX_QUEUE_DESCR32(_d, dpa, field, value) \
90     (vmw_shmem_st32(_d, dpa + offsetof(struct Vmxnet3_TxQueueDesc, field), value))
91 
92 #define VMXNET3_READ_TX_QUEUE_DESCR64(_d, dpa, field) \
93     (vmw_shmem_ld64(_d, dpa + offsetof(struct Vmxnet3_TxQueueDesc, field)))
94 
95 #define VMXNET3_WRITE_TX_QUEUE_DESCR64(_d, dpa, field, value) \
96     (vmw_shmem_st64(_d, dpa + offsetof(struct Vmxnet3_TxQueueDesc, field), value))
97 
98 #define VMXNET3_READ_RX_QUEUE_DESCR64(_d, dpa, field) \
99     (vmw_shmem_ld64(_d, dpa + offsetof(struct Vmxnet3_RxQueueDesc, field)))
100 
101 #define VMXNET3_READ_RX_QUEUE_DESCR32(_d, dpa, field) \
102     (vmw_shmem_ld32(_d, dpa + offsetof(struct Vmxnet3_RxQueueDesc, field)))
103 
104 #define VMXNET3_WRITE_RX_QUEUE_DESCR64(_d, dpa, field, value) \
105     (vmw_shmem_st64(_d, dpa + offsetof(struct Vmxnet3_RxQueueDesc, field), value))
106 
107 #define VMXNET3_WRITE_RX_QUEUE_DESCR8(_d, dpa, field, value) \
108     (vmw_shmem_st8(_d, dpa + offsetof(struct Vmxnet3_RxQueueDesc, field), value))
109 
110 /* Macros for guest driver shared area access */
111 #define VMXNET3_READ_DRV_SHARED64(_d, shpa, field) \
112     (vmw_shmem_ld64(_d, shpa + offsetof(struct Vmxnet3_DriverShared, field)))
113 
114 #define VMXNET3_READ_DRV_SHARED32(_d, shpa, field) \
115     (vmw_shmem_ld32(_d, shpa + offsetof(struct Vmxnet3_DriverShared, field)))
116 
117 #define VMXNET3_WRITE_DRV_SHARED32(_d, shpa, field, val) \
118     (vmw_shmem_st32(_d, shpa + offsetof(struct Vmxnet3_DriverShared, field), val))
119 
120 #define VMXNET3_READ_DRV_SHARED16(_d, shpa, field) \
121     (vmw_shmem_ld16(_d, shpa + offsetof(struct Vmxnet3_DriverShared, field)))
122 
123 #define VMXNET3_READ_DRV_SHARED8(_d, shpa, field) \
124     (vmw_shmem_ld8(_d, shpa + offsetof(struct Vmxnet3_DriverShared, field)))
125 
126 #define VMXNET3_READ_DRV_SHARED(_d, shpa, field, b, l) \
127     (vmw_shmem_read(_d, shpa + offsetof(struct Vmxnet3_DriverShared, field), b, l))
128 
129 #define VMXNET_FLAG_IS_SET(field, flag) (((field) & (flag)) == (flag))
130 
131 typedef struct VMXNET3Class {
132     PCIDeviceClass parent_class;
133     DeviceRealize parent_dc_realize;
134 } VMXNET3Class;
135 
136 #define VMXNET3_DEVICE_CLASS(klass) \
137     OBJECT_CLASS_CHECK(VMXNET3Class, (klass), TYPE_VMXNET3)
138 #define VMXNET3_DEVICE_GET_CLASS(obj) \
139     OBJECT_GET_CLASS(VMXNET3Class, (obj), 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         net_tx_pkt_build_vheader(s->tx_pkt, false, false, 0);
443         break;
444 
445     case VMXNET3_OM_CSUM:
446         net_tx_pkt_build_vheader(s->tx_pkt, false, true, 0);
447         VMW_PKPRN("L4 CSO requested\n");
448         break;
449 
450     case VMXNET3_OM_TSO:
451         net_tx_pkt_build_vheader(s->tx_pkt, true, true,
452             s->cso_or_gso_size);
453         net_tx_pkt_update_ip_checksums(s->tx_pkt);
454         VMW_PKPRN("GSO offload requested.");
455         break;
456 
457     default:
458         g_assert_not_reached();
459         return false;
460     }
461 
462     return true;
463 }
464 
465 static void
466 vmxnet3_tx_retrieve_metadata(VMXNET3State *s,
467                              const struct Vmxnet3_TxDesc *txd)
468 {
469     s->offload_mode = txd->om;
470     s->cso_or_gso_size = txd->msscof;
471     s->tci = txd->tci;
472     s->needs_vlan = txd->ti;
473 }
474 
475 typedef enum {
476     VMXNET3_PKT_STATUS_OK,
477     VMXNET3_PKT_STATUS_ERROR,
478     VMXNET3_PKT_STATUS_DISCARD,/* only for tx */
479     VMXNET3_PKT_STATUS_OUT_OF_BUF /* only for rx */
480 } Vmxnet3PktStatus;
481 
482 static void
483 vmxnet3_on_tx_done_update_stats(VMXNET3State *s, int qidx,
484     Vmxnet3PktStatus status)
485 {
486     size_t tot_len = net_tx_pkt_get_total_len(s->tx_pkt);
487     struct UPT1_TxStats *stats = &s->txq_descr[qidx].txq_stats;
488 
489     switch (status) {
490     case VMXNET3_PKT_STATUS_OK:
491         switch (net_tx_pkt_get_packet_type(s->tx_pkt)) {
492         case ETH_PKT_BCAST:
493             stats->bcastPktsTxOK++;
494             stats->bcastBytesTxOK += tot_len;
495             break;
496         case ETH_PKT_MCAST:
497             stats->mcastPktsTxOK++;
498             stats->mcastBytesTxOK += tot_len;
499             break;
500         case ETH_PKT_UCAST:
501             stats->ucastPktsTxOK++;
502             stats->ucastBytesTxOK += tot_len;
503             break;
504         default:
505             g_assert_not_reached();
506         }
507 
508         if (s->offload_mode == VMXNET3_OM_TSO) {
509             /*
510              * According to VMWARE headers this statistic is a number
511              * of packets after segmentation but since we don't have
512              * this information in QEMU model, the best we can do is to
513              * provide number of non-segmented packets
514              */
515             stats->TSOPktsTxOK++;
516             stats->TSOBytesTxOK += tot_len;
517         }
518         break;
519 
520     case VMXNET3_PKT_STATUS_DISCARD:
521         stats->pktsTxDiscard++;
522         break;
523 
524     case VMXNET3_PKT_STATUS_ERROR:
525         stats->pktsTxError++;
526         break;
527 
528     default:
529         g_assert_not_reached();
530     }
531 }
532 
533 static void
534 vmxnet3_on_rx_done_update_stats(VMXNET3State *s,
535                                 int qidx,
536                                 Vmxnet3PktStatus status)
537 {
538     struct UPT1_RxStats *stats = &s->rxq_descr[qidx].rxq_stats;
539     size_t tot_len = net_rx_pkt_get_total_len(s->rx_pkt);
540 
541     switch (status) {
542     case VMXNET3_PKT_STATUS_OUT_OF_BUF:
543         stats->pktsRxOutOfBuf++;
544         break;
545 
546     case VMXNET3_PKT_STATUS_ERROR:
547         stats->pktsRxError++;
548         break;
549     case VMXNET3_PKT_STATUS_OK:
550         switch (net_rx_pkt_get_packet_type(s->rx_pkt)) {
551         case ETH_PKT_BCAST:
552             stats->bcastPktsRxOK++;
553             stats->bcastBytesRxOK += tot_len;
554             break;
555         case ETH_PKT_MCAST:
556             stats->mcastPktsRxOK++;
557             stats->mcastBytesRxOK += tot_len;
558             break;
559         case ETH_PKT_UCAST:
560             stats->ucastPktsRxOK++;
561             stats->ucastBytesRxOK += tot_len;
562             break;
563         default:
564             g_assert_not_reached();
565         }
566 
567         if (tot_len > s->mtu) {
568             stats->LROPktsRxOK++;
569             stats->LROBytesRxOK += tot_len;
570         }
571         break;
572     default:
573         g_assert_not_reached();
574     }
575 }
576 
577 static inline void
578 vmxnet3_ring_read_curr_txdesc(PCIDevice *pcidev, Vmxnet3Ring *ring,
579                               struct Vmxnet3_TxDesc *txd)
580 {
581     vmxnet3_ring_read_curr_cell(pcidev, ring, txd);
582     txd->addr = le64_to_cpu(txd->addr);
583     txd->val1 = le32_to_cpu(txd->val1);
584     txd->val2 = le32_to_cpu(txd->val2);
585 }
586 
587 static inline bool
588 vmxnet3_pop_next_tx_descr(VMXNET3State *s,
589                           int qidx,
590                           struct Vmxnet3_TxDesc *txd,
591                           uint32_t *descr_idx)
592 {
593     Vmxnet3Ring *ring = &s->txq_descr[qidx].tx_ring;
594     PCIDevice *d = PCI_DEVICE(s);
595 
596     vmxnet3_ring_read_curr_txdesc(d, ring, txd);
597     if (txd->gen == vmxnet3_ring_curr_gen(ring)) {
598         /* Only read after generation field verification */
599         smp_rmb();
600         /* Re-read to be sure we got the latest version */
601         vmxnet3_ring_read_curr_txdesc(d, ring, txd);
602         VMXNET3_RING_DUMP(VMW_RIPRN, "TX", qidx, ring);
603         *descr_idx = vmxnet3_ring_curr_cell_idx(ring);
604         vmxnet3_inc_tx_consumption_counter(s, qidx);
605         return true;
606     }
607 
608     return false;
609 }
610 
611 static bool
612 vmxnet3_send_packet(VMXNET3State *s, uint32_t qidx)
613 {
614     Vmxnet3PktStatus status = VMXNET3_PKT_STATUS_OK;
615 
616     if (!vmxnet3_setup_tx_offloads(s)) {
617         status = VMXNET3_PKT_STATUS_ERROR;
618         goto func_exit;
619     }
620 
621     /* debug prints */
622     vmxnet3_dump_virt_hdr(net_tx_pkt_get_vhdr(s->tx_pkt));
623     net_tx_pkt_dump(s->tx_pkt);
624 
625     if (!net_tx_pkt_send(s->tx_pkt, qemu_get_queue(s->nic))) {
626         status = VMXNET3_PKT_STATUS_DISCARD;
627         goto func_exit;
628     }
629 
630 func_exit:
631     vmxnet3_on_tx_done_update_stats(s, qidx, status);
632     return (status == VMXNET3_PKT_STATUS_OK);
633 }
634 
635 static void vmxnet3_process_tx_queue(VMXNET3State *s, int qidx)
636 {
637     struct Vmxnet3_TxDesc txd;
638     uint32_t txd_idx;
639     uint32_t data_len;
640     hwaddr data_pa;
641 
642     for (;;) {
643         if (!vmxnet3_pop_next_tx_descr(s, qidx, &txd, &txd_idx)) {
644             break;
645         }
646 
647         vmxnet3_dump_tx_descr(&txd);
648 
649         if (!s->skip_current_tx_pkt) {
650             data_len = (txd.len > 0) ? txd.len : VMXNET3_MAX_TX_BUF_SIZE;
651             data_pa = txd.addr;
652 
653             if (!net_tx_pkt_add_raw_fragment(s->tx_pkt,
654                                                 data_pa,
655                                                 data_len)) {
656                 s->skip_current_tx_pkt = true;
657             }
658         }
659 
660         if (s->tx_sop) {
661             vmxnet3_tx_retrieve_metadata(s, &txd);
662             s->tx_sop = false;
663         }
664 
665         if (txd.eop) {
666             if (!s->skip_current_tx_pkt && net_tx_pkt_parse(s->tx_pkt)) {
667                 if (s->needs_vlan) {
668                     net_tx_pkt_setup_vlan_header(s->tx_pkt, s->tci);
669                 }
670 
671                 vmxnet3_send_packet(s, qidx);
672             } else {
673                 vmxnet3_on_tx_done_update_stats(s, qidx,
674                                                 VMXNET3_PKT_STATUS_ERROR);
675             }
676 
677             vmxnet3_complete_packet(s, qidx, txd_idx);
678             s->tx_sop = true;
679             s->skip_current_tx_pkt = false;
680             net_tx_pkt_reset(s->tx_pkt);
681         }
682     }
683 }
684 
685 static inline void
686 vmxnet3_read_next_rx_descr(VMXNET3State *s, int qidx, int ridx,
687                            struct Vmxnet3_RxDesc *dbuf, uint32_t *didx)
688 {
689     PCIDevice *d = PCI_DEVICE(s);
690 
691     Vmxnet3Ring *ring = &s->rxq_descr[qidx].rx_ring[ridx];
692     *didx = vmxnet3_ring_curr_cell_idx(ring);
693     vmxnet3_ring_read_curr_cell(d, ring, dbuf);
694     dbuf->addr = le64_to_cpu(dbuf->addr);
695     dbuf->val1 = le32_to_cpu(dbuf->val1);
696     dbuf->ext1 = le32_to_cpu(dbuf->ext1);
697 }
698 
699 static inline uint8_t
700 vmxnet3_get_rx_ring_gen(VMXNET3State *s, int qidx, int ridx)
701 {
702     return s->rxq_descr[qidx].rx_ring[ridx].gen;
703 }
704 
705 static inline hwaddr
706 vmxnet3_pop_rxc_descr(VMXNET3State *s, int qidx, uint32_t *descr_gen)
707 {
708     uint8_t ring_gen;
709     struct Vmxnet3_RxCompDesc rxcd;
710 
711     hwaddr daddr =
712         vmxnet3_ring_curr_cell_pa(&s->rxq_descr[qidx].comp_ring);
713 
714     pci_dma_read(PCI_DEVICE(s),
715                  daddr, &rxcd, sizeof(struct Vmxnet3_RxCompDesc));
716     rxcd.val1 = le32_to_cpu(rxcd.val1);
717     rxcd.val2 = le32_to_cpu(rxcd.val2);
718     rxcd.val3 = le32_to_cpu(rxcd.val3);
719     ring_gen = vmxnet3_ring_curr_gen(&s->rxq_descr[qidx].comp_ring);
720 
721     if (rxcd.gen != ring_gen) {
722         *descr_gen = ring_gen;
723         vmxnet3_inc_rx_completion_counter(s, qidx);
724         return daddr;
725     }
726 
727     return 0;
728 }
729 
730 static inline void
731 vmxnet3_revert_rxc_descr(VMXNET3State *s, int qidx)
732 {
733     vmxnet3_dec_rx_completion_counter(s, qidx);
734 }
735 
736 #define RXQ_IDX      (0)
737 #define RX_HEAD_BODY_RING (0)
738 #define RX_BODY_ONLY_RING (1)
739 
740 static bool
741 vmxnet3_get_next_head_rx_descr(VMXNET3State *s,
742                                struct Vmxnet3_RxDesc *descr_buf,
743                                uint32_t *descr_idx,
744                                uint32_t *ridx)
745 {
746     for (;;) {
747         uint32_t ring_gen;
748         vmxnet3_read_next_rx_descr(s, RXQ_IDX, RX_HEAD_BODY_RING,
749                                    descr_buf, descr_idx);
750 
751         /* If no more free descriptors - return */
752         ring_gen = vmxnet3_get_rx_ring_gen(s, RXQ_IDX, RX_HEAD_BODY_RING);
753         if (descr_buf->gen != ring_gen) {
754             return false;
755         }
756 
757         /* Only read after generation field verification */
758         smp_rmb();
759         /* Re-read to be sure we got the latest version */
760         vmxnet3_read_next_rx_descr(s, RXQ_IDX, RX_HEAD_BODY_RING,
761                                    descr_buf, descr_idx);
762 
763         /* Mark current descriptor as used/skipped */
764         vmxnet3_inc_rx_consumption_counter(s, RXQ_IDX, RX_HEAD_BODY_RING);
765 
766         /* If this is what we are looking for - return */
767         if (descr_buf->btype == VMXNET3_RXD_BTYPE_HEAD) {
768             *ridx = RX_HEAD_BODY_RING;
769             return true;
770         }
771     }
772 }
773 
774 static bool
775 vmxnet3_get_next_body_rx_descr(VMXNET3State *s,
776                                struct Vmxnet3_RxDesc *d,
777                                uint32_t *didx,
778                                uint32_t *ridx)
779 {
780     vmxnet3_read_next_rx_descr(s, RXQ_IDX, RX_HEAD_BODY_RING, d, didx);
781 
782     /* Try to find corresponding descriptor in head/body ring */
783     if (d->gen == vmxnet3_get_rx_ring_gen(s, RXQ_IDX, RX_HEAD_BODY_RING)) {
784         /* Only read after generation field verification */
785         smp_rmb();
786         /* Re-read to be sure we got the latest version */
787         vmxnet3_read_next_rx_descr(s, RXQ_IDX, RX_HEAD_BODY_RING, d, didx);
788         if (d->btype == VMXNET3_RXD_BTYPE_BODY) {
789             vmxnet3_inc_rx_consumption_counter(s, RXQ_IDX, RX_HEAD_BODY_RING);
790             *ridx = RX_HEAD_BODY_RING;
791             return true;
792         }
793     }
794 
795     /*
796      * If there is no free descriptors on head/body ring or next free
797      * descriptor is a head descriptor switch to body only ring
798      */
799     vmxnet3_read_next_rx_descr(s, RXQ_IDX, RX_BODY_ONLY_RING, d, didx);
800 
801     /* If no more free descriptors - return */
802     if (d->gen == vmxnet3_get_rx_ring_gen(s, RXQ_IDX, RX_BODY_ONLY_RING)) {
803         /* Only read after generation field verification */
804         smp_rmb();
805         /* Re-read to be sure we got the latest version */
806         vmxnet3_read_next_rx_descr(s, RXQ_IDX, RX_BODY_ONLY_RING, d, didx);
807         assert(d->btype == VMXNET3_RXD_BTYPE_BODY);
808         *ridx = RX_BODY_ONLY_RING;
809         vmxnet3_inc_rx_consumption_counter(s, RXQ_IDX, RX_BODY_ONLY_RING);
810         return true;
811     }
812 
813     return false;
814 }
815 
816 static inline bool
817 vmxnet3_get_next_rx_descr(VMXNET3State *s, bool is_head,
818                           struct Vmxnet3_RxDesc *descr_buf,
819                           uint32_t *descr_idx,
820                           uint32_t *ridx)
821 {
822     if (is_head || !s->rx_packets_compound) {
823         return vmxnet3_get_next_head_rx_descr(s, descr_buf, descr_idx, ridx);
824     } else {
825         return vmxnet3_get_next_body_rx_descr(s, descr_buf, descr_idx, ridx);
826     }
827 }
828 
829 /* In case packet was csum offloaded (either NEEDS_CSUM or DATA_VALID),
830  * the implementation always passes an RxCompDesc with a "Checksum
831  * calculated and found correct" to the OS (cnc=0 and tuc=1, see
832  * vmxnet3_rx_update_descr). This emulates the observed ESXi behavior.
833  *
834  * Therefore, if packet has the NEEDS_CSUM set, we must calculate
835  * and place a fully computed checksum into the tcp/udp header.
836  * Otherwise, the OS driver will receive a checksum-correct indication
837  * (CHECKSUM_UNNECESSARY), but with the actual tcp/udp checksum field
838  * having just the pseudo header csum value.
839  *
840  * While this is not a problem if packet is destined for local delivery,
841  * in the case the host OS performs forwarding, it will forward an
842  * incorrectly checksummed packet.
843  */
844 static void vmxnet3_rx_need_csum_calculate(struct NetRxPkt *pkt,
845                                            const void *pkt_data,
846                                            size_t pkt_len)
847 {
848     struct virtio_net_hdr *vhdr;
849     bool isip4, isip6, istcp, isudp;
850     uint8_t *data;
851     int len;
852 
853     if (!net_rx_pkt_has_virt_hdr(pkt)) {
854         return;
855     }
856 
857     vhdr = net_rx_pkt_get_vhdr(pkt);
858     if (!VMXNET_FLAG_IS_SET(vhdr->flags, VIRTIO_NET_HDR_F_NEEDS_CSUM)) {
859         return;
860     }
861 
862     net_rx_pkt_get_protocols(pkt, &isip4, &isip6, &isudp, &istcp);
863     if (!(isip4 || isip6) || !(istcp || isudp)) {
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 isip4, isip6, istcp, isudp;
892     struct virtio_net_hdr *vhdr;
893     uint8_t offload_type;
894 
895     if (net_rx_pkt_is_vlan_stripped(pkt)) {
896         rxcd->ts = 1;
897         rxcd->tci = net_rx_pkt_get_vlan_tag(pkt);
898     }
899 
900     if (!net_rx_pkt_has_virt_hdr(pkt)) {
901         goto nocsum;
902     }
903 
904     vhdr = net_rx_pkt_get_vhdr(pkt);
905     /*
906      * Checksum is valid when lower level tell so or when lower level
907      * requires checksum offload telling that packet produced/bridged
908      * locally and did travel over network after last checksum calculation
909      * or production
910      */
911     csum_ok = VMXNET_FLAG_IS_SET(vhdr->flags, VIRTIO_NET_HDR_F_DATA_VALID) ||
912               VMXNET_FLAG_IS_SET(vhdr->flags, VIRTIO_NET_HDR_F_NEEDS_CSUM);
913 
914     offload_type = vhdr->gso_type & ~VIRTIO_NET_HDR_GSO_ECN;
915     is_gso = (offload_type != VIRTIO_NET_HDR_GSO_NONE) ? 1 : 0;
916 
917     if (!csum_ok && !is_gso) {
918         goto nocsum;
919     }
920 
921     net_rx_pkt_get_protocols(pkt, &isip4, &isip6, &isudp, &istcp);
922     if ((!istcp && !isudp) || (!isip4 && !isip6)) {
923         goto nocsum;
924     }
925 
926     rxcd->cnc = 0;
927     rxcd->v4 = isip4 ? 1 : 0;
928     rxcd->v6 = isip6 ? 1 : 0;
929     rxcd->tcp = istcp ? 1 : 0;
930     rxcd->udp = isudp ? 1 : 0;
931     rxcd->fcs = rxcd->tuc = rxcd->ipc = 1;
932     return;
933 
934 nocsum:
935     rxcd->cnc = 1;
936     return;
937 }
938 
939 static void
940 vmxnet3_pci_dma_writev(PCIDevice *pci_dev,
941                        const struct iovec *iov,
942                        size_t start_iov_off,
943                        hwaddr target_addr,
944                        size_t bytes_to_copy)
945 {
946     size_t curr_off = 0;
947     size_t copied = 0;
948 
949     while (bytes_to_copy) {
950         if (start_iov_off < (curr_off + iov->iov_len)) {
951             size_t chunk_len =
952                 MIN((curr_off + iov->iov_len) - start_iov_off, bytes_to_copy);
953 
954             pci_dma_write(pci_dev, target_addr + copied,
955                           iov->iov_base + start_iov_off - curr_off,
956                           chunk_len);
957 
958             copied += chunk_len;
959             start_iov_off += chunk_len;
960             curr_off = start_iov_off;
961             bytes_to_copy -= chunk_len;
962         } else {
963             curr_off += iov->iov_len;
964         }
965         iov++;
966     }
967 }
968 
969 static void
970 vmxnet3_pci_dma_write_rxcd(PCIDevice *pcidev, dma_addr_t pa,
971                            struct Vmxnet3_RxCompDesc *rxcd)
972 {
973     rxcd->val1 = cpu_to_le32(rxcd->val1);
974     rxcd->val2 = cpu_to_le32(rxcd->val2);
975     rxcd->val3 = cpu_to_le32(rxcd->val3);
976     pci_dma_write(pcidev, pa, rxcd, sizeof(*rxcd));
977 }
978 
979 static bool
980 vmxnet3_indicate_packet(VMXNET3State *s)
981 {
982     struct Vmxnet3_RxDesc rxd;
983     PCIDevice *d = PCI_DEVICE(s);
984     bool is_head = true;
985     uint32_t rxd_idx;
986     uint32_t rx_ridx = 0;
987 
988     struct Vmxnet3_RxCompDesc rxcd;
989     uint32_t new_rxcd_gen = VMXNET3_INIT_GEN;
990     hwaddr new_rxcd_pa = 0;
991     hwaddr ready_rxcd_pa = 0;
992     struct iovec *data = net_rx_pkt_get_iovec(s->rx_pkt);
993     size_t bytes_copied = 0;
994     size_t bytes_left = net_rx_pkt_get_total_len(s->rx_pkt);
995     uint16_t num_frags = 0;
996     size_t chunk_size;
997 
998     net_rx_pkt_dump(s->rx_pkt);
999 
1000     while (bytes_left > 0) {
1001 
1002         /* cannot add more frags to packet */
1003         if (num_frags == s->max_rx_frags) {
1004             break;
1005         }
1006 
1007         new_rxcd_pa = vmxnet3_pop_rxc_descr(s, RXQ_IDX, &new_rxcd_gen);
1008         if (!new_rxcd_pa) {
1009             break;
1010         }
1011 
1012         if (!vmxnet3_get_next_rx_descr(s, is_head, &rxd, &rxd_idx, &rx_ridx)) {
1013             break;
1014         }
1015 
1016         chunk_size = MIN(bytes_left, rxd.len);
1017         vmxnet3_pci_dma_writev(d, data, bytes_copied, rxd.addr, chunk_size);
1018         bytes_copied += chunk_size;
1019         bytes_left -= chunk_size;
1020 
1021         vmxnet3_dump_rx_descr(&rxd);
1022 
1023         if (ready_rxcd_pa != 0) {
1024             vmxnet3_pci_dma_write_rxcd(d, ready_rxcd_pa, &rxcd);
1025         }
1026 
1027         memset(&rxcd, 0, sizeof(struct Vmxnet3_RxCompDesc));
1028         rxcd.rxdIdx = rxd_idx;
1029         rxcd.len = chunk_size;
1030         rxcd.sop = is_head;
1031         rxcd.gen = new_rxcd_gen;
1032         rxcd.rqID = RXQ_IDX + rx_ridx * s->rxq_num;
1033 
1034         if (bytes_left == 0) {
1035             vmxnet3_rx_update_descr(s->rx_pkt, &rxcd);
1036         }
1037 
1038         VMW_RIPRN("RX Completion descriptor: rxRing: %lu rxIdx %lu len %lu "
1039                   "sop %d csum_correct %lu",
1040                   (unsigned long) rx_ridx,
1041                   (unsigned long) rxcd.rxdIdx,
1042                   (unsigned long) rxcd.len,
1043                   (int) rxcd.sop,
1044                   (unsigned long) rxcd.tuc);
1045 
1046         is_head = false;
1047         ready_rxcd_pa = new_rxcd_pa;
1048         new_rxcd_pa = 0;
1049         num_frags++;
1050     }
1051 
1052     if (ready_rxcd_pa != 0) {
1053         rxcd.eop = 1;
1054         rxcd.err = (bytes_left != 0);
1055 
1056         vmxnet3_pci_dma_write_rxcd(d, ready_rxcd_pa, &rxcd);
1057 
1058         /* Flush RX descriptor changes */
1059         smp_wmb();
1060     }
1061 
1062     if (new_rxcd_pa != 0) {
1063         vmxnet3_revert_rxc_descr(s, RXQ_IDX);
1064     }
1065 
1066     vmxnet3_trigger_interrupt(s, s->rxq_descr[RXQ_IDX].intr_idx);
1067 
1068     if (bytes_left == 0) {
1069         vmxnet3_on_rx_done_update_stats(s, RXQ_IDX, VMXNET3_PKT_STATUS_OK);
1070         return true;
1071     } else if (num_frags == s->max_rx_frags) {
1072         vmxnet3_on_rx_done_update_stats(s, RXQ_IDX, VMXNET3_PKT_STATUS_ERROR);
1073         return false;
1074     } else {
1075         vmxnet3_on_rx_done_update_stats(s, RXQ_IDX,
1076                                         VMXNET3_PKT_STATUS_OUT_OF_BUF);
1077         return false;
1078     }
1079 }
1080 
1081 static void
1082 vmxnet3_io_bar0_write(void *opaque, hwaddr addr,
1083                       uint64_t val, unsigned size)
1084 {
1085     VMXNET3State *s = opaque;
1086 
1087     if (!s->device_active) {
1088         return;
1089     }
1090 
1091     if (VMW_IS_MULTIREG_ADDR(addr, VMXNET3_REG_TXPROD,
1092                         VMXNET3_DEVICE_MAX_TX_QUEUES, VMXNET3_REG_ALIGN)) {
1093         int tx_queue_idx =
1094             VMW_MULTIREG_IDX_BY_ADDR(addr, VMXNET3_REG_TXPROD,
1095                                      VMXNET3_REG_ALIGN);
1096         assert(tx_queue_idx <= s->txq_num);
1097         vmxnet3_process_tx_queue(s, tx_queue_idx);
1098         return;
1099     }
1100 
1101     if (VMW_IS_MULTIREG_ADDR(addr, VMXNET3_REG_IMR,
1102                         VMXNET3_MAX_INTRS, VMXNET3_REG_ALIGN)) {
1103         int l = VMW_MULTIREG_IDX_BY_ADDR(addr, VMXNET3_REG_IMR,
1104                                          VMXNET3_REG_ALIGN);
1105 
1106         VMW_CBPRN("Interrupt mask for line %d written: 0x%" PRIx64, l, val);
1107 
1108         vmxnet3_on_interrupt_mask_changed(s, l, val);
1109         return;
1110     }
1111 
1112     if (VMW_IS_MULTIREG_ADDR(addr, VMXNET3_REG_RXPROD,
1113                         VMXNET3_DEVICE_MAX_RX_QUEUES, VMXNET3_REG_ALIGN) ||
1114        VMW_IS_MULTIREG_ADDR(addr, VMXNET3_REG_RXPROD2,
1115                         VMXNET3_DEVICE_MAX_RX_QUEUES, VMXNET3_REG_ALIGN)) {
1116         return;
1117     }
1118 
1119     VMW_WRPRN("BAR0 unknown write [%" PRIx64 "] = %" PRIx64 ", size %d",
1120               (uint64_t) addr, val, size);
1121 }
1122 
1123 static uint64_t
1124 vmxnet3_io_bar0_read(void *opaque, hwaddr addr, unsigned size)
1125 {
1126     VMXNET3State *s = opaque;
1127 
1128     if (VMW_IS_MULTIREG_ADDR(addr, VMXNET3_REG_IMR,
1129                         VMXNET3_MAX_INTRS, VMXNET3_REG_ALIGN)) {
1130         int l = VMW_MULTIREG_IDX_BY_ADDR(addr, VMXNET3_REG_IMR,
1131                                          VMXNET3_REG_ALIGN);
1132         return s->interrupt_states[l].is_masked;
1133     }
1134 
1135     VMW_CBPRN("BAR0 unknown read [%" PRIx64 "], size %d", addr, size);
1136     return 0;
1137 }
1138 
1139 static void vmxnet3_reset_interrupt_states(VMXNET3State *s)
1140 {
1141     int i;
1142     for (i = 0; i < ARRAY_SIZE(s->interrupt_states); i++) {
1143         s->interrupt_states[i].is_asserted = false;
1144         s->interrupt_states[i].is_pending = false;
1145         s->interrupt_states[i].is_masked = true;
1146     }
1147 }
1148 
1149 static void vmxnet3_reset_mac(VMXNET3State *s)
1150 {
1151     memcpy(&s->conf.macaddr.a, &s->perm_mac.a, sizeof(s->perm_mac.a));
1152     VMW_CFPRN("MAC address set to: " MAC_FMT, MAC_ARG(s->conf.macaddr.a));
1153 }
1154 
1155 static void vmxnet3_deactivate_device(VMXNET3State *s)
1156 {
1157     if (s->device_active) {
1158         VMW_CBPRN("Deactivating vmxnet3...");
1159         net_tx_pkt_reset(s->tx_pkt);
1160         net_tx_pkt_uninit(s->tx_pkt);
1161         net_rx_pkt_uninit(s->rx_pkt);
1162         s->device_active = false;
1163     }
1164 }
1165 
1166 static void vmxnet3_reset(VMXNET3State *s)
1167 {
1168     VMW_CBPRN("Resetting vmxnet3...");
1169 
1170     vmxnet3_deactivate_device(s);
1171     vmxnet3_reset_interrupt_states(s);
1172     s->drv_shmem = 0;
1173     s->tx_sop = true;
1174     s->skip_current_tx_pkt = false;
1175 }
1176 
1177 static void vmxnet3_update_rx_mode(VMXNET3State *s)
1178 {
1179     PCIDevice *d = PCI_DEVICE(s);
1180 
1181     s->rx_mode = VMXNET3_READ_DRV_SHARED32(d, s->drv_shmem,
1182                                            devRead.rxFilterConf.rxMode);
1183     VMW_CFPRN("RX mode: 0x%08X", s->rx_mode);
1184 }
1185 
1186 static void vmxnet3_update_vlan_filters(VMXNET3State *s)
1187 {
1188     int i;
1189     PCIDevice *d = PCI_DEVICE(s);
1190 
1191     /* Copy configuration from shared memory */
1192     VMXNET3_READ_DRV_SHARED(d, s->drv_shmem,
1193                             devRead.rxFilterConf.vfTable,
1194                             s->vlan_table,
1195                             sizeof(s->vlan_table));
1196 
1197     /* Invert byte order when needed */
1198     for (i = 0; i < ARRAY_SIZE(s->vlan_table); i++) {
1199         s->vlan_table[i] = le32_to_cpu(s->vlan_table[i]);
1200     }
1201 
1202     /* Dump configuration for debugging purposes */
1203     VMW_CFPRN("Configured VLANs:");
1204     for (i = 0; i < sizeof(s->vlan_table) * 8; i++) {
1205         if (VMXNET3_VFTABLE_ENTRY_IS_SET(s->vlan_table, i)) {
1206             VMW_CFPRN("\tVLAN %d is present", i);
1207         }
1208     }
1209 }
1210 
1211 static void vmxnet3_update_mcast_filters(VMXNET3State *s)
1212 {
1213     PCIDevice *d = PCI_DEVICE(s);
1214 
1215     uint16_t list_bytes =
1216         VMXNET3_READ_DRV_SHARED16(d, s->drv_shmem,
1217                                   devRead.rxFilterConf.mfTableLen);
1218 
1219     s->mcast_list_len = list_bytes / sizeof(s->mcast_list[0]);
1220 
1221     s->mcast_list = g_realloc(s->mcast_list, list_bytes);
1222     if (!s->mcast_list) {
1223         if (s->mcast_list_len == 0) {
1224             VMW_CFPRN("Current multicast list is empty");
1225         } else {
1226             VMW_ERPRN("Failed to allocate multicast list of %d elements",
1227                       s->mcast_list_len);
1228         }
1229         s->mcast_list_len = 0;
1230     } else {
1231         int i;
1232         hwaddr mcast_list_pa =
1233             VMXNET3_READ_DRV_SHARED64(d, s->drv_shmem,
1234                                       devRead.rxFilterConf.mfTablePA);
1235 
1236         pci_dma_read(d, mcast_list_pa, s->mcast_list, list_bytes);
1237 
1238         VMW_CFPRN("Current multicast list len is %d:", s->mcast_list_len);
1239         for (i = 0; i < s->mcast_list_len; i++) {
1240             VMW_CFPRN("\t" MAC_FMT, MAC_ARG(s->mcast_list[i].a));
1241         }
1242     }
1243 }
1244 
1245 static void vmxnet3_setup_rx_filtering(VMXNET3State *s)
1246 {
1247     vmxnet3_update_rx_mode(s);
1248     vmxnet3_update_vlan_filters(s);
1249     vmxnet3_update_mcast_filters(s);
1250 }
1251 
1252 static uint32_t vmxnet3_get_interrupt_config(VMXNET3State *s)
1253 {
1254     uint32_t interrupt_mode = VMXNET3_IT_AUTO | (VMXNET3_IMM_AUTO << 2);
1255     VMW_CFPRN("Interrupt config is 0x%X", interrupt_mode);
1256     return interrupt_mode;
1257 }
1258 
1259 static void vmxnet3_fill_stats(VMXNET3State *s)
1260 {
1261     int i;
1262     PCIDevice *d = PCI_DEVICE(s);
1263 
1264     if (!s->device_active)
1265         return;
1266 
1267     for (i = 0; i < s->txq_num; i++) {
1268         pci_dma_write(d,
1269                       s->txq_descr[i].tx_stats_pa,
1270                       &s->txq_descr[i].txq_stats,
1271                       sizeof(s->txq_descr[i].txq_stats));
1272     }
1273 
1274     for (i = 0; i < s->rxq_num; i++) {
1275         pci_dma_write(d,
1276                       s->rxq_descr[i].rx_stats_pa,
1277                       &s->rxq_descr[i].rxq_stats,
1278                       sizeof(s->rxq_descr[i].rxq_stats));
1279     }
1280 }
1281 
1282 static void vmxnet3_adjust_by_guest_type(VMXNET3State *s)
1283 {
1284     struct Vmxnet3_GOSInfo gos;
1285     PCIDevice *d = PCI_DEVICE(s);
1286 
1287     VMXNET3_READ_DRV_SHARED(d, s->drv_shmem, devRead.misc.driverInfo.gos,
1288                             &gos, sizeof(gos));
1289     s->rx_packets_compound =
1290         (gos.gosType == VMXNET3_GOS_TYPE_WIN) ? false : true;
1291 
1292     VMW_CFPRN("Guest type specifics: RXCOMPOUND: %d", s->rx_packets_compound);
1293 }
1294 
1295 static void
1296 vmxnet3_dump_conf_descr(const char *name,
1297                         struct Vmxnet3_VariableLenConfDesc *pm_descr)
1298 {
1299     VMW_CFPRN("%s descriptor dump: Version %u, Length %u",
1300               name, pm_descr->confVer, pm_descr->confLen);
1301 
1302 };
1303 
1304 static void vmxnet3_update_pm_state(VMXNET3State *s)
1305 {
1306     struct Vmxnet3_VariableLenConfDesc pm_descr;
1307     PCIDevice *d = PCI_DEVICE(s);
1308 
1309     pm_descr.confLen =
1310         VMXNET3_READ_DRV_SHARED32(d, s->drv_shmem, devRead.pmConfDesc.confLen);
1311     pm_descr.confVer =
1312         VMXNET3_READ_DRV_SHARED32(d, s->drv_shmem, devRead.pmConfDesc.confVer);
1313     pm_descr.confPA =
1314         VMXNET3_READ_DRV_SHARED64(d, s->drv_shmem, devRead.pmConfDesc.confPA);
1315 
1316     vmxnet3_dump_conf_descr("PM State", &pm_descr);
1317 }
1318 
1319 static void vmxnet3_update_features(VMXNET3State *s)
1320 {
1321     uint32_t guest_features;
1322     int rxcso_supported;
1323     PCIDevice *d = PCI_DEVICE(s);
1324 
1325     guest_features = VMXNET3_READ_DRV_SHARED32(d, s->drv_shmem,
1326                                                devRead.misc.uptFeatures);
1327 
1328     rxcso_supported = VMXNET_FLAG_IS_SET(guest_features, UPT1_F_RXCSUM);
1329     s->rx_vlan_stripping = VMXNET_FLAG_IS_SET(guest_features, UPT1_F_RXVLAN);
1330     s->lro_supported = VMXNET_FLAG_IS_SET(guest_features, UPT1_F_LRO);
1331 
1332     VMW_CFPRN("Features configuration: LRO: %d, RXCSUM: %d, VLANSTRIP: %d",
1333               s->lro_supported, rxcso_supported,
1334               s->rx_vlan_stripping);
1335     if (s->peer_has_vhdr) {
1336         qemu_set_offload(qemu_get_queue(s->nic)->peer,
1337                          rxcso_supported,
1338                          s->lro_supported,
1339                          s->lro_supported,
1340                          0,
1341                          0);
1342     }
1343 }
1344 
1345 static bool vmxnet3_verify_intx(VMXNET3State *s, int intx)
1346 {
1347     return s->msix_used || msi_enabled(PCI_DEVICE(s))
1348         || intx == pci_get_byte(s->parent_obj.config + PCI_INTERRUPT_PIN) - 1;
1349 }
1350 
1351 static void vmxnet3_validate_interrupt_idx(bool is_msix, int idx)
1352 {
1353     int max_ints = is_msix ? VMXNET3_MAX_INTRS : VMXNET3_MAX_NMSIX_INTRS;
1354     if (idx >= max_ints) {
1355         hw_error("Bad interrupt index: %d\n", idx);
1356     }
1357 }
1358 
1359 static void vmxnet3_validate_interrupts(VMXNET3State *s)
1360 {
1361     int i;
1362 
1363     VMW_CFPRN("Verifying event interrupt index (%d)", s->event_int_idx);
1364     vmxnet3_validate_interrupt_idx(s->msix_used, s->event_int_idx);
1365 
1366     for (i = 0; i < s->txq_num; i++) {
1367         int idx = s->txq_descr[i].intr_idx;
1368         VMW_CFPRN("Verifying TX queue %d interrupt index (%d)", i, idx);
1369         vmxnet3_validate_interrupt_idx(s->msix_used, idx);
1370     }
1371 
1372     for (i = 0; i < s->rxq_num; i++) {
1373         int idx = s->rxq_descr[i].intr_idx;
1374         VMW_CFPRN("Verifying RX queue %d interrupt index (%d)", i, idx);
1375         vmxnet3_validate_interrupt_idx(s->msix_used, idx);
1376     }
1377 }
1378 
1379 static void vmxnet3_validate_queues(VMXNET3State *s)
1380 {
1381     /*
1382     * txq_num and rxq_num are total number of queues
1383     * configured by guest. These numbers must not
1384     * exceed corresponding maximal values.
1385     */
1386 
1387     if (s->txq_num > VMXNET3_DEVICE_MAX_TX_QUEUES) {
1388         hw_error("Bad TX queues number: %d\n", s->txq_num);
1389     }
1390 
1391     if (s->rxq_num > VMXNET3_DEVICE_MAX_RX_QUEUES) {
1392         hw_error("Bad RX queues number: %d\n", s->rxq_num);
1393     }
1394 }
1395 
1396 static void vmxnet3_activate_device(VMXNET3State *s)
1397 {
1398     int i;
1399     static const uint32_t VMXNET3_DEF_TX_THRESHOLD = 1;
1400     PCIDevice *d = PCI_DEVICE(s);
1401     hwaddr qdescr_table_pa;
1402     uint64_t pa;
1403     uint32_t size;
1404 
1405     /* Verify configuration consistency */
1406     if (!vmxnet3_verify_driver_magic(d, s->drv_shmem)) {
1407         VMW_ERPRN("Device configuration received from driver is invalid");
1408         return;
1409     }
1410 
1411     /* Verify if device is active */
1412     if (s->device_active) {
1413         VMW_CFPRN("Vmxnet3 device is active");
1414         return;
1415     }
1416 
1417     vmxnet3_adjust_by_guest_type(s);
1418     vmxnet3_update_features(s);
1419     vmxnet3_update_pm_state(s);
1420     vmxnet3_setup_rx_filtering(s);
1421     /* Cache fields from shared memory */
1422     s->mtu = VMXNET3_READ_DRV_SHARED32(d, s->drv_shmem, devRead.misc.mtu);
1423     VMW_CFPRN("MTU is %u", s->mtu);
1424 
1425     s->max_rx_frags =
1426         VMXNET3_READ_DRV_SHARED16(d, s->drv_shmem, devRead.misc.maxNumRxSG);
1427 
1428     if (s->max_rx_frags == 0) {
1429         s->max_rx_frags = 1;
1430     }
1431 
1432     VMW_CFPRN("Max RX fragments is %u", s->max_rx_frags);
1433 
1434     s->event_int_idx =
1435         VMXNET3_READ_DRV_SHARED8(d, s->drv_shmem, devRead.intrConf.eventIntrIdx);
1436     assert(vmxnet3_verify_intx(s, s->event_int_idx));
1437     VMW_CFPRN("Events interrupt line is %u", s->event_int_idx);
1438 
1439     s->auto_int_masking =
1440         VMXNET3_READ_DRV_SHARED8(d, s->drv_shmem, devRead.intrConf.autoMask);
1441     VMW_CFPRN("Automatic interrupt masking is %d", (int)s->auto_int_masking);
1442 
1443     s->txq_num =
1444         VMXNET3_READ_DRV_SHARED8(d, s->drv_shmem, devRead.misc.numTxQueues);
1445     s->rxq_num =
1446         VMXNET3_READ_DRV_SHARED8(d, s->drv_shmem, devRead.misc.numRxQueues);
1447 
1448     VMW_CFPRN("Number of TX/RX queues %u/%u", s->txq_num, s->rxq_num);
1449     vmxnet3_validate_queues(s);
1450 
1451     qdescr_table_pa =
1452         VMXNET3_READ_DRV_SHARED64(d, s->drv_shmem, devRead.misc.queueDescPA);
1453     VMW_CFPRN("TX queues descriptors table is at 0x%" PRIx64, qdescr_table_pa);
1454 
1455     /*
1456      * Worst-case scenario is a packet that holds all TX rings space so
1457      * we calculate total size of all TX rings for max TX fragments number
1458      */
1459     s->max_tx_frags = 0;
1460 
1461     /* TX queues */
1462     for (i = 0; i < s->txq_num; i++) {
1463         hwaddr qdescr_pa =
1464             qdescr_table_pa + i * sizeof(struct Vmxnet3_TxQueueDesc);
1465 
1466         /* Read interrupt number for this TX queue */
1467         s->txq_descr[i].intr_idx =
1468             VMXNET3_READ_TX_QUEUE_DESCR8(d, qdescr_pa, conf.intrIdx);
1469         assert(vmxnet3_verify_intx(s, s->txq_descr[i].intr_idx));
1470 
1471         VMW_CFPRN("TX Queue %d interrupt: %d", i, s->txq_descr[i].intr_idx);
1472 
1473         /* Read rings memory locations for TX queues */
1474         pa = VMXNET3_READ_TX_QUEUE_DESCR64(d, qdescr_pa, conf.txRingBasePA);
1475         size = VMXNET3_READ_TX_QUEUE_DESCR32(d, qdescr_pa, conf.txRingSize);
1476 
1477         vmxnet3_ring_init(d, &s->txq_descr[i].tx_ring, pa, size,
1478                           sizeof(struct Vmxnet3_TxDesc), false);
1479         VMXNET3_RING_DUMP(VMW_CFPRN, "TX", i, &s->txq_descr[i].tx_ring);
1480 
1481         s->max_tx_frags += size;
1482 
1483         /* TXC ring */
1484         pa = VMXNET3_READ_TX_QUEUE_DESCR64(d, qdescr_pa, conf.compRingBasePA);
1485         size = VMXNET3_READ_TX_QUEUE_DESCR32(d, qdescr_pa, conf.compRingSize);
1486         vmxnet3_ring_init(d, &s->txq_descr[i].comp_ring, pa, size,
1487                           sizeof(struct Vmxnet3_TxCompDesc), true);
1488         VMXNET3_RING_DUMP(VMW_CFPRN, "TXC", i, &s->txq_descr[i].comp_ring);
1489 
1490         s->txq_descr[i].tx_stats_pa =
1491             qdescr_pa + offsetof(struct Vmxnet3_TxQueueDesc, stats);
1492 
1493         memset(&s->txq_descr[i].txq_stats, 0,
1494                sizeof(s->txq_descr[i].txq_stats));
1495 
1496         /* Fill device-managed parameters for queues */
1497         VMXNET3_WRITE_TX_QUEUE_DESCR32(d, qdescr_pa,
1498                                        ctrl.txThreshold,
1499                                        VMXNET3_DEF_TX_THRESHOLD);
1500     }
1501 
1502     /* Preallocate TX packet wrapper */
1503     VMW_CFPRN("Max TX fragments is %u", s->max_tx_frags);
1504     net_tx_pkt_init(&s->tx_pkt, PCI_DEVICE(s),
1505                     s->max_tx_frags, s->peer_has_vhdr);
1506     net_rx_pkt_init(&s->rx_pkt, s->peer_has_vhdr);
1507 
1508     /* Read rings memory locations for RX queues */
1509     for (i = 0; i < s->rxq_num; i++) {
1510         int j;
1511         hwaddr qd_pa =
1512             qdescr_table_pa + s->txq_num * sizeof(struct Vmxnet3_TxQueueDesc) +
1513             i * sizeof(struct Vmxnet3_RxQueueDesc);
1514 
1515         /* Read interrupt number for this RX queue */
1516         s->rxq_descr[i].intr_idx =
1517             VMXNET3_READ_TX_QUEUE_DESCR8(d, qd_pa, conf.intrIdx);
1518         assert(vmxnet3_verify_intx(s, s->rxq_descr[i].intr_idx));
1519 
1520         VMW_CFPRN("RX Queue %d interrupt: %d", i, s->rxq_descr[i].intr_idx);
1521 
1522         /* Read rings memory locations */
1523         for (j = 0; j < VMXNET3_RX_RINGS_PER_QUEUE; j++) {
1524             /* RX rings */
1525             pa = VMXNET3_READ_RX_QUEUE_DESCR64(d, qd_pa, conf.rxRingBasePA[j]);
1526             size = VMXNET3_READ_RX_QUEUE_DESCR32(d, qd_pa, conf.rxRingSize[j]);
1527             vmxnet3_ring_init(d, &s->rxq_descr[i].rx_ring[j], pa, size,
1528                               sizeof(struct Vmxnet3_RxDesc), false);
1529             VMW_CFPRN("RX queue %d:%d: Base: %" PRIx64 ", Size: %d",
1530                       i, j, pa, size);
1531         }
1532 
1533         /* RXC ring */
1534         pa = VMXNET3_READ_RX_QUEUE_DESCR64(d, qd_pa, conf.compRingBasePA);
1535         size = VMXNET3_READ_RX_QUEUE_DESCR32(d, qd_pa, conf.compRingSize);
1536         vmxnet3_ring_init(d, &s->rxq_descr[i].comp_ring, pa, size,
1537                           sizeof(struct Vmxnet3_RxCompDesc), true);
1538         VMW_CFPRN("RXC queue %d: Base: %" PRIx64 ", Size: %d", i, pa, size);
1539 
1540         s->rxq_descr[i].rx_stats_pa =
1541             qd_pa + offsetof(struct Vmxnet3_RxQueueDesc, stats);
1542         memset(&s->rxq_descr[i].rxq_stats, 0,
1543                sizeof(s->rxq_descr[i].rxq_stats));
1544     }
1545 
1546     vmxnet3_validate_interrupts(s);
1547 
1548     /* Make sure everything is in place before device activation */
1549     smp_wmb();
1550 
1551     vmxnet3_reset_mac(s);
1552 
1553     s->device_active = true;
1554 }
1555 
1556 static void vmxnet3_handle_command(VMXNET3State *s, uint64_t cmd)
1557 {
1558     s->last_command = cmd;
1559 
1560     switch (cmd) {
1561     case VMXNET3_CMD_GET_PERM_MAC_HI:
1562         VMW_CBPRN("Set: Get upper part of permanent MAC");
1563         break;
1564 
1565     case VMXNET3_CMD_GET_PERM_MAC_LO:
1566         VMW_CBPRN("Set: Get lower part of permanent MAC");
1567         break;
1568 
1569     case VMXNET3_CMD_GET_STATS:
1570         VMW_CBPRN("Set: Get device statistics");
1571         vmxnet3_fill_stats(s);
1572         break;
1573 
1574     case VMXNET3_CMD_ACTIVATE_DEV:
1575         VMW_CBPRN("Set: Activating vmxnet3 device");
1576         vmxnet3_activate_device(s);
1577         break;
1578 
1579     case VMXNET3_CMD_UPDATE_RX_MODE:
1580         VMW_CBPRN("Set: Update rx mode");
1581         vmxnet3_update_rx_mode(s);
1582         break;
1583 
1584     case VMXNET3_CMD_UPDATE_VLAN_FILTERS:
1585         VMW_CBPRN("Set: Update VLAN filters");
1586         vmxnet3_update_vlan_filters(s);
1587         break;
1588 
1589     case VMXNET3_CMD_UPDATE_MAC_FILTERS:
1590         VMW_CBPRN("Set: Update MAC filters");
1591         vmxnet3_update_mcast_filters(s);
1592         break;
1593 
1594     case VMXNET3_CMD_UPDATE_FEATURE:
1595         VMW_CBPRN("Set: Update features");
1596         vmxnet3_update_features(s);
1597         break;
1598 
1599     case VMXNET3_CMD_UPDATE_PMCFG:
1600         VMW_CBPRN("Set: Update power management config");
1601         vmxnet3_update_pm_state(s);
1602         break;
1603 
1604     case VMXNET3_CMD_GET_LINK:
1605         VMW_CBPRN("Set: Get link");
1606         break;
1607 
1608     case VMXNET3_CMD_RESET_DEV:
1609         VMW_CBPRN("Set: Reset device");
1610         vmxnet3_reset(s);
1611         break;
1612 
1613     case VMXNET3_CMD_QUIESCE_DEV:
1614         VMW_CBPRN("Set: VMXNET3_CMD_QUIESCE_DEV - deactivate the device");
1615         vmxnet3_deactivate_device(s);
1616         break;
1617 
1618     case VMXNET3_CMD_GET_CONF_INTR:
1619         VMW_CBPRN("Set: VMXNET3_CMD_GET_CONF_INTR - interrupt configuration");
1620         break;
1621 
1622     case VMXNET3_CMD_GET_ADAPTIVE_RING_INFO:
1623         VMW_CBPRN("Set: VMXNET3_CMD_GET_ADAPTIVE_RING_INFO - "
1624                   "adaptive ring info flags");
1625         break;
1626 
1627     case VMXNET3_CMD_GET_DID_LO:
1628         VMW_CBPRN("Set: Get lower part of device ID");
1629         break;
1630 
1631     case VMXNET3_CMD_GET_DID_HI:
1632         VMW_CBPRN("Set: Get upper part of device ID");
1633         break;
1634 
1635     case VMXNET3_CMD_GET_DEV_EXTRA_INFO:
1636         VMW_CBPRN("Set: Get device extra info");
1637         break;
1638 
1639     default:
1640         VMW_CBPRN("Received unknown command: %" PRIx64, cmd);
1641         break;
1642     }
1643 }
1644 
1645 static uint64_t vmxnet3_get_command_status(VMXNET3State *s)
1646 {
1647     uint64_t ret;
1648 
1649     switch (s->last_command) {
1650     case VMXNET3_CMD_ACTIVATE_DEV:
1651         ret = (s->device_active) ? 0 : 1;
1652         VMW_CFPRN("Device active: %" PRIx64, ret);
1653         break;
1654 
1655     case VMXNET3_CMD_RESET_DEV:
1656     case VMXNET3_CMD_QUIESCE_DEV:
1657     case VMXNET3_CMD_GET_QUEUE_STATUS:
1658     case VMXNET3_CMD_GET_DEV_EXTRA_INFO:
1659         ret = 0;
1660         break;
1661 
1662     case VMXNET3_CMD_GET_LINK:
1663         ret = s->link_status_and_speed;
1664         VMW_CFPRN("Link and speed: %" PRIx64, ret);
1665         break;
1666 
1667     case VMXNET3_CMD_GET_PERM_MAC_LO:
1668         ret = vmxnet3_get_mac_low(&s->perm_mac);
1669         break;
1670 
1671     case VMXNET3_CMD_GET_PERM_MAC_HI:
1672         ret = vmxnet3_get_mac_high(&s->perm_mac);
1673         break;
1674 
1675     case VMXNET3_CMD_GET_CONF_INTR:
1676         ret = vmxnet3_get_interrupt_config(s);
1677         break;
1678 
1679     case VMXNET3_CMD_GET_ADAPTIVE_RING_INFO:
1680         ret = VMXNET3_DISABLE_ADAPTIVE_RING;
1681         break;
1682 
1683     case VMXNET3_CMD_GET_DID_LO:
1684         ret = PCI_DEVICE_ID_VMWARE_VMXNET3;
1685         break;
1686 
1687     case VMXNET3_CMD_GET_DID_HI:
1688         ret = VMXNET3_DEVICE_REVISION;
1689         break;
1690 
1691     default:
1692         VMW_WRPRN("Received request for unknown command: %x", s->last_command);
1693         ret = 0;
1694         break;
1695     }
1696 
1697     return ret;
1698 }
1699 
1700 static void vmxnet3_set_events(VMXNET3State *s, uint32_t val)
1701 {
1702     uint32_t events;
1703     PCIDevice *d = PCI_DEVICE(s);
1704 
1705     VMW_CBPRN("Setting events: 0x%x", val);
1706     events = VMXNET3_READ_DRV_SHARED32(d, s->drv_shmem, ecr) | val;
1707     VMXNET3_WRITE_DRV_SHARED32(d, s->drv_shmem, ecr, events);
1708 }
1709 
1710 static void vmxnet3_ack_events(VMXNET3State *s, uint32_t val)
1711 {
1712     PCIDevice *d = PCI_DEVICE(s);
1713     uint32_t events;
1714 
1715     VMW_CBPRN("Clearing events: 0x%x", val);
1716     events = VMXNET3_READ_DRV_SHARED32(d, s->drv_shmem, ecr) & ~val;
1717     VMXNET3_WRITE_DRV_SHARED32(d, s->drv_shmem, ecr, events);
1718 }
1719 
1720 static void
1721 vmxnet3_io_bar1_write(void *opaque,
1722                       hwaddr addr,
1723                       uint64_t val,
1724                       unsigned size)
1725 {
1726     VMXNET3State *s = opaque;
1727 
1728     switch (addr) {
1729     /* Vmxnet3 Revision Report Selection */
1730     case VMXNET3_REG_VRRS:
1731         VMW_CBPRN("Write BAR1 [VMXNET3_REG_VRRS] = %" PRIx64 ", size %d",
1732                   val, size);
1733         break;
1734 
1735     /* UPT Version Report Selection */
1736     case VMXNET3_REG_UVRS:
1737         VMW_CBPRN("Write BAR1 [VMXNET3_REG_UVRS] = %" PRIx64 ", size %d",
1738                   val, size);
1739         break;
1740 
1741     /* Driver Shared Address Low */
1742     case VMXNET3_REG_DSAL:
1743         VMW_CBPRN("Write BAR1 [VMXNET3_REG_DSAL] = %" PRIx64 ", size %d",
1744                   val, size);
1745         /*
1746          * Guest driver will first write the low part of the shared
1747          * memory address. We save it to temp variable and set the
1748          * shared address only after we get the high part
1749          */
1750         if (val == 0) {
1751             vmxnet3_deactivate_device(s);
1752         }
1753         s->temp_shared_guest_driver_memory = val;
1754         s->drv_shmem = 0;
1755         break;
1756 
1757     /* Driver Shared Address High */
1758     case VMXNET3_REG_DSAH:
1759         VMW_CBPRN("Write BAR1 [VMXNET3_REG_DSAH] = %" PRIx64 ", size %d",
1760                   val, size);
1761         /*
1762          * Set the shared memory between guest driver and device.
1763          * We already should have low address part.
1764          */
1765         s->drv_shmem = s->temp_shared_guest_driver_memory | (val << 32);
1766         break;
1767 
1768     /* Command */
1769     case VMXNET3_REG_CMD:
1770         VMW_CBPRN("Write BAR1 [VMXNET3_REG_CMD] = %" PRIx64 ", size %d",
1771                   val, size);
1772         vmxnet3_handle_command(s, val);
1773         break;
1774 
1775     /* MAC Address Low */
1776     case VMXNET3_REG_MACL:
1777         VMW_CBPRN("Write BAR1 [VMXNET3_REG_MACL] = %" PRIx64 ", size %d",
1778                   val, size);
1779         s->temp_mac = val;
1780         break;
1781 
1782     /* MAC Address High */
1783     case VMXNET3_REG_MACH:
1784         VMW_CBPRN("Write BAR1 [VMXNET3_REG_MACH] = %" PRIx64 ", size %d",
1785                   val, size);
1786         vmxnet3_set_variable_mac(s, val, s->temp_mac);
1787         break;
1788 
1789     /* Interrupt Cause Register */
1790     case VMXNET3_REG_ICR:
1791         VMW_CBPRN("Write BAR1 [VMXNET3_REG_ICR] = %" PRIx64 ", size %d",
1792                   val, size);
1793         g_assert_not_reached();
1794         break;
1795 
1796     /* Event Cause Register */
1797     case VMXNET3_REG_ECR:
1798         VMW_CBPRN("Write BAR1 [VMXNET3_REG_ECR] = %" PRIx64 ", size %d",
1799                   val, size);
1800         vmxnet3_ack_events(s, val);
1801         break;
1802 
1803     default:
1804         VMW_CBPRN("Unknown Write to BAR1 [%" PRIx64 "] = %" PRIx64 ", size %d",
1805                   addr, val, size);
1806         break;
1807     }
1808 }
1809 
1810 static uint64_t
1811 vmxnet3_io_bar1_read(void *opaque, hwaddr addr, unsigned size)
1812 {
1813         VMXNET3State *s = opaque;
1814         uint64_t ret = 0;
1815 
1816         switch (addr) {
1817         /* Vmxnet3 Revision Report Selection */
1818         case VMXNET3_REG_VRRS:
1819             VMW_CBPRN("Read BAR1 [VMXNET3_REG_VRRS], size %d", size);
1820             ret = VMXNET3_DEVICE_REVISION;
1821             break;
1822 
1823         /* UPT Version Report Selection */
1824         case VMXNET3_REG_UVRS:
1825             VMW_CBPRN("Read BAR1 [VMXNET3_REG_UVRS], size %d", size);
1826             ret = VMXNET3_UPT_REVISION;
1827             break;
1828 
1829         /* Command */
1830         case VMXNET3_REG_CMD:
1831             VMW_CBPRN("Read BAR1 [VMXNET3_REG_CMD], size %d", size);
1832             ret = vmxnet3_get_command_status(s);
1833             break;
1834 
1835         /* MAC Address Low */
1836         case VMXNET3_REG_MACL:
1837             VMW_CBPRN("Read BAR1 [VMXNET3_REG_MACL], size %d", size);
1838             ret = vmxnet3_get_mac_low(&s->conf.macaddr);
1839             break;
1840 
1841         /* MAC Address High */
1842         case VMXNET3_REG_MACH:
1843             VMW_CBPRN("Read BAR1 [VMXNET3_REG_MACH], size %d", size);
1844             ret = vmxnet3_get_mac_high(&s->conf.macaddr);
1845             break;
1846 
1847         /*
1848          * Interrupt Cause Register
1849          * Used for legacy interrupts only so interrupt index always 0
1850          */
1851         case VMXNET3_REG_ICR:
1852             VMW_CBPRN("Read BAR1 [VMXNET3_REG_ICR], size %d", size);
1853             if (vmxnet3_interrupt_asserted(s, 0)) {
1854                 vmxnet3_clear_interrupt(s, 0);
1855                 ret = true;
1856             } else {
1857                 ret = false;
1858             }
1859             break;
1860 
1861         default:
1862             VMW_CBPRN("Unknow read BAR1[%" PRIx64 "], %d bytes", addr, size);
1863             break;
1864         }
1865 
1866         return ret;
1867 }
1868 
1869 static int
1870 vmxnet3_can_receive(NetClientState *nc)
1871 {
1872     VMXNET3State *s = qemu_get_nic_opaque(nc);
1873     return s->device_active &&
1874            VMXNET_FLAG_IS_SET(s->link_status_and_speed, VMXNET3_LINK_STATUS_UP);
1875 }
1876 
1877 static inline bool
1878 vmxnet3_is_registered_vlan(VMXNET3State *s, const void *data)
1879 {
1880     uint16_t vlan_tag = eth_get_pkt_tci(data) & VLAN_VID_MASK;
1881     if (IS_SPECIAL_VLAN_ID(vlan_tag)) {
1882         return true;
1883     }
1884 
1885     return VMXNET3_VFTABLE_ENTRY_IS_SET(s->vlan_table, vlan_tag);
1886 }
1887 
1888 static bool
1889 vmxnet3_is_allowed_mcast_group(VMXNET3State *s, const uint8_t *group_mac)
1890 {
1891     int i;
1892     for (i = 0; i < s->mcast_list_len; i++) {
1893         if (!memcmp(group_mac, s->mcast_list[i].a, sizeof(s->mcast_list[i]))) {
1894             return true;
1895         }
1896     }
1897     return false;
1898 }
1899 
1900 static bool
1901 vmxnet3_rx_filter_may_indicate(VMXNET3State *s, const void *data,
1902     size_t size)
1903 {
1904     struct eth_header *ehdr = PKT_GET_ETH_HDR(data);
1905 
1906     if (VMXNET_FLAG_IS_SET(s->rx_mode, VMXNET3_RXM_PROMISC)) {
1907         return true;
1908     }
1909 
1910     if (!vmxnet3_is_registered_vlan(s, data)) {
1911         return false;
1912     }
1913 
1914     switch (net_rx_pkt_get_packet_type(s->rx_pkt)) {
1915     case ETH_PKT_UCAST:
1916         if (!VMXNET_FLAG_IS_SET(s->rx_mode, VMXNET3_RXM_UCAST)) {
1917             return false;
1918         }
1919         if (memcmp(s->conf.macaddr.a, ehdr->h_dest, ETH_ALEN)) {
1920             return false;
1921         }
1922         break;
1923 
1924     case ETH_PKT_BCAST:
1925         if (!VMXNET_FLAG_IS_SET(s->rx_mode, VMXNET3_RXM_BCAST)) {
1926             return false;
1927         }
1928         break;
1929 
1930     case ETH_PKT_MCAST:
1931         if (VMXNET_FLAG_IS_SET(s->rx_mode, VMXNET3_RXM_ALL_MULTI)) {
1932             return true;
1933         }
1934         if (!VMXNET_FLAG_IS_SET(s->rx_mode, VMXNET3_RXM_MCAST)) {
1935             return false;
1936         }
1937         if (!vmxnet3_is_allowed_mcast_group(s, ehdr->h_dest)) {
1938             return false;
1939         }
1940         break;
1941 
1942     default:
1943         g_assert_not_reached();
1944     }
1945 
1946     return true;
1947 }
1948 
1949 static ssize_t
1950 vmxnet3_receive(NetClientState *nc, const uint8_t *buf, size_t size)
1951 {
1952     VMXNET3State *s = qemu_get_nic_opaque(nc);
1953     size_t bytes_indicated;
1954     uint8_t min_buf[MIN_BUF_SIZE];
1955 
1956     if (!vmxnet3_can_receive(nc)) {
1957         VMW_PKPRN("Cannot receive now");
1958         return -1;
1959     }
1960 
1961     if (s->peer_has_vhdr) {
1962         net_rx_pkt_set_vhdr(s->rx_pkt, (struct virtio_net_hdr *)buf);
1963         buf += sizeof(struct virtio_net_hdr);
1964         size -= sizeof(struct virtio_net_hdr);
1965     }
1966 
1967     /* Pad to minimum Ethernet frame length */
1968     if (size < sizeof(min_buf)) {
1969         memcpy(min_buf, buf, size);
1970         memset(&min_buf[size], 0, sizeof(min_buf) - size);
1971         buf = min_buf;
1972         size = sizeof(min_buf);
1973     }
1974 
1975     net_rx_pkt_set_packet_type(s->rx_pkt,
1976         get_eth_packet_type(PKT_GET_ETH_HDR(buf)));
1977 
1978     if (vmxnet3_rx_filter_may_indicate(s, buf, size)) {
1979         net_rx_pkt_set_protocols(s->rx_pkt, buf, size);
1980         vmxnet3_rx_need_csum_calculate(s->rx_pkt, buf, size);
1981         net_rx_pkt_attach_data(s->rx_pkt, buf, size, s->rx_vlan_stripping);
1982         bytes_indicated = vmxnet3_indicate_packet(s) ? size : -1;
1983         if (bytes_indicated < size) {
1984             VMW_PKPRN("RX: %zu of %zu bytes indicated", bytes_indicated, size);
1985         }
1986     } else {
1987         VMW_PKPRN("Packet dropped by RX filter");
1988         bytes_indicated = size;
1989     }
1990 
1991     assert(size > 0);
1992     assert(bytes_indicated != 0);
1993     return bytes_indicated;
1994 }
1995 
1996 static void vmxnet3_set_link_status(NetClientState *nc)
1997 {
1998     VMXNET3State *s = qemu_get_nic_opaque(nc);
1999 
2000     if (nc->link_down) {
2001         s->link_status_and_speed &= ~VMXNET3_LINK_STATUS_UP;
2002     } else {
2003         s->link_status_and_speed |= VMXNET3_LINK_STATUS_UP;
2004     }
2005 
2006     vmxnet3_set_events(s, VMXNET3_ECR_LINK);
2007     vmxnet3_trigger_interrupt(s, s->event_int_idx);
2008 }
2009 
2010 static NetClientInfo net_vmxnet3_info = {
2011         .type = NET_CLIENT_DRIVER_NIC,
2012         .size = sizeof(NICState),
2013         .receive = vmxnet3_receive,
2014         .link_status_changed = vmxnet3_set_link_status,
2015 };
2016 
2017 static bool vmxnet3_peer_has_vnet_hdr(VMXNET3State *s)
2018 {
2019     NetClientState *nc = qemu_get_queue(s->nic);
2020 
2021     if (qemu_has_vnet_hdr(nc->peer)) {
2022         return true;
2023     }
2024 
2025     return false;
2026 }
2027 
2028 static void vmxnet3_net_uninit(VMXNET3State *s)
2029 {
2030     g_free(s->mcast_list);
2031     vmxnet3_deactivate_device(s);
2032     qemu_del_nic(s->nic);
2033 }
2034 
2035 static void vmxnet3_net_init(VMXNET3State *s)
2036 {
2037     DeviceState *d = DEVICE(s);
2038 
2039     VMW_CBPRN("vmxnet3_net_init called...");
2040 
2041     qemu_macaddr_default_if_unset(&s->conf.macaddr);
2042 
2043     /* Windows guest will query the address that was set on init */
2044     memcpy(&s->perm_mac.a, &s->conf.macaddr.a, sizeof(s->perm_mac.a));
2045 
2046     s->mcast_list = NULL;
2047     s->mcast_list_len = 0;
2048 
2049     s->link_status_and_speed = VMXNET3_LINK_SPEED | VMXNET3_LINK_STATUS_UP;
2050 
2051     VMW_CFPRN("Permanent MAC: " MAC_FMT, MAC_ARG(s->perm_mac.a));
2052 
2053     s->nic = qemu_new_nic(&net_vmxnet3_info, &s->conf,
2054                           object_get_typename(OBJECT(s)),
2055                           d->id, s);
2056 
2057     s->peer_has_vhdr = vmxnet3_peer_has_vnet_hdr(s);
2058     s->tx_sop = true;
2059     s->skip_current_tx_pkt = false;
2060     s->tx_pkt = NULL;
2061     s->rx_pkt = NULL;
2062     s->rx_vlan_stripping = false;
2063     s->lro_supported = false;
2064 
2065     if (s->peer_has_vhdr) {
2066         qemu_set_vnet_hdr_len(qemu_get_queue(s->nic)->peer,
2067             sizeof(struct virtio_net_hdr));
2068 
2069         qemu_using_vnet_hdr(qemu_get_queue(s->nic)->peer, 1);
2070     }
2071 
2072     qemu_format_nic_info_str(qemu_get_queue(s->nic), s->conf.macaddr.a);
2073 }
2074 
2075 static void
2076 vmxnet3_unuse_msix_vectors(VMXNET3State *s, int num_vectors)
2077 {
2078     PCIDevice *d = PCI_DEVICE(s);
2079     int i;
2080     for (i = 0; i < num_vectors; i++) {
2081         msix_vector_unuse(d, i);
2082     }
2083 }
2084 
2085 static bool
2086 vmxnet3_use_msix_vectors(VMXNET3State *s, int num_vectors)
2087 {
2088     PCIDevice *d = PCI_DEVICE(s);
2089     int i;
2090     for (i = 0; i < num_vectors; i++) {
2091         int res = msix_vector_use(d, i);
2092         if (0 > res) {
2093             VMW_WRPRN("Failed to use MSI-X vector %d, error %d", i, res);
2094             vmxnet3_unuse_msix_vectors(s, i);
2095             return false;
2096         }
2097     }
2098     return true;
2099 }
2100 
2101 static bool
2102 vmxnet3_init_msix(VMXNET3State *s)
2103 {
2104     PCIDevice *d = PCI_DEVICE(s);
2105     int res = msix_init(d, VMXNET3_MAX_INTRS,
2106                         &s->msix_bar,
2107                         VMXNET3_MSIX_BAR_IDX, VMXNET3_OFF_MSIX_TABLE,
2108                         &s->msix_bar,
2109                         VMXNET3_MSIX_BAR_IDX, VMXNET3_OFF_MSIX_PBA(s),
2110                         VMXNET3_MSIX_OFFSET(s), NULL);
2111 
2112     if (0 > res) {
2113         VMW_WRPRN("Failed to initialize MSI-X, error %d", res);
2114         s->msix_used = false;
2115     } else {
2116         if (!vmxnet3_use_msix_vectors(s, VMXNET3_MAX_INTRS)) {
2117             VMW_WRPRN("Failed to use MSI-X vectors, error %d", res);
2118             msix_uninit(d, &s->msix_bar, &s->msix_bar);
2119             s->msix_used = false;
2120         } else {
2121             s->msix_used = true;
2122         }
2123     }
2124     return s->msix_used;
2125 }
2126 
2127 static void
2128 vmxnet3_cleanup_msix(VMXNET3State *s)
2129 {
2130     PCIDevice *d = PCI_DEVICE(s);
2131 
2132     if (s->msix_used) {
2133         vmxnet3_unuse_msix_vectors(s, VMXNET3_MAX_INTRS);
2134         msix_uninit(d, &s->msix_bar, &s->msix_bar);
2135     }
2136 }
2137 
2138 static void
2139 vmxnet3_cleanup_msi(VMXNET3State *s)
2140 {
2141     PCIDevice *d = PCI_DEVICE(s);
2142 
2143     msi_uninit(d);
2144 }
2145 
2146 static const MemoryRegionOps b0_ops = {
2147     .read = vmxnet3_io_bar0_read,
2148     .write = vmxnet3_io_bar0_write,
2149     .endianness = DEVICE_LITTLE_ENDIAN,
2150     .impl = {
2151             .min_access_size = 4,
2152             .max_access_size = 4,
2153     },
2154 };
2155 
2156 static const MemoryRegionOps b1_ops = {
2157     .read = vmxnet3_io_bar1_read,
2158     .write = vmxnet3_io_bar1_write,
2159     .endianness = DEVICE_LITTLE_ENDIAN,
2160     .impl = {
2161             .min_access_size = 4,
2162             .max_access_size = 4,
2163     },
2164 };
2165 
2166 static uint64_t vmxnet3_device_serial_num(VMXNET3State *s)
2167 {
2168     uint64_t dsn_payload;
2169     uint8_t *dsnp = (uint8_t *)&dsn_payload;
2170 
2171     dsnp[0] = 0xfe;
2172     dsnp[1] = s->conf.macaddr.a[3];
2173     dsnp[2] = s->conf.macaddr.a[4];
2174     dsnp[3] = s->conf.macaddr.a[5];
2175     dsnp[4] = s->conf.macaddr.a[0];
2176     dsnp[5] = s->conf.macaddr.a[1];
2177     dsnp[6] = s->conf.macaddr.a[2];
2178     dsnp[7] = 0xff;
2179     return dsn_payload;
2180 }
2181 
2182 
2183 #define VMXNET3_USE_64BIT         (true)
2184 #define VMXNET3_PER_VECTOR_MASK   (false)
2185 
2186 static void vmxnet3_pci_realize(PCIDevice *pci_dev, Error **errp)
2187 {
2188     VMXNET3State *s = VMXNET3(pci_dev);
2189     int ret;
2190 
2191     VMW_CBPRN("Starting init...");
2192 
2193     memory_region_init_io(&s->bar0, OBJECT(s), &b0_ops, s,
2194                           "vmxnet3-b0", VMXNET3_PT_REG_SIZE);
2195     pci_register_bar(pci_dev, VMXNET3_BAR0_IDX,
2196                      PCI_BASE_ADDRESS_SPACE_MEMORY, &s->bar0);
2197 
2198     memory_region_init_io(&s->bar1, OBJECT(s), &b1_ops, s,
2199                           "vmxnet3-b1", VMXNET3_VD_REG_SIZE);
2200     pci_register_bar(pci_dev, VMXNET3_BAR1_IDX,
2201                      PCI_BASE_ADDRESS_SPACE_MEMORY, &s->bar1);
2202 
2203     memory_region_init(&s->msix_bar, OBJECT(s), "vmxnet3-msix-bar",
2204                        VMXNET3_MSIX_BAR_SIZE);
2205     pci_register_bar(pci_dev, VMXNET3_MSIX_BAR_IDX,
2206                      PCI_BASE_ADDRESS_SPACE_MEMORY, &s->msix_bar);
2207 
2208     vmxnet3_reset_interrupt_states(s);
2209 
2210     /* Interrupt pin A */
2211     pci_dev->config[PCI_INTERRUPT_PIN] = 0x01;
2212 
2213     ret = msi_init(pci_dev, VMXNET3_MSI_OFFSET(s), VMXNET3_MAX_NMSIX_INTRS,
2214                    VMXNET3_USE_64BIT, VMXNET3_PER_VECTOR_MASK, NULL);
2215     /* Any error other than -ENOTSUP(board's MSI support is broken)
2216      * is a programming error. Fall back to INTx silently on -ENOTSUP */
2217     assert(!ret || ret == -ENOTSUP);
2218 
2219     if (!vmxnet3_init_msix(s)) {
2220         VMW_WRPRN("Failed to initialize MSI-X, configuration is inconsistent.");
2221     }
2222 
2223     vmxnet3_net_init(s);
2224 
2225     if (pci_is_express(pci_dev)) {
2226         if (pci_bus_is_express(pci_get_bus(pci_dev))) {
2227             pcie_endpoint_cap_init(pci_dev, VMXNET3_EXP_EP_OFFSET);
2228         }
2229 
2230         pcie_dev_ser_num_init(pci_dev, VMXNET3_DSN_OFFSET,
2231                               vmxnet3_device_serial_num(s));
2232     }
2233 }
2234 
2235 static void vmxnet3_instance_init(Object *obj)
2236 {
2237     VMXNET3State *s = VMXNET3(obj);
2238     device_add_bootindex_property(obj, &s->conf.bootindex,
2239                                   "bootindex", "/ethernet-phy@0",
2240                                   DEVICE(obj), NULL);
2241 }
2242 
2243 static void vmxnet3_pci_uninit(PCIDevice *pci_dev)
2244 {
2245     VMXNET3State *s = VMXNET3(pci_dev);
2246 
2247     VMW_CBPRN("Starting uninit...");
2248 
2249     vmxnet3_net_uninit(s);
2250 
2251     vmxnet3_cleanup_msix(s);
2252 
2253     vmxnet3_cleanup_msi(s);
2254 }
2255 
2256 static void vmxnet3_qdev_reset(DeviceState *dev)
2257 {
2258     PCIDevice *d = PCI_DEVICE(dev);
2259     VMXNET3State *s = VMXNET3(d);
2260 
2261     VMW_CBPRN("Starting QDEV reset...");
2262     vmxnet3_reset(s);
2263 }
2264 
2265 static bool vmxnet3_mc_list_needed(void *opaque)
2266 {
2267     return true;
2268 }
2269 
2270 static int vmxnet3_mcast_list_pre_load(void *opaque)
2271 {
2272     VMXNET3State *s = opaque;
2273 
2274     s->mcast_list = g_malloc(s->mcast_list_buff_size);
2275 
2276     return 0;
2277 }
2278 
2279 
2280 static int vmxnet3_pre_save(void *opaque)
2281 {
2282     VMXNET3State *s = opaque;
2283 
2284     s->mcast_list_buff_size = s->mcast_list_len * sizeof(MACAddr);
2285 
2286     return 0;
2287 }
2288 
2289 static const VMStateDescription vmxstate_vmxnet3_mcast_list = {
2290     .name = "vmxnet3/mcast_list",
2291     .version_id = 1,
2292     .minimum_version_id = 1,
2293     .pre_load = vmxnet3_mcast_list_pre_load,
2294     .needed = vmxnet3_mc_list_needed,
2295     .fields = (VMStateField[]) {
2296         VMSTATE_VBUFFER_UINT32(mcast_list, VMXNET3State, 0, NULL,
2297             mcast_list_buff_size),
2298         VMSTATE_END_OF_LIST()
2299     }
2300 };
2301 
2302 static const VMStateDescription vmstate_vmxnet3_ring = {
2303     .name = "vmxnet3-ring",
2304     .version_id = 0,
2305     .fields = (VMStateField[]) {
2306         VMSTATE_UINT64(pa, Vmxnet3Ring),
2307         VMSTATE_UINT32(size, Vmxnet3Ring),
2308         VMSTATE_UINT32(cell_size, Vmxnet3Ring),
2309         VMSTATE_UINT32(next, Vmxnet3Ring),
2310         VMSTATE_UINT8(gen, Vmxnet3Ring),
2311         VMSTATE_END_OF_LIST()
2312     }
2313 };
2314 
2315 static const VMStateDescription vmstate_vmxnet3_tx_stats = {
2316     .name = "vmxnet3-tx-stats",
2317     .version_id = 0,
2318     .fields = (VMStateField[]) {
2319         VMSTATE_UINT64(TSOPktsTxOK, struct UPT1_TxStats),
2320         VMSTATE_UINT64(TSOBytesTxOK, struct UPT1_TxStats),
2321         VMSTATE_UINT64(ucastPktsTxOK, struct UPT1_TxStats),
2322         VMSTATE_UINT64(ucastBytesTxOK, struct UPT1_TxStats),
2323         VMSTATE_UINT64(mcastPktsTxOK, struct UPT1_TxStats),
2324         VMSTATE_UINT64(mcastBytesTxOK, struct UPT1_TxStats),
2325         VMSTATE_UINT64(bcastPktsTxOK, struct UPT1_TxStats),
2326         VMSTATE_UINT64(bcastBytesTxOK, struct UPT1_TxStats),
2327         VMSTATE_UINT64(pktsTxError, struct UPT1_TxStats),
2328         VMSTATE_UINT64(pktsTxDiscard, struct UPT1_TxStats),
2329         VMSTATE_END_OF_LIST()
2330     }
2331 };
2332 
2333 static const VMStateDescription vmstate_vmxnet3_txq_descr = {
2334     .name = "vmxnet3-txq-descr",
2335     .version_id = 0,
2336     .fields = (VMStateField[]) {
2337         VMSTATE_STRUCT(tx_ring, Vmxnet3TxqDescr, 0, vmstate_vmxnet3_ring,
2338                        Vmxnet3Ring),
2339         VMSTATE_STRUCT(comp_ring, Vmxnet3TxqDescr, 0, vmstate_vmxnet3_ring,
2340                        Vmxnet3Ring),
2341         VMSTATE_UINT8(intr_idx, Vmxnet3TxqDescr),
2342         VMSTATE_UINT64(tx_stats_pa, Vmxnet3TxqDescr),
2343         VMSTATE_STRUCT(txq_stats, Vmxnet3TxqDescr, 0, vmstate_vmxnet3_tx_stats,
2344                        struct UPT1_TxStats),
2345         VMSTATE_END_OF_LIST()
2346     }
2347 };
2348 
2349 static const VMStateDescription vmstate_vmxnet3_rx_stats = {
2350     .name = "vmxnet3-rx-stats",
2351     .version_id = 0,
2352     .fields = (VMStateField[]) {
2353         VMSTATE_UINT64(LROPktsRxOK, struct UPT1_RxStats),
2354         VMSTATE_UINT64(LROBytesRxOK, struct UPT1_RxStats),
2355         VMSTATE_UINT64(ucastPktsRxOK, struct UPT1_RxStats),
2356         VMSTATE_UINT64(ucastBytesRxOK, struct UPT1_RxStats),
2357         VMSTATE_UINT64(mcastPktsRxOK, struct UPT1_RxStats),
2358         VMSTATE_UINT64(mcastBytesRxOK, struct UPT1_RxStats),
2359         VMSTATE_UINT64(bcastPktsRxOK, struct UPT1_RxStats),
2360         VMSTATE_UINT64(bcastBytesRxOK, struct UPT1_RxStats),
2361         VMSTATE_UINT64(pktsRxOutOfBuf, struct UPT1_RxStats),
2362         VMSTATE_UINT64(pktsRxError, struct UPT1_RxStats),
2363         VMSTATE_END_OF_LIST()
2364     }
2365 };
2366 
2367 static const VMStateDescription vmstate_vmxnet3_rxq_descr = {
2368     .name = "vmxnet3-rxq-descr",
2369     .version_id = 0,
2370     .fields = (VMStateField[]) {
2371         VMSTATE_STRUCT_ARRAY(rx_ring, Vmxnet3RxqDescr,
2372                              VMXNET3_RX_RINGS_PER_QUEUE, 0,
2373                              vmstate_vmxnet3_ring, Vmxnet3Ring),
2374         VMSTATE_STRUCT(comp_ring, Vmxnet3RxqDescr, 0, vmstate_vmxnet3_ring,
2375                        Vmxnet3Ring),
2376         VMSTATE_UINT8(intr_idx, Vmxnet3RxqDescr),
2377         VMSTATE_UINT64(rx_stats_pa, Vmxnet3RxqDescr),
2378         VMSTATE_STRUCT(rxq_stats, Vmxnet3RxqDescr, 0, vmstate_vmxnet3_rx_stats,
2379                        struct UPT1_RxStats),
2380         VMSTATE_END_OF_LIST()
2381     }
2382 };
2383 
2384 static int vmxnet3_post_load(void *opaque, int version_id)
2385 {
2386     VMXNET3State *s = opaque;
2387     PCIDevice *d = PCI_DEVICE(s);
2388 
2389     net_tx_pkt_init(&s->tx_pkt, PCI_DEVICE(s),
2390                     s->max_tx_frags, s->peer_has_vhdr);
2391     net_rx_pkt_init(&s->rx_pkt, s->peer_has_vhdr);
2392 
2393     if (s->msix_used) {
2394         if  (!vmxnet3_use_msix_vectors(s, VMXNET3_MAX_INTRS)) {
2395             VMW_WRPRN("Failed to re-use MSI-X vectors");
2396             msix_uninit(d, &s->msix_bar, &s->msix_bar);
2397             s->msix_used = false;
2398             return -1;
2399         }
2400     }
2401 
2402     vmxnet3_validate_queues(s);
2403     vmxnet3_validate_interrupts(s);
2404 
2405     return 0;
2406 }
2407 
2408 static const VMStateDescription vmstate_vmxnet3_int_state = {
2409     .name = "vmxnet3-int-state",
2410     .version_id = 0,
2411     .fields = (VMStateField[]) {
2412         VMSTATE_BOOL(is_masked, Vmxnet3IntState),
2413         VMSTATE_BOOL(is_pending, Vmxnet3IntState),
2414         VMSTATE_BOOL(is_asserted, Vmxnet3IntState),
2415         VMSTATE_END_OF_LIST()
2416     }
2417 };
2418 
2419 static const VMStateDescription vmstate_vmxnet3 = {
2420     .name = "vmxnet3",
2421     .version_id = 1,
2422     .minimum_version_id = 1,
2423     .pre_save = vmxnet3_pre_save,
2424     .post_load = vmxnet3_post_load,
2425     .fields = (VMStateField[]) {
2426             VMSTATE_PCI_DEVICE(parent_obj, VMXNET3State),
2427             VMSTATE_MSIX(parent_obj, VMXNET3State),
2428             VMSTATE_BOOL(rx_packets_compound, VMXNET3State),
2429             VMSTATE_BOOL(rx_vlan_stripping, VMXNET3State),
2430             VMSTATE_BOOL(lro_supported, VMXNET3State),
2431             VMSTATE_UINT32(rx_mode, VMXNET3State),
2432             VMSTATE_UINT32(mcast_list_len, VMXNET3State),
2433             VMSTATE_UINT32(mcast_list_buff_size, VMXNET3State),
2434             VMSTATE_UINT32_ARRAY(vlan_table, VMXNET3State, VMXNET3_VFT_SIZE),
2435             VMSTATE_UINT32(mtu, VMXNET3State),
2436             VMSTATE_UINT16(max_rx_frags, VMXNET3State),
2437             VMSTATE_UINT32(max_tx_frags, VMXNET3State),
2438             VMSTATE_UINT8(event_int_idx, VMXNET3State),
2439             VMSTATE_BOOL(auto_int_masking, VMXNET3State),
2440             VMSTATE_UINT8(txq_num, VMXNET3State),
2441             VMSTATE_UINT8(rxq_num, VMXNET3State),
2442             VMSTATE_UINT32(device_active, VMXNET3State),
2443             VMSTATE_UINT32(last_command, VMXNET3State),
2444             VMSTATE_UINT32(link_status_and_speed, VMXNET3State),
2445             VMSTATE_UINT32(temp_mac, VMXNET3State),
2446             VMSTATE_UINT64(drv_shmem, VMXNET3State),
2447             VMSTATE_UINT64(temp_shared_guest_driver_memory, VMXNET3State),
2448 
2449             VMSTATE_STRUCT_ARRAY(txq_descr, VMXNET3State,
2450                 VMXNET3_DEVICE_MAX_TX_QUEUES, 0, vmstate_vmxnet3_txq_descr,
2451                 Vmxnet3TxqDescr),
2452             VMSTATE_STRUCT_ARRAY(rxq_descr, VMXNET3State,
2453                 VMXNET3_DEVICE_MAX_RX_QUEUES, 0, vmstate_vmxnet3_rxq_descr,
2454                 Vmxnet3RxqDescr),
2455             VMSTATE_STRUCT_ARRAY(interrupt_states, VMXNET3State,
2456                 VMXNET3_MAX_INTRS, 0, vmstate_vmxnet3_int_state,
2457                 Vmxnet3IntState),
2458 
2459             VMSTATE_END_OF_LIST()
2460     },
2461     .subsections = (const VMStateDescription*[]) {
2462         &vmxstate_vmxnet3_mcast_list,
2463         NULL
2464     }
2465 };
2466 
2467 static Property vmxnet3_properties[] = {
2468     DEFINE_NIC_PROPERTIES(VMXNET3State, conf),
2469     DEFINE_PROP_BIT("x-old-msi-offsets", VMXNET3State, compat_flags,
2470                     VMXNET3_COMPAT_FLAG_OLD_MSI_OFFSETS_BIT, false),
2471     DEFINE_PROP_BIT("x-disable-pcie", VMXNET3State, compat_flags,
2472                     VMXNET3_COMPAT_FLAG_DISABLE_PCIE_BIT, false),
2473     DEFINE_PROP_END_OF_LIST(),
2474 };
2475 
2476 static void vmxnet3_realize(DeviceState *qdev, Error **errp)
2477 {
2478     VMXNET3Class *vc = VMXNET3_DEVICE_GET_CLASS(qdev);
2479     PCIDevice *pci_dev = PCI_DEVICE(qdev);
2480     VMXNET3State *s = VMXNET3(qdev);
2481 
2482     if (!(s->compat_flags & VMXNET3_COMPAT_FLAG_DISABLE_PCIE)) {
2483         pci_dev->cap_present |= QEMU_PCI_CAP_EXPRESS;
2484     }
2485 
2486     vc->parent_dc_realize(qdev, errp);
2487 }
2488 
2489 static void vmxnet3_class_init(ObjectClass *class, void *data)
2490 {
2491     DeviceClass *dc = DEVICE_CLASS(class);
2492     PCIDeviceClass *c = PCI_DEVICE_CLASS(class);
2493     VMXNET3Class *vc = VMXNET3_DEVICE_CLASS(class);
2494 
2495     c->realize = vmxnet3_pci_realize;
2496     c->exit = vmxnet3_pci_uninit;
2497     c->vendor_id = PCI_VENDOR_ID_VMWARE;
2498     c->device_id = PCI_DEVICE_ID_VMWARE_VMXNET3;
2499     c->revision = PCI_DEVICE_ID_VMWARE_VMXNET3_REVISION;
2500     c->romfile = "efi-vmxnet3.rom";
2501     c->class_id = PCI_CLASS_NETWORK_ETHERNET;
2502     c->subsystem_vendor_id = PCI_VENDOR_ID_VMWARE;
2503     c->subsystem_id = PCI_DEVICE_ID_VMWARE_VMXNET3;
2504     device_class_set_parent_realize(dc, vmxnet3_realize,
2505                                     &vc->parent_dc_realize);
2506     dc->desc = "VMWare Paravirtualized Ethernet v3";
2507     dc->reset = vmxnet3_qdev_reset;
2508     dc->vmsd = &vmstate_vmxnet3;
2509     device_class_set_props(dc, vmxnet3_properties);
2510     set_bit(DEVICE_CATEGORY_NETWORK, dc->categories);
2511 }
2512 
2513 static const TypeInfo vmxnet3_info = {
2514     .name          = TYPE_VMXNET3,
2515     .parent        = TYPE_PCI_DEVICE,
2516     .class_size    = sizeof(VMXNET3Class),
2517     .instance_size = sizeof(VMXNET3State),
2518     .class_init    = vmxnet3_class_init,
2519     .instance_init = vmxnet3_instance_init,
2520     .interfaces = (InterfaceInfo[]) {
2521         { INTERFACE_PCIE_DEVICE },
2522         { INTERFACE_CONVENTIONAL_PCI_DEVICE },
2523         { }
2524     },
2525 };
2526 
2527 static void vmxnet3_register_types(void)
2528 {
2529     VMW_CBPRN("vmxnet3_register_types called...");
2530     type_register_static(&vmxnet3_info);
2531 }
2532 
2533 type_init(vmxnet3_register_types)
2534