xref: /openbmc/qemu/hw/net/vmxnet3.c (revision a495eba0)
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 #define MIN_BUF_SIZE 60
44 
45 /* Compatibility flags for migration */
46 #define VMXNET3_COMPAT_FLAG_OLD_MSI_OFFSETS_BIT 0
47 #define VMXNET3_COMPAT_FLAG_OLD_MSI_OFFSETS \
48     (1 << VMXNET3_COMPAT_FLAG_OLD_MSI_OFFSETS_BIT)
49 #define VMXNET3_COMPAT_FLAG_DISABLE_PCIE_BIT 1
50 #define VMXNET3_COMPAT_FLAG_DISABLE_PCIE \
51     (1 << VMXNET3_COMPAT_FLAG_DISABLE_PCIE_BIT)
52 
53 #define VMXNET3_EXP_EP_OFFSET (0x48)
54 #define VMXNET3_MSI_OFFSET(s) \
55     ((s)->compat_flags & VMXNET3_COMPAT_FLAG_OLD_MSI_OFFSETS ? 0x50 : 0x84)
56 #define VMXNET3_MSIX_OFFSET(s) \
57     ((s)->compat_flags & VMXNET3_COMPAT_FLAG_OLD_MSI_OFFSETS ? 0 : 0x9c)
58 #define VMXNET3_DSN_OFFSET     (0x100)
59 
60 #define VMXNET3_BAR0_IDX      (0)
61 #define VMXNET3_BAR1_IDX      (1)
62 #define VMXNET3_MSIX_BAR_IDX  (2)
63 
64 #define VMXNET3_OFF_MSIX_TABLE (0x000)
65 #define VMXNET3_OFF_MSIX_PBA(s) \
66     ((s)->compat_flags & VMXNET3_COMPAT_FLAG_OLD_MSI_OFFSETS ? 0x800 : 0x1000)
67 
68 /* Link speed in Mbps should be shifted by 16 */
69 #define VMXNET3_LINK_SPEED      (1000 << 16)
70 
71 /* Link status: 1 - up, 0 - down. */
72 #define VMXNET3_LINK_STATUS_UP  0x1
73 
74 /* Least significant bit should be set for revision and version */
75 #define VMXNET3_UPT_REVISION      0x1
76 #define VMXNET3_DEVICE_REVISION   0x1
77 
78 /* Number of interrupt vectors for non-MSIx modes */
79 #define VMXNET3_MAX_NMSIX_INTRS   (1)
80 
81 /* Macros for rings descriptors access */
82 #define VMXNET3_READ_TX_QUEUE_DESCR8(_d, dpa, field) \
83     (vmw_shmem_ld8(_d, dpa + offsetof(struct Vmxnet3_TxQueueDesc, field)))
84 
85 #define VMXNET3_WRITE_TX_QUEUE_DESCR8(_d, dpa, field, value) \
86     (vmw_shmem_st8(_d, dpa + offsetof(struct Vmxnet3_TxQueueDesc, field, value)))
87 
88 #define VMXNET3_READ_TX_QUEUE_DESCR32(_d, dpa, field) \
89     (vmw_shmem_ld32(_d, dpa + offsetof(struct Vmxnet3_TxQueueDesc, field)))
90 
91 #define VMXNET3_WRITE_TX_QUEUE_DESCR32(_d, dpa, field, value) \
92     (vmw_shmem_st32(_d, dpa + offsetof(struct Vmxnet3_TxQueueDesc, field), value))
93 
94 #define VMXNET3_READ_TX_QUEUE_DESCR64(_d, dpa, field) \
95     (vmw_shmem_ld64(_d, dpa + offsetof(struct Vmxnet3_TxQueueDesc, field)))
96 
97 #define VMXNET3_WRITE_TX_QUEUE_DESCR64(_d, dpa, field, value) \
98     (vmw_shmem_st64(_d, dpa + offsetof(struct Vmxnet3_TxQueueDesc, field), value))
99 
100 #define VMXNET3_READ_RX_QUEUE_DESCR64(_d, dpa, field) \
101     (vmw_shmem_ld64(_d, dpa + offsetof(struct Vmxnet3_RxQueueDesc, field)))
102 
103 #define VMXNET3_READ_RX_QUEUE_DESCR32(_d, dpa, field) \
104     (vmw_shmem_ld32(_d, dpa + offsetof(struct Vmxnet3_RxQueueDesc, field)))
105 
106 #define VMXNET3_WRITE_RX_QUEUE_DESCR64(_d, dpa, field, value) \
107     (vmw_shmem_st64(_d, dpa + offsetof(struct Vmxnet3_RxQueueDesc, field), value))
108 
109 #define VMXNET3_WRITE_RX_QUEUE_DESCR8(_d, dpa, field, value) \
110     (vmw_shmem_st8(_d, dpa + offsetof(struct Vmxnet3_RxQueueDesc, field), value))
111 
112 /* Macros for guest driver shared area access */
113 #define VMXNET3_READ_DRV_SHARED64(_d, shpa, field) \
114     (vmw_shmem_ld64(_d, shpa + offsetof(struct Vmxnet3_DriverShared, field)))
115 
116 #define VMXNET3_READ_DRV_SHARED32(_d, shpa, field) \
117     (vmw_shmem_ld32(_d, shpa + offsetof(struct Vmxnet3_DriverShared, field)))
118 
119 #define VMXNET3_WRITE_DRV_SHARED32(_d, shpa, field, val) \
120     (vmw_shmem_st32(_d, shpa + offsetof(struct Vmxnet3_DriverShared, field), val))
121 
122 #define VMXNET3_READ_DRV_SHARED16(_d, shpa, field) \
123     (vmw_shmem_ld16(_d, shpa + offsetof(struct Vmxnet3_DriverShared, field)))
124 
125 #define VMXNET3_READ_DRV_SHARED8(_d, shpa, field) \
126     (vmw_shmem_ld8(_d, shpa + offsetof(struct Vmxnet3_DriverShared, field)))
127 
128 #define VMXNET3_READ_DRV_SHARED(_d, shpa, field, b, l) \
129     (vmw_shmem_read(_d, shpa + offsetof(struct Vmxnet3_DriverShared, field), b, l))
130 
131 #define VMXNET_FLAG_IS_SET(field, flag) (((field) & (flag)) == (flag))
132 
133 struct VMXNET3Class {
134     PCIDeviceClass parent_class;
135     DeviceRealize parent_dc_realize;
136 };
137 typedef struct VMXNET3Class VMXNET3Class;
138 
139 DECLARE_CLASS_CHECKERS(VMXNET3Class, VMXNET3_DEVICE,
140                        TYPE_VMXNET3)
141 
142 static inline void vmxnet3_ring_init(PCIDevice *d,
143                                      Vmxnet3Ring *ring,
144                                      hwaddr pa,
145                                      uint32_t size,
146                                      uint32_t cell_size,
147                                      bool zero_region)
148 {
149     ring->pa = pa;
150     ring->size = size;
151     ring->cell_size = cell_size;
152     ring->gen = VMXNET3_INIT_GEN;
153     ring->next = 0;
154 
155     if (zero_region) {
156         vmw_shmem_set(d, pa, 0, size * cell_size);
157     }
158 }
159 
160 #define VMXNET3_RING_DUMP(macro, ring_name, ridx, r)                         \
161     macro("%s#%d: base %" PRIx64 " size %u cell_size %u gen %d next %u",  \
162           (ring_name), (ridx),                                               \
163           (r)->pa, (r)->size, (r)->cell_size, (r)->gen, (r)->next)
164 
165 static inline void vmxnet3_ring_inc(Vmxnet3Ring *ring)
166 {
167     if (++ring->next >= ring->size) {
168         ring->next = 0;
169         ring->gen ^= 1;
170     }
171 }
172 
173 static inline void vmxnet3_ring_dec(Vmxnet3Ring *ring)
174 {
175     if (ring->next-- == 0) {
176         ring->next = ring->size - 1;
177         ring->gen ^= 1;
178     }
179 }
180 
181 static inline hwaddr vmxnet3_ring_curr_cell_pa(Vmxnet3Ring *ring)
182 {
183     return ring->pa + ring->next * ring->cell_size;
184 }
185 
186 static inline void vmxnet3_ring_read_curr_cell(PCIDevice *d, Vmxnet3Ring *ring,
187                                                void *buff)
188 {
189     vmw_shmem_read(d, vmxnet3_ring_curr_cell_pa(ring), buff, ring->cell_size);
190 }
191 
192 static inline void vmxnet3_ring_write_curr_cell(PCIDevice *d, Vmxnet3Ring *ring,
193                                                 void *buff)
194 {
195     vmw_shmem_write(d, vmxnet3_ring_curr_cell_pa(ring), buff, ring->cell_size);
196 }
197 
198 static inline size_t vmxnet3_ring_curr_cell_idx(Vmxnet3Ring *ring)
199 {
200     return ring->next;
201 }
202 
203 static inline uint8_t vmxnet3_ring_curr_gen(Vmxnet3Ring *ring)
204 {
205     return ring->gen;
206 }
207 
208 /* Debug trace-related functions */
209 static inline void
210 vmxnet3_dump_tx_descr(struct Vmxnet3_TxDesc *descr)
211 {
212     VMW_PKPRN("TX DESCR: "
213               "addr %" PRIx64 ", len: %d, gen: %d, rsvd: %d, "
214               "dtype: %d, ext1: %d, msscof: %d, hlen: %d, om: %d, "
215               "eop: %d, cq: %d, ext2: %d, ti: %d, tci: %d",
216               descr->addr, descr->len, descr->gen, descr->rsvd,
217               descr->dtype, descr->ext1, descr->msscof, descr->hlen, descr->om,
218               descr->eop, descr->cq, descr->ext2, descr->ti, descr->tci);
219 }
220 
221 static inline void
222 vmxnet3_dump_virt_hdr(struct virtio_net_hdr *vhdr)
223 {
224     VMW_PKPRN("VHDR: flags 0x%x, gso_type: 0x%x, hdr_len: %d, gso_size: %d, "
225               "csum_start: %d, csum_offset: %d",
226               vhdr->flags, vhdr->gso_type, vhdr->hdr_len, vhdr->gso_size,
227               vhdr->csum_start, vhdr->csum_offset);
228 }
229 
230 static inline void
231 vmxnet3_dump_rx_descr(struct Vmxnet3_RxDesc *descr)
232 {
233     VMW_PKPRN("RX DESCR: addr %" PRIx64 ", len: %d, gen: %d, rsvd: %d, "
234               "dtype: %d, ext1: %d, btype: %d",
235               descr->addr, descr->len, descr->gen,
236               descr->rsvd, descr->dtype, descr->ext1, descr->btype);
237 }
238 
239 /* Interrupt management */
240 
241 /*
242  * This function returns sign whether interrupt line is in asserted state
243  * This depends on the type of interrupt used. For INTX interrupt line will
244  * be asserted until explicit deassertion, for MSI(X) interrupt line will
245  * be deasserted automatically due to notification semantics of the MSI(X)
246  * interrupts
247  */
248 static bool _vmxnet3_assert_interrupt_line(VMXNET3State *s, uint32_t int_idx)
249 {
250     PCIDevice *d = PCI_DEVICE(s);
251 
252     if (s->msix_used && msix_enabled(d)) {
253         VMW_IRPRN("Sending MSI-X notification for vector %u", int_idx);
254         msix_notify(d, int_idx);
255         return false;
256     }
257     if (msi_enabled(d)) {
258         VMW_IRPRN("Sending MSI notification for vector %u", int_idx);
259         msi_notify(d, int_idx);
260         return false;
261     }
262 
263     VMW_IRPRN("Asserting line for interrupt %u", int_idx);
264     pci_irq_assert(d);
265     return true;
266 }
267 
268 static void _vmxnet3_deassert_interrupt_line(VMXNET3State *s, int lidx)
269 {
270     PCIDevice *d = PCI_DEVICE(s);
271 
272     /*
273      * This function should never be called for MSI(X) interrupts
274      * because deassertion never required for message interrupts
275      */
276     assert(!s->msix_used || !msix_enabled(d));
277     /*
278      * This function should never be called for MSI(X) interrupts
279      * because deassertion never required for message interrupts
280      */
281     assert(!msi_enabled(d));
282 
283     VMW_IRPRN("Deasserting line for interrupt %u", lidx);
284     pci_irq_deassert(d);
285 }
286 
287 static void vmxnet3_update_interrupt_line_state(VMXNET3State *s, int lidx)
288 {
289     if (!s->interrupt_states[lidx].is_pending &&
290        s->interrupt_states[lidx].is_asserted) {
291         VMW_IRPRN("New interrupt line state for index %d is DOWN", lidx);
292         _vmxnet3_deassert_interrupt_line(s, lidx);
293         s->interrupt_states[lidx].is_asserted = false;
294         return;
295     }
296 
297     if (s->interrupt_states[lidx].is_pending &&
298        !s->interrupt_states[lidx].is_masked &&
299        !s->interrupt_states[lidx].is_asserted) {
300         VMW_IRPRN("New interrupt line state for index %d is UP", lidx);
301         s->interrupt_states[lidx].is_asserted =
302             _vmxnet3_assert_interrupt_line(s, lidx);
303         s->interrupt_states[lidx].is_pending = false;
304         return;
305     }
306 }
307 
308 static void vmxnet3_trigger_interrupt(VMXNET3State *s, int lidx)
309 {
310     PCIDevice *d = PCI_DEVICE(s);
311     s->interrupt_states[lidx].is_pending = true;
312     vmxnet3_update_interrupt_line_state(s, lidx);
313 
314     if (s->msix_used && msix_enabled(d) && s->auto_int_masking) {
315         goto do_automask;
316     }
317 
318     if (msi_enabled(d) && s->auto_int_masking) {
319         goto do_automask;
320     }
321 
322     return;
323 
324 do_automask:
325     s->interrupt_states[lidx].is_masked = true;
326     vmxnet3_update_interrupt_line_state(s, lidx);
327 }
328 
329 static bool vmxnet3_interrupt_asserted(VMXNET3State *s, int lidx)
330 {
331     return s->interrupt_states[lidx].is_asserted;
332 }
333 
334 static void vmxnet3_clear_interrupt(VMXNET3State *s, int int_idx)
335 {
336     s->interrupt_states[int_idx].is_pending = false;
337     if (s->auto_int_masking) {
338         s->interrupt_states[int_idx].is_masked = true;
339     }
340     vmxnet3_update_interrupt_line_state(s, int_idx);
341 }
342 
343 static void
344 vmxnet3_on_interrupt_mask_changed(VMXNET3State *s, int lidx, bool is_masked)
345 {
346     s->interrupt_states[lidx].is_masked = is_masked;
347     vmxnet3_update_interrupt_line_state(s, lidx);
348 }
349 
350 static bool vmxnet3_verify_driver_magic(PCIDevice *d, hwaddr dshmem)
351 {
352     return (VMXNET3_READ_DRV_SHARED32(d, dshmem, magic) == VMXNET3_REV1_MAGIC);
353 }
354 
355 #define VMXNET3_GET_BYTE(x, byte_num) (((x) >> (byte_num)*8) & 0xFF)
356 #define VMXNET3_MAKE_BYTE(byte_num, val) \
357     (((uint32_t)((val) & 0xFF)) << (byte_num)*8)
358 
359 static void vmxnet3_set_variable_mac(VMXNET3State *s, uint32_t h, uint32_t l)
360 {
361     s->conf.macaddr.a[0] = VMXNET3_GET_BYTE(l,  0);
362     s->conf.macaddr.a[1] = VMXNET3_GET_BYTE(l,  1);
363     s->conf.macaddr.a[2] = VMXNET3_GET_BYTE(l,  2);
364     s->conf.macaddr.a[3] = VMXNET3_GET_BYTE(l,  3);
365     s->conf.macaddr.a[4] = VMXNET3_GET_BYTE(h, 0);
366     s->conf.macaddr.a[5] = VMXNET3_GET_BYTE(h, 1);
367 
368     VMW_CFPRN("Variable MAC: " MAC_FMT, MAC_ARG(s->conf.macaddr.a));
369 
370     qemu_format_nic_info_str(qemu_get_queue(s->nic), s->conf.macaddr.a);
371 }
372 
373 static uint64_t vmxnet3_get_mac_low(MACAddr *addr)
374 {
375     return VMXNET3_MAKE_BYTE(0, addr->a[0]) |
376            VMXNET3_MAKE_BYTE(1, addr->a[1]) |
377            VMXNET3_MAKE_BYTE(2, addr->a[2]) |
378            VMXNET3_MAKE_BYTE(3, addr->a[3]);
379 }
380 
381 static uint64_t vmxnet3_get_mac_high(MACAddr *addr)
382 {
383     return VMXNET3_MAKE_BYTE(0, addr->a[4]) |
384            VMXNET3_MAKE_BYTE(1, addr->a[5]);
385 }
386 
387 static void
388 vmxnet3_inc_tx_consumption_counter(VMXNET3State *s, int qidx)
389 {
390     vmxnet3_ring_inc(&s->txq_descr[qidx].tx_ring);
391 }
392 
393 static inline void
394 vmxnet3_inc_rx_consumption_counter(VMXNET3State *s, int qidx, int ridx)
395 {
396     vmxnet3_ring_inc(&s->rxq_descr[qidx].rx_ring[ridx]);
397 }
398 
399 static inline void
400 vmxnet3_inc_tx_completion_counter(VMXNET3State *s, int qidx)
401 {
402     vmxnet3_ring_inc(&s->txq_descr[qidx].comp_ring);
403 }
404 
405 static void
406 vmxnet3_inc_rx_completion_counter(VMXNET3State *s, int qidx)
407 {
408     vmxnet3_ring_inc(&s->rxq_descr[qidx].comp_ring);
409 }
410 
411 static void
412 vmxnet3_dec_rx_completion_counter(VMXNET3State *s, int qidx)
413 {
414     vmxnet3_ring_dec(&s->rxq_descr[qidx].comp_ring);
415 }
416 
417 static void vmxnet3_complete_packet(VMXNET3State *s, int qidx, uint32_t tx_ridx)
418 {
419     struct Vmxnet3_TxCompDesc txcq_descr;
420     PCIDevice *d = PCI_DEVICE(s);
421 
422     VMXNET3_RING_DUMP(VMW_RIPRN, "TXC", qidx, &s->txq_descr[qidx].comp_ring);
423 
424     memset(&txcq_descr, 0, sizeof(txcq_descr));
425     txcq_descr.txdIdx = tx_ridx;
426     txcq_descr.gen = vmxnet3_ring_curr_gen(&s->txq_descr[qidx].comp_ring);
427     txcq_descr.val1 = cpu_to_le32(txcq_descr.val1);
428     txcq_descr.val2 = cpu_to_le32(txcq_descr.val2);
429     vmxnet3_ring_write_curr_cell(d, &s->txq_descr[qidx].comp_ring, &txcq_descr);
430 
431     /* Flush changes in TX descriptor before changing the counter value */
432     smp_wmb();
433 
434     vmxnet3_inc_tx_completion_counter(s, qidx);
435     vmxnet3_trigger_interrupt(s, s->txq_descr[qidx].intr_idx);
436 }
437 
438 static bool
439 vmxnet3_setup_tx_offloads(VMXNET3State *s)
440 {
441     switch (s->offload_mode) {
442     case VMXNET3_OM_NONE:
443         net_tx_pkt_build_vheader(s->tx_pkt, false, false, 0);
444         break;
445 
446     case VMXNET3_OM_CSUM:
447         net_tx_pkt_build_vheader(s->tx_pkt, false, true, 0);
448         VMW_PKPRN("L4 CSO requested\n");
449         break;
450 
451     case VMXNET3_OM_TSO:
452         net_tx_pkt_build_vheader(s->tx_pkt, true, true,
453             s->cso_or_gso_size);
454         net_tx_pkt_update_ip_checksums(s->tx_pkt);
455         VMW_PKPRN("GSO offload requested.");
456         break;
457 
458     default:
459         g_assert_not_reached();
460         return false;
461     }
462 
463     return true;
464 }
465 
466 static void
467 vmxnet3_tx_retrieve_metadata(VMXNET3State *s,
468                              const struct Vmxnet3_TxDesc *txd)
469 {
470     s->offload_mode = txd->om;
471     s->cso_or_gso_size = txd->msscof;
472     s->tci = txd->tci;
473     s->needs_vlan = txd->ti;
474 }
475 
476 typedef enum {
477     VMXNET3_PKT_STATUS_OK,
478     VMXNET3_PKT_STATUS_ERROR,
479     VMXNET3_PKT_STATUS_DISCARD,/* only for tx */
480     VMXNET3_PKT_STATUS_OUT_OF_BUF /* only for rx */
481 } Vmxnet3PktStatus;
482 
483 static void
484 vmxnet3_on_tx_done_update_stats(VMXNET3State *s, int qidx,
485     Vmxnet3PktStatus status)
486 {
487     size_t tot_len = net_tx_pkt_get_total_len(s->tx_pkt);
488     struct UPT1_TxStats *stats = &s->txq_descr[qidx].txq_stats;
489 
490     switch (status) {
491     case VMXNET3_PKT_STATUS_OK:
492         switch (net_tx_pkt_get_packet_type(s->tx_pkt)) {
493         case ETH_PKT_BCAST:
494             stats->bcastPktsTxOK++;
495             stats->bcastBytesTxOK += tot_len;
496             break;
497         case ETH_PKT_MCAST:
498             stats->mcastPktsTxOK++;
499             stats->mcastBytesTxOK += tot_len;
500             break;
501         case ETH_PKT_UCAST:
502             stats->ucastPktsTxOK++;
503             stats->ucastBytesTxOK += tot_len;
504             break;
505         default:
506             g_assert_not_reached();
507         }
508 
509         if (s->offload_mode == VMXNET3_OM_TSO) {
510             /*
511              * According to VMWARE headers this statistic is a number
512              * of packets after segmentation but since we don't have
513              * this information in QEMU model, the best we can do is to
514              * provide number of non-segmented packets
515              */
516             stats->TSOPktsTxOK++;
517             stats->TSOBytesTxOK += tot_len;
518         }
519         break;
520 
521     case VMXNET3_PKT_STATUS_DISCARD:
522         stats->pktsTxDiscard++;
523         break;
524 
525     case VMXNET3_PKT_STATUS_ERROR:
526         stats->pktsTxError++;
527         break;
528 
529     default:
530         g_assert_not_reached();
531     }
532 }
533 
534 static void
535 vmxnet3_on_rx_done_update_stats(VMXNET3State *s,
536                                 int qidx,
537                                 Vmxnet3PktStatus status)
538 {
539     struct UPT1_RxStats *stats = &s->rxq_descr[qidx].rxq_stats;
540     size_t tot_len = net_rx_pkt_get_total_len(s->rx_pkt);
541 
542     switch (status) {
543     case VMXNET3_PKT_STATUS_OUT_OF_BUF:
544         stats->pktsRxOutOfBuf++;
545         break;
546 
547     case VMXNET3_PKT_STATUS_ERROR:
548         stats->pktsRxError++;
549         break;
550     case VMXNET3_PKT_STATUS_OK:
551         switch (net_rx_pkt_get_packet_type(s->rx_pkt)) {
552         case ETH_PKT_BCAST:
553             stats->bcastPktsRxOK++;
554             stats->bcastBytesRxOK += tot_len;
555             break;
556         case ETH_PKT_MCAST:
557             stats->mcastPktsRxOK++;
558             stats->mcastBytesRxOK += tot_len;
559             break;
560         case ETH_PKT_UCAST:
561             stats->ucastPktsRxOK++;
562             stats->ucastBytesRxOK += tot_len;
563             break;
564         default:
565             g_assert_not_reached();
566         }
567 
568         if (tot_len > s->mtu) {
569             stats->LROPktsRxOK++;
570             stats->LROBytesRxOK += tot_len;
571         }
572         break;
573     default:
574         g_assert_not_reached();
575     }
576 }
577 
578 static inline void
579 vmxnet3_ring_read_curr_txdesc(PCIDevice *pcidev, Vmxnet3Ring *ring,
580                               struct Vmxnet3_TxDesc *txd)
581 {
582     vmxnet3_ring_read_curr_cell(pcidev, ring, txd);
583     txd->addr = le64_to_cpu(txd->addr);
584     txd->val1 = le32_to_cpu(txd->val1);
585     txd->val2 = le32_to_cpu(txd->val2);
586 }
587 
588 static inline bool
589 vmxnet3_pop_next_tx_descr(VMXNET3State *s,
590                           int qidx,
591                           struct Vmxnet3_TxDesc *txd,
592                           uint32_t *descr_idx)
593 {
594     Vmxnet3Ring *ring = &s->txq_descr[qidx].tx_ring;
595     PCIDevice *d = PCI_DEVICE(s);
596 
597     vmxnet3_ring_read_curr_txdesc(d, ring, txd);
598     if (txd->gen == vmxnet3_ring_curr_gen(ring)) {
599         /* Only read after generation field verification */
600         smp_rmb();
601         /* Re-read to be sure we got the latest version */
602         vmxnet3_ring_read_curr_txdesc(d, ring, txd);
603         VMXNET3_RING_DUMP(VMW_RIPRN, "TX", qidx, ring);
604         *descr_idx = vmxnet3_ring_curr_cell_idx(ring);
605         vmxnet3_inc_tx_consumption_counter(s, qidx);
606         return true;
607     }
608 
609     return false;
610 }
611 
612 static bool
613 vmxnet3_send_packet(VMXNET3State *s, uint32_t qidx)
614 {
615     Vmxnet3PktStatus status = VMXNET3_PKT_STATUS_OK;
616 
617     if (!vmxnet3_setup_tx_offloads(s)) {
618         status = VMXNET3_PKT_STATUS_ERROR;
619         goto func_exit;
620     }
621 
622     /* debug prints */
623     vmxnet3_dump_virt_hdr(net_tx_pkt_get_vhdr(s->tx_pkt));
624     net_tx_pkt_dump(s->tx_pkt);
625 
626     if (!net_tx_pkt_send(s->tx_pkt, qemu_get_queue(s->nic))) {
627         status = VMXNET3_PKT_STATUS_DISCARD;
628         goto func_exit;
629     }
630 
631 func_exit:
632     vmxnet3_on_tx_done_update_stats(s, qidx, status);
633     return (status == VMXNET3_PKT_STATUS_OK);
634 }
635 
636 static void vmxnet3_process_tx_queue(VMXNET3State *s, int qidx)
637 {
638     struct Vmxnet3_TxDesc txd;
639     uint32_t txd_idx;
640     uint32_t data_len;
641     hwaddr data_pa;
642 
643     for (;;) {
644         if (!vmxnet3_pop_next_tx_descr(s, qidx, &txd, &txd_idx)) {
645             break;
646         }
647 
648         vmxnet3_dump_tx_descr(&txd);
649 
650         if (!s->skip_current_tx_pkt) {
651             data_len = (txd.len > 0) ? txd.len : VMXNET3_MAX_TX_BUF_SIZE;
652             data_pa = txd.addr;
653 
654             if (!net_tx_pkt_add_raw_fragment(s->tx_pkt,
655                                                 data_pa,
656                                                 data_len)) {
657                 s->skip_current_tx_pkt = true;
658             }
659         }
660 
661         if (s->tx_sop) {
662             vmxnet3_tx_retrieve_metadata(s, &txd);
663             s->tx_sop = false;
664         }
665 
666         if (txd.eop) {
667             if (!s->skip_current_tx_pkt && net_tx_pkt_parse(s->tx_pkt)) {
668                 if (s->needs_vlan) {
669                     net_tx_pkt_setup_vlan_header(s->tx_pkt, s->tci);
670                 }
671 
672                 vmxnet3_send_packet(s, qidx);
673             } else {
674                 vmxnet3_on_tx_done_update_stats(s, qidx,
675                                                 VMXNET3_PKT_STATUS_ERROR);
676             }
677 
678             vmxnet3_complete_packet(s, qidx, txd_idx);
679             s->tx_sop = true;
680             s->skip_current_tx_pkt = false;
681             net_tx_pkt_reset(s->tx_pkt);
682         }
683     }
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 isip4, isip6, istcp, isudp;
851     uint8_t *data;
852     int len;
853 
854     if (!net_rx_pkt_has_virt_hdr(pkt)) {
855         return;
856     }
857 
858     vhdr = net_rx_pkt_get_vhdr(pkt);
859     if (!VMXNET_FLAG_IS_SET(vhdr->flags, VIRTIO_NET_HDR_F_NEEDS_CSUM)) {
860         return;
861     }
862 
863     net_rx_pkt_get_protocols(pkt, &isip4, &isip6, &isudp, &istcp);
864     if (!(isip4 || isip6) || !(istcp || isudp)) {
865         return;
866     }
867 
868     vmxnet3_dump_virt_hdr(vhdr);
869 
870     /* Validate packet len: csum_start + scum_offset + length of csum field */
871     if (pkt_len < (vhdr->csum_start + vhdr->csum_offset + 2)) {
872         VMW_PKPRN("packet len:%zu < csum_start(%d) + csum_offset(%d) + 2, "
873                   "cannot calculate checksum",
874                   pkt_len, vhdr->csum_start, vhdr->csum_offset);
875         return;
876     }
877 
878     data = (uint8_t *)pkt_data + vhdr->csum_start;
879     len = pkt_len - vhdr->csum_start;
880     /* Put the checksum obtained into the packet */
881     stw_be_p(data + vhdr->csum_offset,
882              net_checksum_finish_nozero(net_checksum_add(len, data)));
883 
884     vhdr->flags &= ~VIRTIO_NET_HDR_F_NEEDS_CSUM;
885     vhdr->flags |= VIRTIO_NET_HDR_F_DATA_VALID;
886 }
887 
888 static void vmxnet3_rx_update_descr(struct NetRxPkt *pkt,
889     struct Vmxnet3_RxCompDesc *rxcd)
890 {
891     int csum_ok, is_gso;
892     bool isip4, isip6, istcp, isudp;
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     if (!net_rx_pkt_has_virt_hdr(pkt)) {
902         goto nocsum;
903     }
904 
905     vhdr = net_rx_pkt_get_vhdr(pkt);
906     /*
907      * Checksum is valid when lower level tell so or when lower level
908      * requires checksum offload telling that packet produced/bridged
909      * locally and did travel over network after last checksum calculation
910      * or production
911      */
912     csum_ok = VMXNET_FLAG_IS_SET(vhdr->flags, VIRTIO_NET_HDR_F_DATA_VALID) ||
913               VMXNET_FLAG_IS_SET(vhdr->flags, VIRTIO_NET_HDR_F_NEEDS_CSUM);
914 
915     offload_type = vhdr->gso_type & ~VIRTIO_NET_HDR_GSO_ECN;
916     is_gso = (offload_type != VIRTIO_NET_HDR_GSO_NONE) ? 1 : 0;
917 
918     if (!csum_ok && !is_gso) {
919         goto nocsum;
920     }
921 
922     net_rx_pkt_get_protocols(pkt, &isip4, &isip6, &isudp, &istcp);
923     if ((!istcp && !isudp) || (!isip4 && !isip6)) {
924         goto nocsum;
925     }
926 
927     rxcd->cnc = 0;
928     rxcd->v4 = isip4 ? 1 : 0;
929     rxcd->v6 = isip6 ? 1 : 0;
930     rxcd->tcp = istcp ? 1 : 0;
931     rxcd->udp = isudp ? 1 : 0;
932     rxcd->fcs = rxcd->tuc = rxcd->ipc = 1;
933     return;
934 
935 nocsum:
936     rxcd->cnc = 1;
937     return;
938 }
939 
940 static void
941 vmxnet3_pci_dma_writev(PCIDevice *pci_dev,
942                        const struct iovec *iov,
943                        size_t start_iov_off,
944                        hwaddr target_addr,
945                        size_t bytes_to_copy)
946 {
947     size_t curr_off = 0;
948     size_t copied = 0;
949 
950     while (bytes_to_copy) {
951         if (start_iov_off < (curr_off + iov->iov_len)) {
952             size_t chunk_len =
953                 MIN((curr_off + iov->iov_len) - start_iov_off, bytes_to_copy);
954 
955             pci_dma_write(pci_dev, target_addr + copied,
956                           iov->iov_base + start_iov_off - curr_off,
957                           chunk_len);
958 
959             copied += chunk_len;
960             start_iov_off += chunk_len;
961             curr_off = start_iov_off;
962             bytes_to_copy -= chunk_len;
963         } else {
964             curr_off += iov->iov_len;
965         }
966         iov++;
967     }
968 }
969 
970 static void
971 vmxnet3_pci_dma_write_rxcd(PCIDevice *pcidev, dma_addr_t pa,
972                            struct Vmxnet3_RxCompDesc *rxcd)
973 {
974     rxcd->val1 = cpu_to_le32(rxcd->val1);
975     rxcd->val2 = cpu_to_le32(rxcd->val2);
976     rxcd->val3 = cpu_to_le32(rxcd->val3);
977     pci_dma_write(pcidev, pa, rxcd, sizeof(*rxcd));
978 }
979 
980 static bool
981 vmxnet3_indicate_packet(VMXNET3State *s)
982 {
983     struct Vmxnet3_RxDesc rxd;
984     PCIDevice *d = PCI_DEVICE(s);
985     bool is_head = true;
986     uint32_t rxd_idx;
987     uint32_t rx_ridx = 0;
988 
989     struct Vmxnet3_RxCompDesc rxcd;
990     uint32_t new_rxcd_gen = VMXNET3_INIT_GEN;
991     hwaddr new_rxcd_pa = 0;
992     hwaddr ready_rxcd_pa = 0;
993     struct iovec *data = net_rx_pkt_get_iovec(s->rx_pkt);
994     size_t bytes_copied = 0;
995     size_t bytes_left = net_rx_pkt_get_total_len(s->rx_pkt);
996     uint16_t num_frags = 0;
997     size_t chunk_size;
998 
999     net_rx_pkt_dump(s->rx_pkt);
1000 
1001     while (bytes_left > 0) {
1002 
1003         /* cannot add more frags to packet */
1004         if (num_frags == s->max_rx_frags) {
1005             break;
1006         }
1007 
1008         new_rxcd_pa = vmxnet3_pop_rxc_descr(s, RXQ_IDX, &new_rxcd_gen);
1009         if (!new_rxcd_pa) {
1010             break;
1011         }
1012 
1013         if (!vmxnet3_get_next_rx_descr(s, is_head, &rxd, &rxd_idx, &rx_ridx)) {
1014             break;
1015         }
1016 
1017         chunk_size = MIN(bytes_left, rxd.len);
1018         vmxnet3_pci_dma_writev(d, data, bytes_copied, rxd.addr, chunk_size);
1019         bytes_copied += chunk_size;
1020         bytes_left -= chunk_size;
1021 
1022         vmxnet3_dump_rx_descr(&rxd);
1023 
1024         if (ready_rxcd_pa != 0) {
1025             vmxnet3_pci_dma_write_rxcd(d, ready_rxcd_pa, &rxcd);
1026         }
1027 
1028         memset(&rxcd, 0, sizeof(struct Vmxnet3_RxCompDesc));
1029         rxcd.rxdIdx = rxd_idx;
1030         rxcd.len = chunk_size;
1031         rxcd.sop = is_head;
1032         rxcd.gen = new_rxcd_gen;
1033         rxcd.rqID = RXQ_IDX + rx_ridx * s->rxq_num;
1034 
1035         if (bytes_left == 0) {
1036             vmxnet3_rx_update_descr(s->rx_pkt, &rxcd);
1037         }
1038 
1039         VMW_RIPRN("RX Completion descriptor: rxRing: %lu rxIdx %lu len %lu "
1040                   "sop %d csum_correct %lu",
1041                   (unsigned long) rx_ridx,
1042                   (unsigned long) rxcd.rxdIdx,
1043                   (unsigned long) rxcd.len,
1044                   (int) rxcd.sop,
1045                   (unsigned long) rxcd.tuc);
1046 
1047         is_head = false;
1048         ready_rxcd_pa = new_rxcd_pa;
1049         new_rxcd_pa = 0;
1050         num_frags++;
1051     }
1052 
1053     if (ready_rxcd_pa != 0) {
1054         rxcd.eop = 1;
1055         rxcd.err = (bytes_left != 0);
1056 
1057         vmxnet3_pci_dma_write_rxcd(d, ready_rxcd_pa, &rxcd);
1058 
1059         /* Flush RX descriptor changes */
1060         smp_wmb();
1061     }
1062 
1063     if (new_rxcd_pa != 0) {
1064         vmxnet3_revert_rxc_descr(s, RXQ_IDX);
1065     }
1066 
1067     vmxnet3_trigger_interrupt(s, s->rxq_descr[RXQ_IDX].intr_idx);
1068 
1069     if (bytes_left == 0) {
1070         vmxnet3_on_rx_done_update_stats(s, RXQ_IDX, VMXNET3_PKT_STATUS_OK);
1071         return true;
1072     } else if (num_frags == s->max_rx_frags) {
1073         vmxnet3_on_rx_done_update_stats(s, RXQ_IDX, VMXNET3_PKT_STATUS_ERROR);
1074         return false;
1075     } else {
1076         vmxnet3_on_rx_done_update_stats(s, RXQ_IDX,
1077                                         VMXNET3_PKT_STATUS_OUT_OF_BUF);
1078         return false;
1079     }
1080 }
1081 
1082 static void
1083 vmxnet3_io_bar0_write(void *opaque, hwaddr addr,
1084                       uint64_t val, unsigned size)
1085 {
1086     VMXNET3State *s = opaque;
1087 
1088     if (!s->device_active) {
1089         return;
1090     }
1091 
1092     if (VMW_IS_MULTIREG_ADDR(addr, VMXNET3_REG_TXPROD,
1093                         VMXNET3_DEVICE_MAX_TX_QUEUES, VMXNET3_REG_ALIGN)) {
1094         int tx_queue_idx =
1095             VMW_MULTIREG_IDX_BY_ADDR(addr, VMXNET3_REG_TXPROD,
1096                                      VMXNET3_REG_ALIGN);
1097         if (tx_queue_idx <= s->txq_num) {
1098             vmxnet3_process_tx_queue(s, tx_queue_idx);
1099         } else {
1100             qemu_log_mask(LOG_GUEST_ERROR, "vmxnet3: Illegal TX queue %d/%d\n",
1101                           tx_queue_idx, s->txq_num);
1102         }
1103         return;
1104     }
1105 
1106     if (VMW_IS_MULTIREG_ADDR(addr, VMXNET3_REG_IMR,
1107                         VMXNET3_MAX_INTRS, VMXNET3_REG_ALIGN)) {
1108         int l = VMW_MULTIREG_IDX_BY_ADDR(addr, VMXNET3_REG_IMR,
1109                                          VMXNET3_REG_ALIGN);
1110 
1111         VMW_CBPRN("Interrupt mask for line %d written: 0x%" PRIx64, l, val);
1112 
1113         vmxnet3_on_interrupt_mask_changed(s, l, val);
1114         return;
1115     }
1116 
1117     if (VMW_IS_MULTIREG_ADDR(addr, VMXNET3_REG_RXPROD,
1118                         VMXNET3_DEVICE_MAX_RX_QUEUES, VMXNET3_REG_ALIGN) ||
1119        VMW_IS_MULTIREG_ADDR(addr, VMXNET3_REG_RXPROD2,
1120                         VMXNET3_DEVICE_MAX_RX_QUEUES, VMXNET3_REG_ALIGN)) {
1121         return;
1122     }
1123 
1124     VMW_WRPRN("BAR0 unknown write [%" PRIx64 "] = %" PRIx64 ", size %d",
1125               (uint64_t) addr, val, size);
1126 }
1127 
1128 static uint64_t
1129 vmxnet3_io_bar0_read(void *opaque, hwaddr addr, unsigned size)
1130 {
1131     VMXNET3State *s = opaque;
1132 
1133     if (VMW_IS_MULTIREG_ADDR(addr, VMXNET3_REG_IMR,
1134                         VMXNET3_MAX_INTRS, VMXNET3_REG_ALIGN)) {
1135         int l = VMW_MULTIREG_IDX_BY_ADDR(addr, VMXNET3_REG_IMR,
1136                                          VMXNET3_REG_ALIGN);
1137         return s->interrupt_states[l].is_masked;
1138     }
1139 
1140     VMW_CBPRN("BAR0 unknown read [%" PRIx64 "], size %d", addr, size);
1141     return 0;
1142 }
1143 
1144 static void vmxnet3_reset_interrupt_states(VMXNET3State *s)
1145 {
1146     int i;
1147     for (i = 0; i < ARRAY_SIZE(s->interrupt_states); i++) {
1148         s->interrupt_states[i].is_asserted = false;
1149         s->interrupt_states[i].is_pending = false;
1150         s->interrupt_states[i].is_masked = true;
1151     }
1152 }
1153 
1154 static void vmxnet3_reset_mac(VMXNET3State *s)
1155 {
1156     memcpy(&s->conf.macaddr.a, &s->perm_mac.a, sizeof(s->perm_mac.a));
1157     VMW_CFPRN("MAC address set to: " MAC_FMT, MAC_ARG(s->conf.macaddr.a));
1158 }
1159 
1160 static void vmxnet3_deactivate_device(VMXNET3State *s)
1161 {
1162     if (s->device_active) {
1163         VMW_CBPRN("Deactivating vmxnet3...");
1164         net_tx_pkt_reset(s->tx_pkt);
1165         net_tx_pkt_uninit(s->tx_pkt);
1166         net_rx_pkt_uninit(s->rx_pkt);
1167         s->device_active = false;
1168     }
1169 }
1170 
1171 static void vmxnet3_reset(VMXNET3State *s)
1172 {
1173     VMW_CBPRN("Resetting vmxnet3...");
1174 
1175     vmxnet3_deactivate_device(s);
1176     vmxnet3_reset_interrupt_states(s);
1177     s->drv_shmem = 0;
1178     s->tx_sop = true;
1179     s->skip_current_tx_pkt = false;
1180 }
1181 
1182 static void vmxnet3_update_rx_mode(VMXNET3State *s)
1183 {
1184     PCIDevice *d = PCI_DEVICE(s);
1185 
1186     s->rx_mode = VMXNET3_READ_DRV_SHARED32(d, s->drv_shmem,
1187                                            devRead.rxFilterConf.rxMode);
1188     VMW_CFPRN("RX mode: 0x%08X", s->rx_mode);
1189 }
1190 
1191 static void vmxnet3_update_vlan_filters(VMXNET3State *s)
1192 {
1193     int i;
1194     PCIDevice *d = PCI_DEVICE(s);
1195 
1196     /* Copy configuration from shared memory */
1197     VMXNET3_READ_DRV_SHARED(d, s->drv_shmem,
1198                             devRead.rxFilterConf.vfTable,
1199                             s->vlan_table,
1200                             sizeof(s->vlan_table));
1201 
1202     /* Invert byte order when needed */
1203     for (i = 0; i < ARRAY_SIZE(s->vlan_table); i++) {
1204         s->vlan_table[i] = le32_to_cpu(s->vlan_table[i]);
1205     }
1206 
1207     /* Dump configuration for debugging purposes */
1208     VMW_CFPRN("Configured VLANs:");
1209     for (i = 0; i < sizeof(s->vlan_table) * 8; i++) {
1210         if (VMXNET3_VFTABLE_ENTRY_IS_SET(s->vlan_table, i)) {
1211             VMW_CFPRN("\tVLAN %d is present", i);
1212         }
1213     }
1214 }
1215 
1216 static void vmxnet3_update_mcast_filters(VMXNET3State *s)
1217 {
1218     PCIDevice *d = PCI_DEVICE(s);
1219 
1220     uint16_t list_bytes =
1221         VMXNET3_READ_DRV_SHARED16(d, s->drv_shmem,
1222                                   devRead.rxFilterConf.mfTableLen);
1223 
1224     s->mcast_list_len = list_bytes / sizeof(s->mcast_list[0]);
1225 
1226     s->mcast_list = g_realloc(s->mcast_list, list_bytes);
1227     if (!s->mcast_list) {
1228         if (s->mcast_list_len == 0) {
1229             VMW_CFPRN("Current multicast list is empty");
1230         } else {
1231             VMW_ERPRN("Failed to allocate multicast list of %d elements",
1232                       s->mcast_list_len);
1233         }
1234         s->mcast_list_len = 0;
1235     } else {
1236         int i;
1237         hwaddr mcast_list_pa =
1238             VMXNET3_READ_DRV_SHARED64(d, s->drv_shmem,
1239                                       devRead.rxFilterConf.mfTablePA);
1240 
1241         pci_dma_read(d, mcast_list_pa, s->mcast_list, list_bytes);
1242 
1243         VMW_CFPRN("Current multicast list len is %d:", s->mcast_list_len);
1244         for (i = 0; i < s->mcast_list_len; i++) {
1245             VMW_CFPRN("\t" MAC_FMT, MAC_ARG(s->mcast_list[i].a));
1246         }
1247     }
1248 }
1249 
1250 static void vmxnet3_setup_rx_filtering(VMXNET3State *s)
1251 {
1252     vmxnet3_update_rx_mode(s);
1253     vmxnet3_update_vlan_filters(s);
1254     vmxnet3_update_mcast_filters(s);
1255 }
1256 
1257 static uint32_t vmxnet3_get_interrupt_config(VMXNET3State *s)
1258 {
1259     uint32_t interrupt_mode = VMXNET3_IT_AUTO | (VMXNET3_IMM_AUTO << 2);
1260     VMW_CFPRN("Interrupt config is 0x%X", interrupt_mode);
1261     return interrupt_mode;
1262 }
1263 
1264 static void vmxnet3_fill_stats(VMXNET3State *s)
1265 {
1266     int i;
1267     PCIDevice *d = PCI_DEVICE(s);
1268 
1269     if (!s->device_active)
1270         return;
1271 
1272     for (i = 0; i < s->txq_num; i++) {
1273         pci_dma_write(d,
1274                       s->txq_descr[i].tx_stats_pa,
1275                       &s->txq_descr[i].txq_stats,
1276                       sizeof(s->txq_descr[i].txq_stats));
1277     }
1278 
1279     for (i = 0; i < s->rxq_num; i++) {
1280         pci_dma_write(d,
1281                       s->rxq_descr[i].rx_stats_pa,
1282                       &s->rxq_descr[i].rxq_stats,
1283                       sizeof(s->rxq_descr[i].rxq_stats));
1284     }
1285 }
1286 
1287 static void vmxnet3_adjust_by_guest_type(VMXNET3State *s)
1288 {
1289     struct Vmxnet3_GOSInfo gos;
1290     PCIDevice *d = PCI_DEVICE(s);
1291 
1292     VMXNET3_READ_DRV_SHARED(d, s->drv_shmem, devRead.misc.driverInfo.gos,
1293                             &gos, sizeof(gos));
1294     s->rx_packets_compound =
1295         (gos.gosType == VMXNET3_GOS_TYPE_WIN) ? false : true;
1296 
1297     VMW_CFPRN("Guest type specifics: RXCOMPOUND: %d", s->rx_packets_compound);
1298 }
1299 
1300 static void
1301 vmxnet3_dump_conf_descr(const char *name,
1302                         struct Vmxnet3_VariableLenConfDesc *pm_descr)
1303 {
1304     VMW_CFPRN("%s descriptor dump: Version %u, Length %u",
1305               name, pm_descr->confVer, pm_descr->confLen);
1306 
1307 };
1308 
1309 static void vmxnet3_update_pm_state(VMXNET3State *s)
1310 {
1311     struct Vmxnet3_VariableLenConfDesc pm_descr;
1312     PCIDevice *d = PCI_DEVICE(s);
1313 
1314     pm_descr.confLen =
1315         VMXNET3_READ_DRV_SHARED32(d, s->drv_shmem, devRead.pmConfDesc.confLen);
1316     pm_descr.confVer =
1317         VMXNET3_READ_DRV_SHARED32(d, s->drv_shmem, devRead.pmConfDesc.confVer);
1318     pm_descr.confPA =
1319         VMXNET3_READ_DRV_SHARED64(d, s->drv_shmem, devRead.pmConfDesc.confPA);
1320 
1321     vmxnet3_dump_conf_descr("PM State", &pm_descr);
1322 }
1323 
1324 static void vmxnet3_update_features(VMXNET3State *s)
1325 {
1326     uint32_t guest_features;
1327     int rxcso_supported;
1328     PCIDevice *d = PCI_DEVICE(s);
1329 
1330     guest_features = VMXNET3_READ_DRV_SHARED32(d, s->drv_shmem,
1331                                                devRead.misc.uptFeatures);
1332 
1333     rxcso_supported = VMXNET_FLAG_IS_SET(guest_features, UPT1_F_RXCSUM);
1334     s->rx_vlan_stripping = VMXNET_FLAG_IS_SET(guest_features, UPT1_F_RXVLAN);
1335     s->lro_supported = VMXNET_FLAG_IS_SET(guest_features, UPT1_F_LRO);
1336 
1337     VMW_CFPRN("Features configuration: LRO: %d, RXCSUM: %d, VLANSTRIP: %d",
1338               s->lro_supported, rxcso_supported,
1339               s->rx_vlan_stripping);
1340     if (s->peer_has_vhdr) {
1341         qemu_set_offload(qemu_get_queue(s->nic)->peer,
1342                          rxcso_supported,
1343                          s->lro_supported,
1344                          s->lro_supported,
1345                          0,
1346                          0);
1347     }
1348 }
1349 
1350 static bool vmxnet3_verify_intx(VMXNET3State *s, int intx)
1351 {
1352     return s->msix_used || msi_enabled(PCI_DEVICE(s))
1353         || intx == pci_get_byte(s->parent_obj.config + PCI_INTERRUPT_PIN) - 1;
1354 }
1355 
1356 static void vmxnet3_validate_interrupt_idx(bool is_msix, int idx)
1357 {
1358     int max_ints = is_msix ? VMXNET3_MAX_INTRS : VMXNET3_MAX_NMSIX_INTRS;
1359     if (idx >= max_ints) {
1360         hw_error("Bad interrupt index: %d\n", idx);
1361     }
1362 }
1363 
1364 static void vmxnet3_validate_interrupts(VMXNET3State *s)
1365 {
1366     int i;
1367 
1368     VMW_CFPRN("Verifying event interrupt index (%d)", s->event_int_idx);
1369     vmxnet3_validate_interrupt_idx(s->msix_used, s->event_int_idx);
1370 
1371     for (i = 0; i < s->txq_num; i++) {
1372         int idx = s->txq_descr[i].intr_idx;
1373         VMW_CFPRN("Verifying TX queue %d interrupt index (%d)", i, idx);
1374         vmxnet3_validate_interrupt_idx(s->msix_used, idx);
1375     }
1376 
1377     for (i = 0; i < s->rxq_num; i++) {
1378         int idx = s->rxq_descr[i].intr_idx;
1379         VMW_CFPRN("Verifying RX queue %d interrupt index (%d)", i, idx);
1380         vmxnet3_validate_interrupt_idx(s->msix_used, idx);
1381     }
1382 }
1383 
1384 static bool vmxnet3_validate_queues(VMXNET3State *s)
1385 {
1386     /*
1387     * txq_num and rxq_num are total number of queues
1388     * configured by guest. These numbers must not
1389     * exceed corresponding maximal values.
1390     */
1391 
1392     if (s->txq_num > VMXNET3_DEVICE_MAX_TX_QUEUES) {
1393         qemu_log_mask(LOG_GUEST_ERROR, "vmxnet3: Bad TX queues number: %d\n",
1394                       s->txq_num);
1395         return false;
1396     }
1397 
1398     if (s->rxq_num > VMXNET3_DEVICE_MAX_RX_QUEUES) {
1399         qemu_log_mask(LOG_GUEST_ERROR, "vmxnet3: Bad RX queues number: %d\n",
1400                       s->rxq_num);
1401         return false;
1402     }
1403 
1404     return true;
1405 }
1406 
1407 static void vmxnet3_activate_device(VMXNET3State *s)
1408 {
1409     int i;
1410     static const uint32_t VMXNET3_DEF_TX_THRESHOLD = 1;
1411     PCIDevice *d = PCI_DEVICE(s);
1412     hwaddr qdescr_table_pa;
1413     uint64_t pa;
1414     uint32_t size;
1415 
1416     /* Verify configuration consistency */
1417     if (!vmxnet3_verify_driver_magic(d, s->drv_shmem)) {
1418         VMW_ERPRN("Device configuration received from driver is invalid");
1419         return;
1420     }
1421 
1422     /* Verify if device is active */
1423     if (s->device_active) {
1424         VMW_CFPRN("Vmxnet3 device is active");
1425         return;
1426     }
1427 
1428     s->txq_num =
1429         VMXNET3_READ_DRV_SHARED8(d, s->drv_shmem, devRead.misc.numTxQueues);
1430     s->rxq_num =
1431         VMXNET3_READ_DRV_SHARED8(d, s->drv_shmem, devRead.misc.numRxQueues);
1432 
1433     VMW_CFPRN("Number of TX/RX queues %u/%u", s->txq_num, s->rxq_num);
1434     if (!vmxnet3_validate_queues(s)) {
1435         return;
1436     }
1437 
1438     vmxnet3_adjust_by_guest_type(s);
1439     vmxnet3_update_features(s);
1440     vmxnet3_update_pm_state(s);
1441     vmxnet3_setup_rx_filtering(s);
1442     /* Cache fields from shared memory */
1443     s->mtu = VMXNET3_READ_DRV_SHARED32(d, s->drv_shmem, devRead.misc.mtu);
1444     assert(VMXNET3_MIN_MTU <= s->mtu && s->mtu < VMXNET3_MAX_MTU);
1445     VMW_CFPRN("MTU is %u", s->mtu);
1446 
1447     s->max_rx_frags =
1448         VMXNET3_READ_DRV_SHARED16(d, s->drv_shmem, devRead.misc.maxNumRxSG);
1449 
1450     if (s->max_rx_frags == 0) {
1451         s->max_rx_frags = 1;
1452     }
1453 
1454     VMW_CFPRN("Max RX fragments is %u", s->max_rx_frags);
1455 
1456     s->event_int_idx =
1457         VMXNET3_READ_DRV_SHARED8(d, s->drv_shmem, devRead.intrConf.eventIntrIdx);
1458     assert(vmxnet3_verify_intx(s, s->event_int_idx));
1459     VMW_CFPRN("Events interrupt line is %u", s->event_int_idx);
1460 
1461     s->auto_int_masking =
1462         VMXNET3_READ_DRV_SHARED8(d, s->drv_shmem, devRead.intrConf.autoMask);
1463     VMW_CFPRN("Automatic interrupt masking is %d", (int)s->auto_int_masking);
1464 
1465     qdescr_table_pa =
1466         VMXNET3_READ_DRV_SHARED64(d, s->drv_shmem, devRead.misc.queueDescPA);
1467     VMW_CFPRN("TX queues descriptors table is at 0x%" PRIx64, qdescr_table_pa);
1468 
1469     /*
1470      * Worst-case scenario is a packet that holds all TX rings space so
1471      * we calculate total size of all TX rings for max TX fragments number
1472      */
1473     s->max_tx_frags = 0;
1474 
1475     /* TX queues */
1476     for (i = 0; i < s->txq_num; i++) {
1477         hwaddr qdescr_pa =
1478             qdescr_table_pa + i * sizeof(struct Vmxnet3_TxQueueDesc);
1479 
1480         /* Read interrupt number for this TX queue */
1481         s->txq_descr[i].intr_idx =
1482             VMXNET3_READ_TX_QUEUE_DESCR8(d, qdescr_pa, conf.intrIdx);
1483         assert(vmxnet3_verify_intx(s, s->txq_descr[i].intr_idx));
1484 
1485         VMW_CFPRN("TX Queue %d interrupt: %d", i, s->txq_descr[i].intr_idx);
1486 
1487         /* Read rings memory locations for TX queues */
1488         pa = VMXNET3_READ_TX_QUEUE_DESCR64(d, qdescr_pa, conf.txRingBasePA);
1489         size = VMXNET3_READ_TX_QUEUE_DESCR32(d, qdescr_pa, conf.txRingSize);
1490         if (size > VMXNET3_TX_RING_MAX_SIZE) {
1491             size = VMXNET3_TX_RING_MAX_SIZE;
1492         }
1493 
1494         vmxnet3_ring_init(d, &s->txq_descr[i].tx_ring, pa, size,
1495                           sizeof(struct Vmxnet3_TxDesc), false);
1496         VMXNET3_RING_DUMP(VMW_CFPRN, "TX", i, &s->txq_descr[i].tx_ring);
1497 
1498         s->max_tx_frags += size;
1499 
1500         /* TXC ring */
1501         pa = VMXNET3_READ_TX_QUEUE_DESCR64(d, qdescr_pa, conf.compRingBasePA);
1502         size = VMXNET3_READ_TX_QUEUE_DESCR32(d, qdescr_pa, conf.compRingSize);
1503         if (size > VMXNET3_TC_RING_MAX_SIZE) {
1504             size = VMXNET3_TC_RING_MAX_SIZE;
1505         }
1506         vmxnet3_ring_init(d, &s->txq_descr[i].comp_ring, pa, size,
1507                           sizeof(struct Vmxnet3_TxCompDesc), true);
1508         VMXNET3_RING_DUMP(VMW_CFPRN, "TXC", i, &s->txq_descr[i].comp_ring);
1509 
1510         s->txq_descr[i].tx_stats_pa =
1511             qdescr_pa + offsetof(struct Vmxnet3_TxQueueDesc, stats);
1512 
1513         memset(&s->txq_descr[i].txq_stats, 0,
1514                sizeof(s->txq_descr[i].txq_stats));
1515 
1516         /* Fill device-managed parameters for queues */
1517         VMXNET3_WRITE_TX_QUEUE_DESCR32(d, qdescr_pa,
1518                                        ctrl.txThreshold,
1519                                        VMXNET3_DEF_TX_THRESHOLD);
1520     }
1521 
1522     /* Preallocate TX packet wrapper */
1523     VMW_CFPRN("Max TX fragments is %u", s->max_tx_frags);
1524     net_tx_pkt_init(&s->tx_pkt, PCI_DEVICE(s),
1525                     s->max_tx_frags, s->peer_has_vhdr);
1526     net_rx_pkt_init(&s->rx_pkt, s->peer_has_vhdr);
1527 
1528     /* Read rings memory locations for RX queues */
1529     for (i = 0; i < s->rxq_num; i++) {
1530         int j;
1531         hwaddr qd_pa =
1532             qdescr_table_pa + s->txq_num * sizeof(struct Vmxnet3_TxQueueDesc) +
1533             i * sizeof(struct Vmxnet3_RxQueueDesc);
1534 
1535         /* Read interrupt number for this RX queue */
1536         s->rxq_descr[i].intr_idx =
1537             VMXNET3_READ_TX_QUEUE_DESCR8(d, qd_pa, conf.intrIdx);
1538         assert(vmxnet3_verify_intx(s, s->rxq_descr[i].intr_idx));
1539 
1540         VMW_CFPRN("RX Queue %d interrupt: %d", i, s->rxq_descr[i].intr_idx);
1541 
1542         /* Read rings memory locations */
1543         for (j = 0; j < VMXNET3_RX_RINGS_PER_QUEUE; j++) {
1544             /* RX rings */
1545             pa = VMXNET3_READ_RX_QUEUE_DESCR64(d, qd_pa, conf.rxRingBasePA[j]);
1546             size = VMXNET3_READ_RX_QUEUE_DESCR32(d, qd_pa, conf.rxRingSize[j]);
1547             if (size > VMXNET3_RX_RING_MAX_SIZE) {
1548                 size = VMXNET3_RX_RING_MAX_SIZE;
1549             }
1550             vmxnet3_ring_init(d, &s->rxq_descr[i].rx_ring[j], pa, size,
1551                               sizeof(struct Vmxnet3_RxDesc), false);
1552             VMW_CFPRN("RX queue %d:%d: Base: %" PRIx64 ", Size: %d",
1553                       i, j, pa, size);
1554         }
1555 
1556         /* RXC ring */
1557         pa = VMXNET3_READ_RX_QUEUE_DESCR64(d, qd_pa, conf.compRingBasePA);
1558         size = VMXNET3_READ_RX_QUEUE_DESCR32(d, qd_pa, conf.compRingSize);
1559         if (size > VMXNET3_RC_RING_MAX_SIZE) {
1560             size = VMXNET3_RC_RING_MAX_SIZE;
1561         }
1562         vmxnet3_ring_init(d, &s->rxq_descr[i].comp_ring, pa, size,
1563                           sizeof(struct Vmxnet3_RxCompDesc), true);
1564         VMW_CFPRN("RXC queue %d: Base: %" PRIx64 ", Size: %d", i, pa, size);
1565 
1566         s->rxq_descr[i].rx_stats_pa =
1567             qd_pa + offsetof(struct Vmxnet3_RxQueueDesc, stats);
1568         memset(&s->rxq_descr[i].rxq_stats, 0,
1569                sizeof(s->rxq_descr[i].rxq_stats));
1570     }
1571 
1572     vmxnet3_validate_interrupts(s);
1573 
1574     /* Make sure everything is in place before device activation */
1575     smp_wmb();
1576 
1577     vmxnet3_reset_mac(s);
1578 
1579     s->device_active = true;
1580 }
1581 
1582 static void vmxnet3_handle_command(VMXNET3State *s, uint64_t cmd)
1583 {
1584     s->last_command = cmd;
1585 
1586     switch (cmd) {
1587     case VMXNET3_CMD_GET_PERM_MAC_HI:
1588         VMW_CBPRN("Set: Get upper part of permanent MAC");
1589         break;
1590 
1591     case VMXNET3_CMD_GET_PERM_MAC_LO:
1592         VMW_CBPRN("Set: Get lower part of permanent MAC");
1593         break;
1594 
1595     case VMXNET3_CMD_GET_STATS:
1596         VMW_CBPRN("Set: Get device statistics");
1597         vmxnet3_fill_stats(s);
1598         break;
1599 
1600     case VMXNET3_CMD_ACTIVATE_DEV:
1601         VMW_CBPRN("Set: Activating vmxnet3 device");
1602         vmxnet3_activate_device(s);
1603         break;
1604 
1605     case VMXNET3_CMD_UPDATE_RX_MODE:
1606         VMW_CBPRN("Set: Update rx mode");
1607         vmxnet3_update_rx_mode(s);
1608         break;
1609 
1610     case VMXNET3_CMD_UPDATE_VLAN_FILTERS:
1611         VMW_CBPRN("Set: Update VLAN filters");
1612         vmxnet3_update_vlan_filters(s);
1613         break;
1614 
1615     case VMXNET3_CMD_UPDATE_MAC_FILTERS:
1616         VMW_CBPRN("Set: Update MAC filters");
1617         vmxnet3_update_mcast_filters(s);
1618         break;
1619 
1620     case VMXNET3_CMD_UPDATE_FEATURE:
1621         VMW_CBPRN("Set: Update features");
1622         vmxnet3_update_features(s);
1623         break;
1624 
1625     case VMXNET3_CMD_UPDATE_PMCFG:
1626         VMW_CBPRN("Set: Update power management config");
1627         vmxnet3_update_pm_state(s);
1628         break;
1629 
1630     case VMXNET3_CMD_GET_LINK:
1631         VMW_CBPRN("Set: Get link");
1632         break;
1633 
1634     case VMXNET3_CMD_RESET_DEV:
1635         VMW_CBPRN("Set: Reset device");
1636         vmxnet3_reset(s);
1637         break;
1638 
1639     case VMXNET3_CMD_QUIESCE_DEV:
1640         VMW_CBPRN("Set: VMXNET3_CMD_QUIESCE_DEV - deactivate the device");
1641         vmxnet3_deactivate_device(s);
1642         break;
1643 
1644     case VMXNET3_CMD_GET_CONF_INTR:
1645         VMW_CBPRN("Set: VMXNET3_CMD_GET_CONF_INTR - interrupt configuration");
1646         break;
1647 
1648     case VMXNET3_CMD_GET_ADAPTIVE_RING_INFO:
1649         VMW_CBPRN("Set: VMXNET3_CMD_GET_ADAPTIVE_RING_INFO - "
1650                   "adaptive ring info flags");
1651         break;
1652 
1653     case VMXNET3_CMD_GET_DID_LO:
1654         VMW_CBPRN("Set: Get lower part of device ID");
1655         break;
1656 
1657     case VMXNET3_CMD_GET_DID_HI:
1658         VMW_CBPRN("Set: Get upper part of device ID");
1659         break;
1660 
1661     case VMXNET3_CMD_GET_DEV_EXTRA_INFO:
1662         VMW_CBPRN("Set: Get device extra info");
1663         break;
1664 
1665     default:
1666         VMW_CBPRN("Received unknown command: %" PRIx64, cmd);
1667         break;
1668     }
1669 }
1670 
1671 static uint64_t vmxnet3_get_command_status(VMXNET3State *s)
1672 {
1673     uint64_t ret;
1674 
1675     switch (s->last_command) {
1676     case VMXNET3_CMD_ACTIVATE_DEV:
1677         ret = (s->device_active) ? 0 : 1;
1678         VMW_CFPRN("Device active: %" PRIx64, ret);
1679         break;
1680 
1681     case VMXNET3_CMD_RESET_DEV:
1682     case VMXNET3_CMD_QUIESCE_DEV:
1683     case VMXNET3_CMD_GET_QUEUE_STATUS:
1684     case VMXNET3_CMD_GET_DEV_EXTRA_INFO:
1685         ret = 0;
1686         break;
1687 
1688     case VMXNET3_CMD_GET_LINK:
1689         ret = s->link_status_and_speed;
1690         VMW_CFPRN("Link and speed: %" PRIx64, ret);
1691         break;
1692 
1693     case VMXNET3_CMD_GET_PERM_MAC_LO:
1694         ret = vmxnet3_get_mac_low(&s->perm_mac);
1695         break;
1696 
1697     case VMXNET3_CMD_GET_PERM_MAC_HI:
1698         ret = vmxnet3_get_mac_high(&s->perm_mac);
1699         break;
1700 
1701     case VMXNET3_CMD_GET_CONF_INTR:
1702         ret = vmxnet3_get_interrupt_config(s);
1703         break;
1704 
1705     case VMXNET3_CMD_GET_ADAPTIVE_RING_INFO:
1706         ret = VMXNET3_DISABLE_ADAPTIVE_RING;
1707         break;
1708 
1709     case VMXNET3_CMD_GET_DID_LO:
1710         ret = PCI_DEVICE_ID_VMWARE_VMXNET3;
1711         break;
1712 
1713     case VMXNET3_CMD_GET_DID_HI:
1714         ret = VMXNET3_DEVICE_REVISION;
1715         break;
1716 
1717     default:
1718         VMW_WRPRN("Received request for unknown command: %x", s->last_command);
1719         ret = 0;
1720         break;
1721     }
1722 
1723     return ret;
1724 }
1725 
1726 static void vmxnet3_set_events(VMXNET3State *s, uint32_t val)
1727 {
1728     uint32_t events;
1729     PCIDevice *d = PCI_DEVICE(s);
1730 
1731     VMW_CBPRN("Setting events: 0x%x", val);
1732     events = VMXNET3_READ_DRV_SHARED32(d, s->drv_shmem, ecr) | val;
1733     VMXNET3_WRITE_DRV_SHARED32(d, s->drv_shmem, ecr, events);
1734 }
1735 
1736 static void vmxnet3_ack_events(VMXNET3State *s, uint32_t val)
1737 {
1738     PCIDevice *d = PCI_DEVICE(s);
1739     uint32_t events;
1740 
1741     VMW_CBPRN("Clearing events: 0x%x", val);
1742     events = VMXNET3_READ_DRV_SHARED32(d, s->drv_shmem, ecr) & ~val;
1743     VMXNET3_WRITE_DRV_SHARED32(d, s->drv_shmem, ecr, events);
1744 }
1745 
1746 static void
1747 vmxnet3_io_bar1_write(void *opaque,
1748                       hwaddr addr,
1749                       uint64_t val,
1750                       unsigned size)
1751 {
1752     VMXNET3State *s = opaque;
1753 
1754     switch (addr) {
1755     /* Vmxnet3 Revision Report Selection */
1756     case VMXNET3_REG_VRRS:
1757         VMW_CBPRN("Write BAR1 [VMXNET3_REG_VRRS] = %" PRIx64 ", size %d",
1758                   val, size);
1759         break;
1760 
1761     /* UPT Version Report Selection */
1762     case VMXNET3_REG_UVRS:
1763         VMW_CBPRN("Write BAR1 [VMXNET3_REG_UVRS] = %" PRIx64 ", size %d",
1764                   val, size);
1765         break;
1766 
1767     /* Driver Shared Address Low */
1768     case VMXNET3_REG_DSAL:
1769         VMW_CBPRN("Write BAR1 [VMXNET3_REG_DSAL] = %" PRIx64 ", size %d",
1770                   val, size);
1771         /*
1772          * Guest driver will first write the low part of the shared
1773          * memory address. We save it to temp variable and set the
1774          * shared address only after we get the high part
1775          */
1776         if (val == 0) {
1777             vmxnet3_deactivate_device(s);
1778         }
1779         s->temp_shared_guest_driver_memory = val;
1780         s->drv_shmem = 0;
1781         break;
1782 
1783     /* Driver Shared Address High */
1784     case VMXNET3_REG_DSAH:
1785         VMW_CBPRN("Write BAR1 [VMXNET3_REG_DSAH] = %" PRIx64 ", size %d",
1786                   val, size);
1787         /*
1788          * Set the shared memory between guest driver and device.
1789          * We already should have low address part.
1790          */
1791         s->drv_shmem = s->temp_shared_guest_driver_memory | (val << 32);
1792         break;
1793 
1794     /* Command */
1795     case VMXNET3_REG_CMD:
1796         VMW_CBPRN("Write BAR1 [VMXNET3_REG_CMD] = %" PRIx64 ", size %d",
1797                   val, size);
1798         vmxnet3_handle_command(s, val);
1799         break;
1800 
1801     /* MAC Address Low */
1802     case VMXNET3_REG_MACL:
1803         VMW_CBPRN("Write BAR1 [VMXNET3_REG_MACL] = %" PRIx64 ", size %d",
1804                   val, size);
1805         s->temp_mac = val;
1806         break;
1807 
1808     /* MAC Address High */
1809     case VMXNET3_REG_MACH:
1810         VMW_CBPRN("Write BAR1 [VMXNET3_REG_MACH] = %" PRIx64 ", size %d",
1811                   val, size);
1812         vmxnet3_set_variable_mac(s, val, s->temp_mac);
1813         break;
1814 
1815     /* Interrupt Cause Register */
1816     case VMXNET3_REG_ICR:
1817         VMW_CBPRN("Write BAR1 [VMXNET3_REG_ICR] = %" PRIx64 ", size %d",
1818                   val, size);
1819         qemu_log_mask(LOG_GUEST_ERROR,
1820                       "%s: write to read-only register VMXNET3_REG_ICR\n",
1821                       TYPE_VMXNET3);
1822         break;
1823 
1824     /* Event Cause Register */
1825     case VMXNET3_REG_ECR:
1826         VMW_CBPRN("Write BAR1 [VMXNET3_REG_ECR] = %" PRIx64 ", size %d",
1827                   val, size);
1828         vmxnet3_ack_events(s, val);
1829         break;
1830 
1831     default:
1832         VMW_CBPRN("Unknown Write to BAR1 [%" PRIx64 "] = %" PRIx64 ", size %d",
1833                   addr, val, size);
1834         break;
1835     }
1836 }
1837 
1838 static uint64_t
1839 vmxnet3_io_bar1_read(void *opaque, hwaddr addr, unsigned size)
1840 {
1841         VMXNET3State *s = opaque;
1842         uint64_t ret = 0;
1843 
1844         switch (addr) {
1845         /* Vmxnet3 Revision Report Selection */
1846         case VMXNET3_REG_VRRS:
1847             VMW_CBPRN("Read BAR1 [VMXNET3_REG_VRRS], size %d", size);
1848             ret = VMXNET3_DEVICE_REVISION;
1849             break;
1850 
1851         /* UPT Version Report Selection */
1852         case VMXNET3_REG_UVRS:
1853             VMW_CBPRN("Read BAR1 [VMXNET3_REG_UVRS], size %d", size);
1854             ret = VMXNET3_UPT_REVISION;
1855             break;
1856 
1857         /* Command */
1858         case VMXNET3_REG_CMD:
1859             VMW_CBPRN("Read BAR1 [VMXNET3_REG_CMD], size %d", size);
1860             ret = vmxnet3_get_command_status(s);
1861             break;
1862 
1863         /* MAC Address Low */
1864         case VMXNET3_REG_MACL:
1865             VMW_CBPRN("Read BAR1 [VMXNET3_REG_MACL], size %d", size);
1866             ret = vmxnet3_get_mac_low(&s->conf.macaddr);
1867             break;
1868 
1869         /* MAC Address High */
1870         case VMXNET3_REG_MACH:
1871             VMW_CBPRN("Read BAR1 [VMXNET3_REG_MACH], size %d", size);
1872             ret = vmxnet3_get_mac_high(&s->conf.macaddr);
1873             break;
1874 
1875         /*
1876          * Interrupt Cause Register
1877          * Used for legacy interrupts only so interrupt index always 0
1878          */
1879         case VMXNET3_REG_ICR:
1880             VMW_CBPRN("Read BAR1 [VMXNET3_REG_ICR], size %d", size);
1881             if (vmxnet3_interrupt_asserted(s, 0)) {
1882                 vmxnet3_clear_interrupt(s, 0);
1883                 ret = true;
1884             } else {
1885                 ret = false;
1886             }
1887             break;
1888 
1889         default:
1890             VMW_CBPRN("Unknow read BAR1[%" PRIx64 "], %d bytes", addr, size);
1891             break;
1892         }
1893 
1894         return ret;
1895 }
1896 
1897 static int
1898 vmxnet3_can_receive(NetClientState *nc)
1899 {
1900     VMXNET3State *s = qemu_get_nic_opaque(nc);
1901     return s->device_active &&
1902            VMXNET_FLAG_IS_SET(s->link_status_and_speed, VMXNET3_LINK_STATUS_UP);
1903 }
1904 
1905 static inline bool
1906 vmxnet3_is_registered_vlan(VMXNET3State *s, const void *data)
1907 {
1908     uint16_t vlan_tag = eth_get_pkt_tci(data) & VLAN_VID_MASK;
1909     if (IS_SPECIAL_VLAN_ID(vlan_tag)) {
1910         return true;
1911     }
1912 
1913     return VMXNET3_VFTABLE_ENTRY_IS_SET(s->vlan_table, vlan_tag);
1914 }
1915 
1916 static bool
1917 vmxnet3_is_allowed_mcast_group(VMXNET3State *s, const uint8_t *group_mac)
1918 {
1919     int i;
1920     for (i = 0; i < s->mcast_list_len; i++) {
1921         if (!memcmp(group_mac, s->mcast_list[i].a, sizeof(s->mcast_list[i]))) {
1922             return true;
1923         }
1924     }
1925     return false;
1926 }
1927 
1928 static bool
1929 vmxnet3_rx_filter_may_indicate(VMXNET3State *s, const void *data,
1930     size_t size)
1931 {
1932     struct eth_header *ehdr = PKT_GET_ETH_HDR(data);
1933 
1934     if (VMXNET_FLAG_IS_SET(s->rx_mode, VMXNET3_RXM_PROMISC)) {
1935         return true;
1936     }
1937 
1938     if (!vmxnet3_is_registered_vlan(s, data)) {
1939         return false;
1940     }
1941 
1942     switch (net_rx_pkt_get_packet_type(s->rx_pkt)) {
1943     case ETH_PKT_UCAST:
1944         if (!VMXNET_FLAG_IS_SET(s->rx_mode, VMXNET3_RXM_UCAST)) {
1945             return false;
1946         }
1947         if (memcmp(s->conf.macaddr.a, ehdr->h_dest, ETH_ALEN)) {
1948             return false;
1949         }
1950         break;
1951 
1952     case ETH_PKT_BCAST:
1953         if (!VMXNET_FLAG_IS_SET(s->rx_mode, VMXNET3_RXM_BCAST)) {
1954             return false;
1955         }
1956         break;
1957 
1958     case ETH_PKT_MCAST:
1959         if (VMXNET_FLAG_IS_SET(s->rx_mode, VMXNET3_RXM_ALL_MULTI)) {
1960             return true;
1961         }
1962         if (!VMXNET_FLAG_IS_SET(s->rx_mode, VMXNET3_RXM_MCAST)) {
1963             return false;
1964         }
1965         if (!vmxnet3_is_allowed_mcast_group(s, ehdr->h_dest)) {
1966             return false;
1967         }
1968         break;
1969 
1970     default:
1971         g_assert_not_reached();
1972     }
1973 
1974     return true;
1975 }
1976 
1977 static ssize_t
1978 vmxnet3_receive(NetClientState *nc, const uint8_t *buf, size_t size)
1979 {
1980     VMXNET3State *s = qemu_get_nic_opaque(nc);
1981     size_t bytes_indicated;
1982     uint8_t min_buf[MIN_BUF_SIZE];
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     /* Pad to minimum Ethernet frame length */
1996     if (size < sizeof(min_buf)) {
1997         memcpy(min_buf, buf, size);
1998         memset(&min_buf[size], 0, sizeof(min_buf) - size);
1999         buf = min_buf;
2000         size = sizeof(min_buf);
2001     }
2002 
2003     net_rx_pkt_set_packet_type(s->rx_pkt,
2004         get_eth_packet_type(PKT_GET_ETH_HDR(buf)));
2005 
2006     if (vmxnet3_rx_filter_may_indicate(s, buf, size)) {
2007         net_rx_pkt_set_protocols(s->rx_pkt, buf, size);
2008         vmxnet3_rx_need_csum_calculate(s->rx_pkt, buf, size);
2009         net_rx_pkt_attach_data(s->rx_pkt, buf, size, s->rx_vlan_stripping);
2010         bytes_indicated = vmxnet3_indicate_packet(s) ? size : -1;
2011         if (bytes_indicated < size) {
2012             VMW_PKPRN("RX: %zu of %zu bytes indicated", bytes_indicated, size);
2013         }
2014     } else {
2015         VMW_PKPRN("Packet dropped by RX filter");
2016         bytes_indicated = size;
2017     }
2018 
2019     assert(size > 0);
2020     assert(bytes_indicated != 0);
2021     return bytes_indicated;
2022 }
2023 
2024 static void vmxnet3_set_link_status(NetClientState *nc)
2025 {
2026     VMXNET3State *s = qemu_get_nic_opaque(nc);
2027 
2028     if (nc->link_down) {
2029         s->link_status_and_speed &= ~VMXNET3_LINK_STATUS_UP;
2030     } else {
2031         s->link_status_and_speed |= VMXNET3_LINK_STATUS_UP;
2032     }
2033 
2034     vmxnet3_set_events(s, VMXNET3_ECR_LINK);
2035     vmxnet3_trigger_interrupt(s, s->event_int_idx);
2036 }
2037 
2038 static NetClientInfo net_vmxnet3_info = {
2039         .type = NET_CLIENT_DRIVER_NIC,
2040         .size = sizeof(NICState),
2041         .receive = vmxnet3_receive,
2042         .link_status_changed = vmxnet3_set_link_status,
2043 };
2044 
2045 static bool vmxnet3_peer_has_vnet_hdr(VMXNET3State *s)
2046 {
2047     NetClientState *nc = qemu_get_queue(s->nic);
2048 
2049     if (qemu_has_vnet_hdr(nc->peer)) {
2050         return true;
2051     }
2052 
2053     return false;
2054 }
2055 
2056 static void vmxnet3_net_uninit(VMXNET3State *s)
2057 {
2058     g_free(s->mcast_list);
2059     vmxnet3_deactivate_device(s);
2060     qemu_del_nic(s->nic);
2061 }
2062 
2063 static void vmxnet3_net_init(VMXNET3State *s)
2064 {
2065     DeviceState *d = DEVICE(s);
2066 
2067     VMW_CBPRN("vmxnet3_net_init called...");
2068 
2069     qemu_macaddr_default_if_unset(&s->conf.macaddr);
2070 
2071     /* Windows guest will query the address that was set on init */
2072     memcpy(&s->perm_mac.a, &s->conf.macaddr.a, sizeof(s->perm_mac.a));
2073 
2074     s->mcast_list = NULL;
2075     s->mcast_list_len = 0;
2076 
2077     s->link_status_and_speed = VMXNET3_LINK_SPEED | VMXNET3_LINK_STATUS_UP;
2078 
2079     VMW_CFPRN("Permanent MAC: " MAC_FMT, MAC_ARG(s->perm_mac.a));
2080 
2081     s->nic = qemu_new_nic(&net_vmxnet3_info, &s->conf,
2082                           object_get_typename(OBJECT(s)),
2083                           d->id, s);
2084 
2085     s->peer_has_vhdr = vmxnet3_peer_has_vnet_hdr(s);
2086     s->tx_sop = true;
2087     s->skip_current_tx_pkt = false;
2088     s->tx_pkt = NULL;
2089     s->rx_pkt = NULL;
2090     s->rx_vlan_stripping = false;
2091     s->lro_supported = false;
2092 
2093     if (s->peer_has_vhdr) {
2094         qemu_set_vnet_hdr_len(qemu_get_queue(s->nic)->peer,
2095             sizeof(struct virtio_net_hdr));
2096 
2097         qemu_using_vnet_hdr(qemu_get_queue(s->nic)->peer, 1);
2098     }
2099 
2100     qemu_format_nic_info_str(qemu_get_queue(s->nic), s->conf.macaddr.a);
2101 }
2102 
2103 static void
2104 vmxnet3_unuse_msix_vectors(VMXNET3State *s, int num_vectors)
2105 {
2106     PCIDevice *d = PCI_DEVICE(s);
2107     int i;
2108     for (i = 0; i < num_vectors; i++) {
2109         msix_vector_unuse(d, i);
2110     }
2111 }
2112 
2113 static bool
2114 vmxnet3_use_msix_vectors(VMXNET3State *s, int num_vectors)
2115 {
2116     PCIDevice *d = PCI_DEVICE(s);
2117     int i;
2118     for (i = 0; i < num_vectors; i++) {
2119         int res = msix_vector_use(d, i);
2120         if (0 > res) {
2121             VMW_WRPRN("Failed to use MSI-X vector %d, error %d", i, res);
2122             vmxnet3_unuse_msix_vectors(s, i);
2123             return false;
2124         }
2125     }
2126     return true;
2127 }
2128 
2129 static bool
2130 vmxnet3_init_msix(VMXNET3State *s)
2131 {
2132     PCIDevice *d = PCI_DEVICE(s);
2133     int res = msix_init(d, VMXNET3_MAX_INTRS,
2134                         &s->msix_bar,
2135                         VMXNET3_MSIX_BAR_IDX, VMXNET3_OFF_MSIX_TABLE,
2136                         &s->msix_bar,
2137                         VMXNET3_MSIX_BAR_IDX, VMXNET3_OFF_MSIX_PBA(s),
2138                         VMXNET3_MSIX_OFFSET(s), NULL);
2139 
2140     if (0 > res) {
2141         VMW_WRPRN("Failed to initialize MSI-X, error %d", res);
2142         s->msix_used = false;
2143     } else {
2144         if (!vmxnet3_use_msix_vectors(s, VMXNET3_MAX_INTRS)) {
2145             VMW_WRPRN("Failed to use MSI-X vectors, error %d", res);
2146             msix_uninit(d, &s->msix_bar, &s->msix_bar);
2147             s->msix_used = false;
2148         } else {
2149             s->msix_used = true;
2150         }
2151     }
2152     return s->msix_used;
2153 }
2154 
2155 static void
2156 vmxnet3_cleanup_msix(VMXNET3State *s)
2157 {
2158     PCIDevice *d = PCI_DEVICE(s);
2159 
2160     if (s->msix_used) {
2161         vmxnet3_unuse_msix_vectors(s, VMXNET3_MAX_INTRS);
2162         msix_uninit(d, &s->msix_bar, &s->msix_bar);
2163     }
2164 }
2165 
2166 static void
2167 vmxnet3_cleanup_msi(VMXNET3State *s)
2168 {
2169     PCIDevice *d = PCI_DEVICE(s);
2170 
2171     msi_uninit(d);
2172 }
2173 
2174 static const MemoryRegionOps b0_ops = {
2175     .read = vmxnet3_io_bar0_read,
2176     .write = vmxnet3_io_bar0_write,
2177     .endianness = DEVICE_LITTLE_ENDIAN,
2178     .impl = {
2179             .min_access_size = 4,
2180             .max_access_size = 4,
2181     },
2182 };
2183 
2184 static const MemoryRegionOps b1_ops = {
2185     .read = vmxnet3_io_bar1_read,
2186     .write = vmxnet3_io_bar1_write,
2187     .endianness = DEVICE_LITTLE_ENDIAN,
2188     .impl = {
2189             .min_access_size = 4,
2190             .max_access_size = 4,
2191     },
2192 };
2193 
2194 static uint64_t vmxnet3_device_serial_num(VMXNET3State *s)
2195 {
2196     uint64_t dsn_payload;
2197     uint8_t *dsnp = (uint8_t *)&dsn_payload;
2198 
2199     dsnp[0] = 0xfe;
2200     dsnp[1] = s->conf.macaddr.a[3];
2201     dsnp[2] = s->conf.macaddr.a[4];
2202     dsnp[3] = s->conf.macaddr.a[5];
2203     dsnp[4] = s->conf.macaddr.a[0];
2204     dsnp[5] = s->conf.macaddr.a[1];
2205     dsnp[6] = s->conf.macaddr.a[2];
2206     dsnp[7] = 0xff;
2207     return dsn_payload;
2208 }
2209 
2210 
2211 #define VMXNET3_USE_64BIT         (true)
2212 #define VMXNET3_PER_VECTOR_MASK   (false)
2213 
2214 static void vmxnet3_pci_realize(PCIDevice *pci_dev, Error **errp)
2215 {
2216     VMXNET3State *s = VMXNET3(pci_dev);
2217     int ret;
2218 
2219     VMW_CBPRN("Starting init...");
2220 
2221     memory_region_init_io(&s->bar0, OBJECT(s), &b0_ops, s,
2222                           "vmxnet3-b0", VMXNET3_PT_REG_SIZE);
2223     pci_register_bar(pci_dev, VMXNET3_BAR0_IDX,
2224                      PCI_BASE_ADDRESS_SPACE_MEMORY, &s->bar0);
2225 
2226     memory_region_init_io(&s->bar1, OBJECT(s), &b1_ops, s,
2227                           "vmxnet3-b1", VMXNET3_VD_REG_SIZE);
2228     pci_register_bar(pci_dev, VMXNET3_BAR1_IDX,
2229                      PCI_BASE_ADDRESS_SPACE_MEMORY, &s->bar1);
2230 
2231     memory_region_init(&s->msix_bar, OBJECT(s), "vmxnet3-msix-bar",
2232                        VMXNET3_MSIX_BAR_SIZE);
2233     pci_register_bar(pci_dev, VMXNET3_MSIX_BAR_IDX,
2234                      PCI_BASE_ADDRESS_SPACE_MEMORY, &s->msix_bar);
2235 
2236     vmxnet3_reset_interrupt_states(s);
2237 
2238     /* Interrupt pin A */
2239     pci_dev->config[PCI_INTERRUPT_PIN] = 0x01;
2240 
2241     ret = msi_init(pci_dev, VMXNET3_MSI_OFFSET(s), VMXNET3_MAX_NMSIX_INTRS,
2242                    VMXNET3_USE_64BIT, VMXNET3_PER_VECTOR_MASK, NULL);
2243     /* Any error other than -ENOTSUP(board's MSI support is broken)
2244      * is a programming error. Fall back to INTx silently on -ENOTSUP */
2245     assert(!ret || ret == -ENOTSUP);
2246 
2247     if (!vmxnet3_init_msix(s)) {
2248         VMW_WRPRN("Failed to initialize MSI-X, configuration is inconsistent.");
2249     }
2250 
2251     vmxnet3_net_init(s);
2252 
2253     if (pci_is_express(pci_dev)) {
2254         if (pci_bus_is_express(pci_get_bus(pci_dev))) {
2255             pcie_endpoint_cap_init(pci_dev, VMXNET3_EXP_EP_OFFSET);
2256         }
2257 
2258         pcie_dev_ser_num_init(pci_dev, VMXNET3_DSN_OFFSET,
2259                               vmxnet3_device_serial_num(s));
2260     }
2261 }
2262 
2263 static void vmxnet3_instance_init(Object *obj)
2264 {
2265     VMXNET3State *s = VMXNET3(obj);
2266     device_add_bootindex_property(obj, &s->conf.bootindex,
2267                                   "bootindex", "/ethernet-phy@0",
2268                                   DEVICE(obj));
2269 }
2270 
2271 static void vmxnet3_pci_uninit(PCIDevice *pci_dev)
2272 {
2273     VMXNET3State *s = VMXNET3(pci_dev);
2274 
2275     VMW_CBPRN("Starting uninit...");
2276 
2277     vmxnet3_net_uninit(s);
2278 
2279     vmxnet3_cleanup_msix(s);
2280 
2281     vmxnet3_cleanup_msi(s);
2282 }
2283 
2284 static void vmxnet3_qdev_reset(DeviceState *dev)
2285 {
2286     PCIDevice *d = PCI_DEVICE(dev);
2287     VMXNET3State *s = VMXNET3(d);
2288 
2289     VMW_CBPRN("Starting QDEV reset...");
2290     vmxnet3_reset(s);
2291 }
2292 
2293 static bool vmxnet3_mc_list_needed(void *opaque)
2294 {
2295     return true;
2296 }
2297 
2298 static int vmxnet3_mcast_list_pre_load(void *opaque)
2299 {
2300     VMXNET3State *s = opaque;
2301 
2302     s->mcast_list = g_malloc(s->mcast_list_buff_size);
2303 
2304     return 0;
2305 }
2306 
2307 
2308 static int vmxnet3_pre_save(void *opaque)
2309 {
2310     VMXNET3State *s = opaque;
2311 
2312     s->mcast_list_buff_size = s->mcast_list_len * sizeof(MACAddr);
2313 
2314     return 0;
2315 }
2316 
2317 static const VMStateDescription vmxstate_vmxnet3_mcast_list = {
2318     .name = "vmxnet3/mcast_list",
2319     .version_id = 1,
2320     .minimum_version_id = 1,
2321     .pre_load = vmxnet3_mcast_list_pre_load,
2322     .needed = vmxnet3_mc_list_needed,
2323     .fields = (VMStateField[]) {
2324         VMSTATE_VBUFFER_UINT32(mcast_list, VMXNET3State, 0, NULL,
2325             mcast_list_buff_size),
2326         VMSTATE_END_OF_LIST()
2327     }
2328 };
2329 
2330 static const VMStateDescription vmstate_vmxnet3_ring = {
2331     .name = "vmxnet3-ring",
2332     .version_id = 0,
2333     .fields = (VMStateField[]) {
2334         VMSTATE_UINT64(pa, Vmxnet3Ring),
2335         VMSTATE_UINT32(size, Vmxnet3Ring),
2336         VMSTATE_UINT32(cell_size, Vmxnet3Ring),
2337         VMSTATE_UINT32(next, Vmxnet3Ring),
2338         VMSTATE_UINT8(gen, Vmxnet3Ring),
2339         VMSTATE_END_OF_LIST()
2340     }
2341 };
2342 
2343 static const VMStateDescription vmstate_vmxnet3_tx_stats = {
2344     .name = "vmxnet3-tx-stats",
2345     .version_id = 0,
2346     .fields = (VMStateField[]) {
2347         VMSTATE_UINT64(TSOPktsTxOK, struct UPT1_TxStats),
2348         VMSTATE_UINT64(TSOBytesTxOK, struct UPT1_TxStats),
2349         VMSTATE_UINT64(ucastPktsTxOK, struct UPT1_TxStats),
2350         VMSTATE_UINT64(ucastBytesTxOK, struct UPT1_TxStats),
2351         VMSTATE_UINT64(mcastPktsTxOK, struct UPT1_TxStats),
2352         VMSTATE_UINT64(mcastBytesTxOK, struct UPT1_TxStats),
2353         VMSTATE_UINT64(bcastPktsTxOK, struct UPT1_TxStats),
2354         VMSTATE_UINT64(bcastBytesTxOK, struct UPT1_TxStats),
2355         VMSTATE_UINT64(pktsTxError, struct UPT1_TxStats),
2356         VMSTATE_UINT64(pktsTxDiscard, struct UPT1_TxStats),
2357         VMSTATE_END_OF_LIST()
2358     }
2359 };
2360 
2361 static const VMStateDescription vmstate_vmxnet3_txq_descr = {
2362     .name = "vmxnet3-txq-descr",
2363     .version_id = 0,
2364     .fields = (VMStateField[]) {
2365         VMSTATE_STRUCT(tx_ring, Vmxnet3TxqDescr, 0, vmstate_vmxnet3_ring,
2366                        Vmxnet3Ring),
2367         VMSTATE_STRUCT(comp_ring, Vmxnet3TxqDescr, 0, vmstate_vmxnet3_ring,
2368                        Vmxnet3Ring),
2369         VMSTATE_UINT8(intr_idx, Vmxnet3TxqDescr),
2370         VMSTATE_UINT64(tx_stats_pa, Vmxnet3TxqDescr),
2371         VMSTATE_STRUCT(txq_stats, Vmxnet3TxqDescr, 0, vmstate_vmxnet3_tx_stats,
2372                        struct UPT1_TxStats),
2373         VMSTATE_END_OF_LIST()
2374     }
2375 };
2376 
2377 static const VMStateDescription vmstate_vmxnet3_rx_stats = {
2378     .name = "vmxnet3-rx-stats",
2379     .version_id = 0,
2380     .fields = (VMStateField[]) {
2381         VMSTATE_UINT64(LROPktsRxOK, struct UPT1_RxStats),
2382         VMSTATE_UINT64(LROBytesRxOK, struct UPT1_RxStats),
2383         VMSTATE_UINT64(ucastPktsRxOK, struct UPT1_RxStats),
2384         VMSTATE_UINT64(ucastBytesRxOK, struct UPT1_RxStats),
2385         VMSTATE_UINT64(mcastPktsRxOK, struct UPT1_RxStats),
2386         VMSTATE_UINT64(mcastBytesRxOK, struct UPT1_RxStats),
2387         VMSTATE_UINT64(bcastPktsRxOK, struct UPT1_RxStats),
2388         VMSTATE_UINT64(bcastBytesRxOK, struct UPT1_RxStats),
2389         VMSTATE_UINT64(pktsRxOutOfBuf, struct UPT1_RxStats),
2390         VMSTATE_UINT64(pktsRxError, struct UPT1_RxStats),
2391         VMSTATE_END_OF_LIST()
2392     }
2393 };
2394 
2395 static const VMStateDescription vmstate_vmxnet3_rxq_descr = {
2396     .name = "vmxnet3-rxq-descr",
2397     .version_id = 0,
2398     .fields = (VMStateField[]) {
2399         VMSTATE_STRUCT_ARRAY(rx_ring, Vmxnet3RxqDescr,
2400                              VMXNET3_RX_RINGS_PER_QUEUE, 0,
2401                              vmstate_vmxnet3_ring, Vmxnet3Ring),
2402         VMSTATE_STRUCT(comp_ring, Vmxnet3RxqDescr, 0, vmstate_vmxnet3_ring,
2403                        Vmxnet3Ring),
2404         VMSTATE_UINT8(intr_idx, Vmxnet3RxqDescr),
2405         VMSTATE_UINT64(rx_stats_pa, Vmxnet3RxqDescr),
2406         VMSTATE_STRUCT(rxq_stats, Vmxnet3RxqDescr, 0, vmstate_vmxnet3_rx_stats,
2407                        struct UPT1_RxStats),
2408         VMSTATE_END_OF_LIST()
2409     }
2410 };
2411 
2412 static int vmxnet3_post_load(void *opaque, int version_id)
2413 {
2414     VMXNET3State *s = opaque;
2415     PCIDevice *d = PCI_DEVICE(s);
2416 
2417     net_tx_pkt_init(&s->tx_pkt, PCI_DEVICE(s),
2418                     s->max_tx_frags, s->peer_has_vhdr);
2419     net_rx_pkt_init(&s->rx_pkt, s->peer_has_vhdr);
2420 
2421     if (s->msix_used) {
2422         if  (!vmxnet3_use_msix_vectors(s, VMXNET3_MAX_INTRS)) {
2423             VMW_WRPRN("Failed to re-use MSI-X vectors");
2424             msix_uninit(d, &s->msix_bar, &s->msix_bar);
2425             s->msix_used = false;
2426             return -1;
2427         }
2428     }
2429 
2430     if (!vmxnet3_validate_queues(s)) {
2431         return -1;
2432     }
2433     vmxnet3_validate_interrupts(s);
2434 
2435     return 0;
2436 }
2437 
2438 static const VMStateDescription vmstate_vmxnet3_int_state = {
2439     .name = "vmxnet3-int-state",
2440     .version_id = 0,
2441     .fields = (VMStateField[]) {
2442         VMSTATE_BOOL(is_masked, Vmxnet3IntState),
2443         VMSTATE_BOOL(is_pending, Vmxnet3IntState),
2444         VMSTATE_BOOL(is_asserted, Vmxnet3IntState),
2445         VMSTATE_END_OF_LIST()
2446     }
2447 };
2448 
2449 static const VMStateDescription vmstate_vmxnet3 = {
2450     .name = "vmxnet3",
2451     .version_id = 1,
2452     .minimum_version_id = 1,
2453     .pre_save = vmxnet3_pre_save,
2454     .post_load = vmxnet3_post_load,
2455     .fields = (VMStateField[]) {
2456             VMSTATE_PCI_DEVICE(parent_obj, VMXNET3State),
2457             VMSTATE_MSIX(parent_obj, VMXNET3State),
2458             VMSTATE_BOOL(rx_packets_compound, VMXNET3State),
2459             VMSTATE_BOOL(rx_vlan_stripping, VMXNET3State),
2460             VMSTATE_BOOL(lro_supported, VMXNET3State),
2461             VMSTATE_UINT32(rx_mode, VMXNET3State),
2462             VMSTATE_UINT32(mcast_list_len, VMXNET3State),
2463             VMSTATE_UINT32(mcast_list_buff_size, VMXNET3State),
2464             VMSTATE_UINT32_ARRAY(vlan_table, VMXNET3State, VMXNET3_VFT_SIZE),
2465             VMSTATE_UINT32(mtu, VMXNET3State),
2466             VMSTATE_UINT16(max_rx_frags, VMXNET3State),
2467             VMSTATE_UINT32(max_tx_frags, VMXNET3State),
2468             VMSTATE_UINT8(event_int_idx, VMXNET3State),
2469             VMSTATE_BOOL(auto_int_masking, VMXNET3State),
2470             VMSTATE_UINT8(txq_num, VMXNET3State),
2471             VMSTATE_UINT8(rxq_num, VMXNET3State),
2472             VMSTATE_UINT32(device_active, VMXNET3State),
2473             VMSTATE_UINT32(last_command, VMXNET3State),
2474             VMSTATE_UINT32(link_status_and_speed, VMXNET3State),
2475             VMSTATE_UINT32(temp_mac, VMXNET3State),
2476             VMSTATE_UINT64(drv_shmem, VMXNET3State),
2477             VMSTATE_UINT64(temp_shared_guest_driver_memory, VMXNET3State),
2478 
2479             VMSTATE_STRUCT_ARRAY(txq_descr, VMXNET3State,
2480                 VMXNET3_DEVICE_MAX_TX_QUEUES, 0, vmstate_vmxnet3_txq_descr,
2481                 Vmxnet3TxqDescr),
2482             VMSTATE_STRUCT_ARRAY(rxq_descr, VMXNET3State,
2483                 VMXNET3_DEVICE_MAX_RX_QUEUES, 0, vmstate_vmxnet3_rxq_descr,
2484                 Vmxnet3RxqDescr),
2485             VMSTATE_STRUCT_ARRAY(interrupt_states, VMXNET3State,
2486                 VMXNET3_MAX_INTRS, 0, vmstate_vmxnet3_int_state,
2487                 Vmxnet3IntState),
2488 
2489             VMSTATE_END_OF_LIST()
2490     },
2491     .subsections = (const VMStateDescription*[]) {
2492         &vmxstate_vmxnet3_mcast_list,
2493         NULL
2494     }
2495 };
2496 
2497 static Property vmxnet3_properties[] = {
2498     DEFINE_NIC_PROPERTIES(VMXNET3State, conf),
2499     DEFINE_PROP_BIT("x-old-msi-offsets", VMXNET3State, compat_flags,
2500                     VMXNET3_COMPAT_FLAG_OLD_MSI_OFFSETS_BIT, false),
2501     DEFINE_PROP_BIT("x-disable-pcie", VMXNET3State, compat_flags,
2502                     VMXNET3_COMPAT_FLAG_DISABLE_PCIE_BIT, false),
2503     DEFINE_PROP_END_OF_LIST(),
2504 };
2505 
2506 static void vmxnet3_realize(DeviceState *qdev, Error **errp)
2507 {
2508     VMXNET3Class *vc = VMXNET3_DEVICE_GET_CLASS(qdev);
2509     PCIDevice *pci_dev = PCI_DEVICE(qdev);
2510     VMXNET3State *s = VMXNET3(qdev);
2511 
2512     if (!(s->compat_flags & VMXNET3_COMPAT_FLAG_DISABLE_PCIE)) {
2513         pci_dev->cap_present |= QEMU_PCI_CAP_EXPRESS;
2514     }
2515 
2516     vc->parent_dc_realize(qdev, errp);
2517 }
2518 
2519 static void vmxnet3_class_init(ObjectClass *class, void *data)
2520 {
2521     DeviceClass *dc = DEVICE_CLASS(class);
2522     PCIDeviceClass *c = PCI_DEVICE_CLASS(class);
2523     VMXNET3Class *vc = VMXNET3_DEVICE_CLASS(class);
2524 
2525     c->realize = vmxnet3_pci_realize;
2526     c->exit = vmxnet3_pci_uninit;
2527     c->vendor_id = PCI_VENDOR_ID_VMWARE;
2528     c->device_id = PCI_DEVICE_ID_VMWARE_VMXNET3;
2529     c->revision = PCI_DEVICE_ID_VMWARE_VMXNET3_REVISION;
2530     c->romfile = "efi-vmxnet3.rom";
2531     c->class_id = PCI_CLASS_NETWORK_ETHERNET;
2532     c->subsystem_vendor_id = PCI_VENDOR_ID_VMWARE;
2533     c->subsystem_id = PCI_DEVICE_ID_VMWARE_VMXNET3;
2534     device_class_set_parent_realize(dc, vmxnet3_realize,
2535                                     &vc->parent_dc_realize);
2536     dc->desc = "VMWare Paravirtualized Ethernet v3";
2537     dc->reset = vmxnet3_qdev_reset;
2538     dc->vmsd = &vmstate_vmxnet3;
2539     device_class_set_props(dc, vmxnet3_properties);
2540     set_bit(DEVICE_CATEGORY_NETWORK, dc->categories);
2541 }
2542 
2543 static const TypeInfo vmxnet3_info = {
2544     .name          = TYPE_VMXNET3,
2545     .parent        = TYPE_PCI_DEVICE,
2546     .class_size    = sizeof(VMXNET3Class),
2547     .instance_size = sizeof(VMXNET3State),
2548     .class_init    = vmxnet3_class_init,
2549     .instance_init = vmxnet3_instance_init,
2550     .interfaces = (InterfaceInfo[]) {
2551         { INTERFACE_PCIE_DEVICE },
2552         { INTERFACE_CONVENTIONAL_PCI_DEVICE },
2553         { }
2554     },
2555 };
2556 
2557 static void vmxnet3_register_types(void)
2558 {
2559     VMW_CBPRN("vmxnet3_register_types called...");
2560     type_register_static(&vmxnet3_info);
2561 }
2562 
2563 type_init(vmxnet3_register_types)
2564