xref: /openbmc/qemu/hw/net/vmxnet3.c (revision 651ccdfa)
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         return net_tx_pkt_build_vheader(s->tx_pkt, false, false, 0);
444 
445     case VMXNET3_OM_CSUM:
446         VMW_PKPRN("L4 CSO requested\n");
447         return net_tx_pkt_build_vheader(s->tx_pkt, false, true, 0);
448 
449     case VMXNET3_OM_TSO:
450         VMW_PKPRN("GSO offload requested.");
451         if (!net_tx_pkt_build_vheader(s->tx_pkt, true, true,
452             s->cso_or_gso_size)) {
453             return false;
454         }
455         net_tx_pkt_update_ip_checksums(s->tx_pkt);
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 hasip4, hasip6;
851     EthL4HdrProto l4hdr_proto;
852     uint8_t *data;
853     int len;
854 
855     vhdr = net_rx_pkt_get_vhdr(pkt);
856     if (!VMXNET_FLAG_IS_SET(vhdr->flags, VIRTIO_NET_HDR_F_NEEDS_CSUM)) {
857         return;
858     }
859 
860     net_rx_pkt_get_protocols(pkt, &hasip4, &hasip6, &l4hdr_proto);
861     if (!(hasip4 || hasip6) ||
862         (l4hdr_proto != ETH_L4_HDR_PROTO_TCP &&
863          l4hdr_proto != ETH_L4_HDR_PROTO_UDP)) {
864         return;
865     }
866 
867     vmxnet3_dump_virt_hdr(vhdr);
868 
869     /* Validate packet len: csum_start + scum_offset + length of csum field */
870     if (pkt_len < (vhdr->csum_start + vhdr->csum_offset + 2)) {
871         VMW_PKPRN("packet len:%zu < csum_start(%d) + csum_offset(%d) + 2, "
872                   "cannot calculate checksum",
873                   pkt_len, vhdr->csum_start, vhdr->csum_offset);
874         return;
875     }
876 
877     data = (uint8_t *)pkt_data + vhdr->csum_start;
878     len = pkt_len - vhdr->csum_start;
879     /* Put the checksum obtained into the packet */
880     stw_be_p(data + vhdr->csum_offset,
881              net_checksum_finish_nozero(net_checksum_add(len, data)));
882 
883     vhdr->flags &= ~VIRTIO_NET_HDR_F_NEEDS_CSUM;
884     vhdr->flags |= VIRTIO_NET_HDR_F_DATA_VALID;
885 }
886 
887 static void vmxnet3_rx_update_descr(struct NetRxPkt *pkt,
888     struct Vmxnet3_RxCompDesc *rxcd)
889 {
890     int csum_ok, is_gso;
891     bool hasip4, hasip6;
892     EthL4HdrProto l4hdr_proto;
893     struct virtio_net_hdr *vhdr;
894     uint8_t offload_type;
895 
896     if (net_rx_pkt_is_vlan_stripped(pkt)) {
897         rxcd->ts = 1;
898         rxcd->tci = net_rx_pkt_get_vlan_tag(pkt);
899     }
900 
901     vhdr = net_rx_pkt_get_vhdr(pkt);
902     /*
903      * Checksum is valid when lower level tell so or when lower level
904      * requires checksum offload telling that packet produced/bridged
905      * locally and did travel over network after last checksum calculation
906      * or production
907      */
908     csum_ok = VMXNET_FLAG_IS_SET(vhdr->flags, VIRTIO_NET_HDR_F_DATA_VALID) ||
909               VMXNET_FLAG_IS_SET(vhdr->flags, VIRTIO_NET_HDR_F_NEEDS_CSUM);
910 
911     offload_type = vhdr->gso_type & ~VIRTIO_NET_HDR_GSO_ECN;
912     is_gso = (offload_type != VIRTIO_NET_HDR_GSO_NONE) ? 1 : 0;
913 
914     if (!csum_ok && !is_gso) {
915         goto nocsum;
916     }
917 
918     net_rx_pkt_get_protocols(pkt, &hasip4, &hasip6, &l4hdr_proto);
919     if ((l4hdr_proto != ETH_L4_HDR_PROTO_TCP &&
920          l4hdr_proto != ETH_L4_HDR_PROTO_UDP) ||
921         (!hasip4 && !hasip6)) {
922         goto nocsum;
923     }
924 
925     rxcd->cnc = 0;
926     rxcd->v4 = hasip4 ? 1 : 0;
927     rxcd->v6 = hasip6 ? 1 : 0;
928     rxcd->tcp = l4hdr_proto == ETH_L4_HDR_PROTO_TCP;
929     rxcd->udp = l4hdr_proto == ETH_L4_HDR_PROTO_UDP;
930     rxcd->fcs = rxcd->tuc = rxcd->ipc = 1;
931     return;
932 
933 nocsum:
934     rxcd->cnc = 1;
935     return;
936 }
937 
938 static void
939 vmxnet3_pci_dma_writev(PCIDevice *pci_dev,
940                        const struct iovec *iov,
941                        size_t start_iov_off,
942                        hwaddr target_addr,
943                        size_t bytes_to_copy)
944 {
945     size_t curr_off = 0;
946     size_t copied = 0;
947 
948     while (bytes_to_copy) {
949         if (start_iov_off < (curr_off + iov->iov_len)) {
950             size_t chunk_len =
951                 MIN((curr_off + iov->iov_len) - start_iov_off, bytes_to_copy);
952 
953             pci_dma_write(pci_dev, target_addr + copied,
954                           iov->iov_base + start_iov_off - curr_off,
955                           chunk_len);
956 
957             copied += chunk_len;
958             start_iov_off += chunk_len;
959             curr_off = start_iov_off;
960             bytes_to_copy -= chunk_len;
961         } else {
962             curr_off += iov->iov_len;
963         }
964         iov++;
965     }
966 }
967 
968 static void
969 vmxnet3_pci_dma_write_rxcd(PCIDevice *pcidev, dma_addr_t pa,
970                            struct Vmxnet3_RxCompDesc *rxcd)
971 {
972     rxcd->val1 = cpu_to_le32(rxcd->val1);
973     rxcd->val2 = cpu_to_le32(rxcd->val2);
974     rxcd->val3 = cpu_to_le32(rxcd->val3);
975     pci_dma_write(pcidev, pa, rxcd, sizeof(*rxcd));
976 }
977 
978 static bool
979 vmxnet3_indicate_packet(VMXNET3State *s)
980 {
981     struct Vmxnet3_RxDesc rxd;
982     PCIDevice *d = PCI_DEVICE(s);
983     bool is_head = true;
984     uint32_t rxd_idx;
985     uint32_t rx_ridx = 0;
986 
987     struct Vmxnet3_RxCompDesc rxcd;
988     uint32_t new_rxcd_gen = VMXNET3_INIT_GEN;
989     hwaddr new_rxcd_pa = 0;
990     hwaddr ready_rxcd_pa = 0;
991     struct iovec *data = net_rx_pkt_get_iovec(s->rx_pkt);
992     size_t bytes_copied = 0;
993     size_t bytes_left = net_rx_pkt_get_total_len(s->rx_pkt);
994     uint16_t num_frags = 0;
995     size_t chunk_size;
996 
997     net_rx_pkt_dump(s->rx_pkt);
998 
999     while (bytes_left > 0) {
1000 
1001         /* cannot add more frags to packet */
1002         if (num_frags == s->max_rx_frags) {
1003             break;
1004         }
1005 
1006         new_rxcd_pa = vmxnet3_pop_rxc_descr(s, RXQ_IDX, &new_rxcd_gen);
1007         if (!new_rxcd_pa) {
1008             break;
1009         }
1010 
1011         if (!vmxnet3_get_next_rx_descr(s, is_head, &rxd, &rxd_idx, &rx_ridx)) {
1012             break;
1013         }
1014 
1015         chunk_size = MIN(bytes_left, rxd.len);
1016         vmxnet3_pci_dma_writev(d, data, bytes_copied, rxd.addr, chunk_size);
1017         bytes_copied += chunk_size;
1018         bytes_left -= chunk_size;
1019 
1020         vmxnet3_dump_rx_descr(&rxd);
1021 
1022         if (ready_rxcd_pa != 0) {
1023             vmxnet3_pci_dma_write_rxcd(d, ready_rxcd_pa, &rxcd);
1024         }
1025 
1026         memset(&rxcd, 0, sizeof(struct Vmxnet3_RxCompDesc));
1027         rxcd.rxdIdx = rxd_idx;
1028         rxcd.len = chunk_size;
1029         rxcd.sop = is_head;
1030         rxcd.gen = new_rxcd_gen;
1031         rxcd.rqID = RXQ_IDX + rx_ridx * s->rxq_num;
1032 
1033         if (bytes_left == 0) {
1034             vmxnet3_rx_update_descr(s->rx_pkt, &rxcd);
1035         }
1036 
1037         VMW_RIPRN("RX Completion descriptor: rxRing: %lu rxIdx %lu len %lu "
1038                   "sop %d csum_correct %lu",
1039                   (unsigned long) rx_ridx,
1040                   (unsigned long) rxcd.rxdIdx,
1041                   (unsigned long) rxcd.len,
1042                   (int) rxcd.sop,
1043                   (unsigned long) rxcd.tuc);
1044 
1045         is_head = false;
1046         ready_rxcd_pa = new_rxcd_pa;
1047         new_rxcd_pa = 0;
1048         num_frags++;
1049     }
1050 
1051     if (ready_rxcd_pa != 0) {
1052         rxcd.eop = 1;
1053         rxcd.err = (bytes_left != 0);
1054 
1055         vmxnet3_pci_dma_write_rxcd(d, ready_rxcd_pa, &rxcd);
1056 
1057         /* Flush RX descriptor changes */
1058         smp_wmb();
1059     }
1060 
1061     if (new_rxcd_pa != 0) {
1062         vmxnet3_revert_rxc_descr(s, RXQ_IDX);
1063     }
1064 
1065     vmxnet3_trigger_interrupt(s, s->rxq_descr[RXQ_IDX].intr_idx);
1066 
1067     if (bytes_left == 0) {
1068         vmxnet3_on_rx_done_update_stats(s, RXQ_IDX, VMXNET3_PKT_STATUS_OK);
1069         return true;
1070     } else if (num_frags == s->max_rx_frags) {
1071         vmxnet3_on_rx_done_update_stats(s, RXQ_IDX, VMXNET3_PKT_STATUS_ERROR);
1072         return false;
1073     } else {
1074         vmxnet3_on_rx_done_update_stats(s, RXQ_IDX,
1075                                         VMXNET3_PKT_STATUS_OUT_OF_BUF);
1076         return false;
1077     }
1078 }
1079 
1080 static void
1081 vmxnet3_io_bar0_write(void *opaque, hwaddr addr,
1082                       uint64_t val, unsigned size)
1083 {
1084     VMXNET3State *s = opaque;
1085 
1086     if (!s->device_active) {
1087         return;
1088     }
1089 
1090     if (VMW_IS_MULTIREG_ADDR(addr, VMXNET3_REG_TXPROD,
1091                         VMXNET3_DEVICE_MAX_TX_QUEUES, VMXNET3_REG_ALIGN)) {
1092         int tx_queue_idx =
1093             VMW_MULTIREG_IDX_BY_ADDR(addr, VMXNET3_REG_TXPROD,
1094                                      VMXNET3_REG_ALIGN);
1095         if (tx_queue_idx <= s->txq_num) {
1096             vmxnet3_process_tx_queue(s, tx_queue_idx);
1097         } else {
1098             qemu_log_mask(LOG_GUEST_ERROR, "vmxnet3: Illegal TX queue %d/%d\n",
1099                           tx_queue_idx, s->txq_num);
1100         }
1101         return;
1102     }
1103 
1104     if (VMW_IS_MULTIREG_ADDR(addr, VMXNET3_REG_IMR,
1105                         VMXNET3_MAX_INTRS, VMXNET3_REG_ALIGN)) {
1106         int l = VMW_MULTIREG_IDX_BY_ADDR(addr, VMXNET3_REG_IMR,
1107                                          VMXNET3_REG_ALIGN);
1108 
1109         VMW_CBPRN("Interrupt mask for line %d written: 0x%" PRIx64, l, val);
1110 
1111         vmxnet3_on_interrupt_mask_changed(s, l, val);
1112         return;
1113     }
1114 
1115     if (VMW_IS_MULTIREG_ADDR(addr, VMXNET3_REG_RXPROD,
1116                         VMXNET3_DEVICE_MAX_RX_QUEUES, VMXNET3_REG_ALIGN) ||
1117        VMW_IS_MULTIREG_ADDR(addr, VMXNET3_REG_RXPROD2,
1118                         VMXNET3_DEVICE_MAX_RX_QUEUES, VMXNET3_REG_ALIGN)) {
1119         return;
1120     }
1121 
1122     VMW_WRPRN("BAR0 unknown write [%" PRIx64 "] = %" PRIx64 ", size %d",
1123               (uint64_t) addr, val, size);
1124 }
1125 
1126 static uint64_t
1127 vmxnet3_io_bar0_read(void *opaque, hwaddr addr, unsigned size)
1128 {
1129     VMXNET3State *s = opaque;
1130 
1131     if (VMW_IS_MULTIREG_ADDR(addr, VMXNET3_REG_IMR,
1132                         VMXNET3_MAX_INTRS, VMXNET3_REG_ALIGN)) {
1133         int l = VMW_MULTIREG_IDX_BY_ADDR(addr, VMXNET3_REG_IMR,
1134                                          VMXNET3_REG_ALIGN);
1135         return s->interrupt_states[l].is_masked;
1136     }
1137 
1138     VMW_CBPRN("BAR0 unknown read [%" PRIx64 "], size %d", addr, size);
1139     return 0;
1140 }
1141 
1142 static void vmxnet3_reset_interrupt_states(VMXNET3State *s)
1143 {
1144     int i;
1145     for (i = 0; i < ARRAY_SIZE(s->interrupt_states); i++) {
1146         s->interrupt_states[i].is_asserted = false;
1147         s->interrupt_states[i].is_pending = false;
1148         s->interrupt_states[i].is_masked = true;
1149     }
1150 }
1151 
1152 static void vmxnet3_reset_mac(VMXNET3State *s)
1153 {
1154     memcpy(&s->conf.macaddr.a, &s->perm_mac.a, sizeof(s->perm_mac.a));
1155     VMW_CFPRN("MAC address set to: " MAC_FMT, MAC_ARG(s->conf.macaddr.a));
1156 }
1157 
1158 static void vmxnet3_deactivate_device(VMXNET3State *s)
1159 {
1160     if (s->device_active) {
1161         VMW_CBPRN("Deactivating vmxnet3...");
1162         net_tx_pkt_reset(s->tx_pkt);
1163         net_tx_pkt_uninit(s->tx_pkt);
1164         net_rx_pkt_uninit(s->rx_pkt);
1165         s->device_active = false;
1166     }
1167 }
1168 
1169 static void vmxnet3_reset(VMXNET3State *s)
1170 {
1171     VMW_CBPRN("Resetting vmxnet3...");
1172 
1173     vmxnet3_deactivate_device(s);
1174     vmxnet3_reset_interrupt_states(s);
1175     s->drv_shmem = 0;
1176     s->tx_sop = true;
1177     s->skip_current_tx_pkt = false;
1178 }
1179 
1180 static void vmxnet3_update_rx_mode(VMXNET3State *s)
1181 {
1182     PCIDevice *d = PCI_DEVICE(s);
1183 
1184     s->rx_mode = VMXNET3_READ_DRV_SHARED32(d, s->drv_shmem,
1185                                            devRead.rxFilterConf.rxMode);
1186     VMW_CFPRN("RX mode: 0x%08X", s->rx_mode);
1187 }
1188 
1189 static void vmxnet3_update_vlan_filters(VMXNET3State *s)
1190 {
1191     int i;
1192     PCIDevice *d = PCI_DEVICE(s);
1193 
1194     /* Copy configuration from shared memory */
1195     VMXNET3_READ_DRV_SHARED(d, s->drv_shmem,
1196                             devRead.rxFilterConf.vfTable,
1197                             s->vlan_table,
1198                             sizeof(s->vlan_table));
1199 
1200     /* Invert byte order when needed */
1201     for (i = 0; i < ARRAY_SIZE(s->vlan_table); i++) {
1202         s->vlan_table[i] = le32_to_cpu(s->vlan_table[i]);
1203     }
1204 
1205     /* Dump configuration for debugging purposes */
1206     VMW_CFPRN("Configured VLANs:");
1207     for (i = 0; i < sizeof(s->vlan_table) * 8; i++) {
1208         if (VMXNET3_VFTABLE_ENTRY_IS_SET(s->vlan_table, i)) {
1209             VMW_CFPRN("\tVLAN %d is present", i);
1210         }
1211     }
1212 }
1213 
1214 static void vmxnet3_update_mcast_filters(VMXNET3State *s)
1215 {
1216     PCIDevice *d = PCI_DEVICE(s);
1217 
1218     uint16_t list_bytes =
1219         VMXNET3_READ_DRV_SHARED16(d, s->drv_shmem,
1220                                   devRead.rxFilterConf.mfTableLen);
1221 
1222     s->mcast_list_len = list_bytes / sizeof(s->mcast_list[0]);
1223 
1224     s->mcast_list = g_realloc(s->mcast_list, list_bytes);
1225     if (!s->mcast_list) {
1226         if (s->mcast_list_len == 0) {
1227             VMW_CFPRN("Current multicast list is empty");
1228         } else {
1229             VMW_ERPRN("Failed to allocate multicast list of %d elements",
1230                       s->mcast_list_len);
1231         }
1232         s->mcast_list_len = 0;
1233     } else {
1234         int i;
1235         hwaddr mcast_list_pa =
1236             VMXNET3_READ_DRV_SHARED64(d, s->drv_shmem,
1237                                       devRead.rxFilterConf.mfTablePA);
1238 
1239         pci_dma_read(d, mcast_list_pa, s->mcast_list, list_bytes);
1240 
1241         VMW_CFPRN("Current multicast list len is %d:", s->mcast_list_len);
1242         for (i = 0; i < s->mcast_list_len; i++) {
1243             VMW_CFPRN("\t" MAC_FMT, MAC_ARG(s->mcast_list[i].a));
1244         }
1245     }
1246 }
1247 
1248 static void vmxnet3_setup_rx_filtering(VMXNET3State *s)
1249 {
1250     vmxnet3_update_rx_mode(s);
1251     vmxnet3_update_vlan_filters(s);
1252     vmxnet3_update_mcast_filters(s);
1253 }
1254 
1255 static uint32_t vmxnet3_get_interrupt_config(VMXNET3State *s)
1256 {
1257     uint32_t interrupt_mode = VMXNET3_IT_AUTO | (VMXNET3_IMM_AUTO << 2);
1258     VMW_CFPRN("Interrupt config is 0x%X", interrupt_mode);
1259     return interrupt_mode;
1260 }
1261 
1262 static void vmxnet3_fill_stats(VMXNET3State *s)
1263 {
1264     int i;
1265     PCIDevice *d = PCI_DEVICE(s);
1266 
1267     if (!s->device_active)
1268         return;
1269 
1270     for (i = 0; i < s->txq_num; i++) {
1271         pci_dma_write(d,
1272                       s->txq_descr[i].tx_stats_pa,
1273                       &s->txq_descr[i].txq_stats,
1274                       sizeof(s->txq_descr[i].txq_stats));
1275     }
1276 
1277     for (i = 0; i < s->rxq_num; i++) {
1278         pci_dma_write(d,
1279                       s->rxq_descr[i].rx_stats_pa,
1280                       &s->rxq_descr[i].rxq_stats,
1281                       sizeof(s->rxq_descr[i].rxq_stats));
1282     }
1283 }
1284 
1285 static void vmxnet3_adjust_by_guest_type(VMXNET3State *s)
1286 {
1287     struct Vmxnet3_GOSInfo gos;
1288     PCIDevice *d = PCI_DEVICE(s);
1289 
1290     VMXNET3_READ_DRV_SHARED(d, s->drv_shmem, devRead.misc.driverInfo.gos,
1291                             &gos, sizeof(gos));
1292     s->rx_packets_compound =
1293         (gos.gosType == VMXNET3_GOS_TYPE_WIN) ? false : true;
1294 
1295     VMW_CFPRN("Guest type specifics: RXCOMPOUND: %d", s->rx_packets_compound);
1296 }
1297 
1298 static void
1299 vmxnet3_dump_conf_descr(const char *name,
1300                         struct Vmxnet3_VariableLenConfDesc *pm_descr)
1301 {
1302     VMW_CFPRN("%s descriptor dump: Version %u, Length %u",
1303               name, pm_descr->confVer, pm_descr->confLen);
1304 
1305 };
1306 
1307 static void vmxnet3_update_pm_state(VMXNET3State *s)
1308 {
1309     struct Vmxnet3_VariableLenConfDesc pm_descr;
1310     PCIDevice *d = PCI_DEVICE(s);
1311 
1312     pm_descr.confLen =
1313         VMXNET3_READ_DRV_SHARED32(d, s->drv_shmem, devRead.pmConfDesc.confLen);
1314     pm_descr.confVer =
1315         VMXNET3_READ_DRV_SHARED32(d, s->drv_shmem, devRead.pmConfDesc.confVer);
1316     pm_descr.confPA =
1317         VMXNET3_READ_DRV_SHARED64(d, s->drv_shmem, devRead.pmConfDesc.confPA);
1318 
1319     vmxnet3_dump_conf_descr("PM State", &pm_descr);
1320 }
1321 
1322 static void vmxnet3_update_features(VMXNET3State *s)
1323 {
1324     uint32_t guest_features;
1325     int rxcso_supported;
1326     PCIDevice *d = PCI_DEVICE(s);
1327 
1328     guest_features = VMXNET3_READ_DRV_SHARED32(d, s->drv_shmem,
1329                                                devRead.misc.uptFeatures);
1330 
1331     rxcso_supported = VMXNET_FLAG_IS_SET(guest_features, UPT1_F_RXCSUM);
1332     s->rx_vlan_stripping = VMXNET_FLAG_IS_SET(guest_features, UPT1_F_RXVLAN);
1333     s->lro_supported = VMXNET_FLAG_IS_SET(guest_features, UPT1_F_LRO);
1334 
1335     VMW_CFPRN("Features configuration: LRO: %d, RXCSUM: %d, VLANSTRIP: %d",
1336               s->lro_supported, rxcso_supported,
1337               s->rx_vlan_stripping);
1338     if (s->peer_has_vhdr) {
1339         qemu_set_offload(qemu_get_queue(s->nic)->peer,
1340                          rxcso_supported,
1341                          s->lro_supported,
1342                          s->lro_supported,
1343                          0,
1344                          0);
1345     }
1346 }
1347 
1348 static bool vmxnet3_verify_intx(VMXNET3State *s, int intx)
1349 {
1350     return s->msix_used || msi_enabled(PCI_DEVICE(s))
1351         || intx == pci_get_byte(s->parent_obj.config + PCI_INTERRUPT_PIN) - 1;
1352 }
1353 
1354 static void vmxnet3_validate_interrupt_idx(bool is_msix, int idx)
1355 {
1356     int max_ints = is_msix ? VMXNET3_MAX_INTRS : VMXNET3_MAX_NMSIX_INTRS;
1357     if (idx >= max_ints) {
1358         hw_error("Bad interrupt index: %d\n", idx);
1359     }
1360 }
1361 
1362 static void vmxnet3_validate_interrupts(VMXNET3State *s)
1363 {
1364     int i;
1365 
1366     VMW_CFPRN("Verifying event interrupt index (%d)", s->event_int_idx);
1367     vmxnet3_validate_interrupt_idx(s->msix_used, s->event_int_idx);
1368 
1369     for (i = 0; i < s->txq_num; i++) {
1370         int idx = s->txq_descr[i].intr_idx;
1371         VMW_CFPRN("Verifying TX queue %d interrupt index (%d)", i, idx);
1372         vmxnet3_validate_interrupt_idx(s->msix_used, idx);
1373     }
1374 
1375     for (i = 0; i < s->rxq_num; i++) {
1376         int idx = s->rxq_descr[i].intr_idx;
1377         VMW_CFPRN("Verifying RX queue %d interrupt index (%d)", i, idx);
1378         vmxnet3_validate_interrupt_idx(s->msix_used, idx);
1379     }
1380 }
1381 
1382 static bool vmxnet3_validate_queues(VMXNET3State *s)
1383 {
1384     /*
1385     * txq_num and rxq_num are total number of queues
1386     * configured by guest. These numbers must not
1387     * exceed corresponding maximal values.
1388     */
1389 
1390     if (s->txq_num > VMXNET3_DEVICE_MAX_TX_QUEUES) {
1391         qemu_log_mask(LOG_GUEST_ERROR, "vmxnet3: Bad TX queues number: %d\n",
1392                       s->txq_num);
1393         return false;
1394     }
1395 
1396     if (s->rxq_num > VMXNET3_DEVICE_MAX_RX_QUEUES) {
1397         qemu_log_mask(LOG_GUEST_ERROR, "vmxnet3: Bad RX queues number: %d\n",
1398                       s->rxq_num);
1399         return false;
1400     }
1401 
1402     return true;
1403 }
1404 
1405 static void vmxnet3_activate_device(VMXNET3State *s)
1406 {
1407     int i;
1408     static const uint32_t VMXNET3_DEF_TX_THRESHOLD = 1;
1409     PCIDevice *d = PCI_DEVICE(s);
1410     hwaddr qdescr_table_pa;
1411     uint64_t pa;
1412     uint32_t size;
1413 
1414     /* Verify configuration consistency */
1415     if (!vmxnet3_verify_driver_magic(d, s->drv_shmem)) {
1416         VMW_ERPRN("Device configuration received from driver is invalid");
1417         return;
1418     }
1419 
1420     /* Verify if device is active */
1421     if (s->device_active) {
1422         VMW_CFPRN("Vmxnet3 device is active");
1423         return;
1424     }
1425 
1426     s->txq_num =
1427         VMXNET3_READ_DRV_SHARED8(d, s->drv_shmem, devRead.misc.numTxQueues);
1428     s->rxq_num =
1429         VMXNET3_READ_DRV_SHARED8(d, s->drv_shmem, devRead.misc.numRxQueues);
1430 
1431     VMW_CFPRN("Number of TX/RX queues %u/%u", s->txq_num, s->rxq_num);
1432     if (!vmxnet3_validate_queues(s)) {
1433         return;
1434     }
1435 
1436     vmxnet3_adjust_by_guest_type(s);
1437     vmxnet3_update_features(s);
1438     vmxnet3_update_pm_state(s);
1439     vmxnet3_setup_rx_filtering(s);
1440     /* Cache fields from shared memory */
1441     s->mtu = VMXNET3_READ_DRV_SHARED32(d, s->drv_shmem, devRead.misc.mtu);
1442     assert(VMXNET3_MIN_MTU <= s->mtu && s->mtu <= VMXNET3_MAX_MTU);
1443     VMW_CFPRN("MTU is %u", s->mtu);
1444 
1445     s->max_rx_frags =
1446         VMXNET3_READ_DRV_SHARED16(d, s->drv_shmem, devRead.misc.maxNumRxSG);
1447 
1448     if (s->max_rx_frags == 0) {
1449         s->max_rx_frags = 1;
1450     }
1451 
1452     VMW_CFPRN("Max RX fragments is %u", s->max_rx_frags);
1453 
1454     s->event_int_idx =
1455         VMXNET3_READ_DRV_SHARED8(d, s->drv_shmem, devRead.intrConf.eventIntrIdx);
1456     assert(vmxnet3_verify_intx(s, s->event_int_idx));
1457     VMW_CFPRN("Events interrupt line is %u", s->event_int_idx);
1458 
1459     s->auto_int_masking =
1460         VMXNET3_READ_DRV_SHARED8(d, s->drv_shmem, devRead.intrConf.autoMask);
1461     VMW_CFPRN("Automatic interrupt masking is %d", (int)s->auto_int_masking);
1462 
1463     qdescr_table_pa =
1464         VMXNET3_READ_DRV_SHARED64(d, s->drv_shmem, devRead.misc.queueDescPA);
1465     VMW_CFPRN("TX queues descriptors table is at 0x%" PRIx64, qdescr_table_pa);
1466 
1467     /*
1468      * Worst-case scenario is a packet that holds all TX rings space so
1469      * we calculate total size of all TX rings for max TX fragments number
1470      */
1471     s->max_tx_frags = 0;
1472 
1473     /* TX queues */
1474     for (i = 0; i < s->txq_num; i++) {
1475         hwaddr qdescr_pa =
1476             qdescr_table_pa + i * sizeof(struct Vmxnet3_TxQueueDesc);
1477 
1478         /* Read interrupt number for this TX queue */
1479         s->txq_descr[i].intr_idx =
1480             VMXNET3_READ_TX_QUEUE_DESCR8(d, qdescr_pa, conf.intrIdx);
1481         assert(vmxnet3_verify_intx(s, s->txq_descr[i].intr_idx));
1482 
1483         VMW_CFPRN("TX Queue %d interrupt: %d", i, s->txq_descr[i].intr_idx);
1484 
1485         /* Read rings memory locations for TX queues */
1486         pa = VMXNET3_READ_TX_QUEUE_DESCR64(d, qdescr_pa, conf.txRingBasePA);
1487         size = VMXNET3_READ_TX_QUEUE_DESCR32(d, qdescr_pa, conf.txRingSize);
1488         if (size > VMXNET3_TX_RING_MAX_SIZE) {
1489             size = VMXNET3_TX_RING_MAX_SIZE;
1490         }
1491 
1492         vmxnet3_ring_init(d, &s->txq_descr[i].tx_ring, pa, size,
1493                           sizeof(struct Vmxnet3_TxDesc), false);
1494         VMXNET3_RING_DUMP(VMW_CFPRN, "TX", i, &s->txq_descr[i].tx_ring);
1495 
1496         s->max_tx_frags += size;
1497 
1498         /* TXC ring */
1499         pa = VMXNET3_READ_TX_QUEUE_DESCR64(d, qdescr_pa, conf.compRingBasePA);
1500         size = VMXNET3_READ_TX_QUEUE_DESCR32(d, qdescr_pa, conf.compRingSize);
1501         if (size > VMXNET3_TC_RING_MAX_SIZE) {
1502             size = VMXNET3_TC_RING_MAX_SIZE;
1503         }
1504         vmxnet3_ring_init(d, &s->txq_descr[i].comp_ring, pa, size,
1505                           sizeof(struct Vmxnet3_TxCompDesc), true);
1506         VMXNET3_RING_DUMP(VMW_CFPRN, "TXC", i, &s->txq_descr[i].comp_ring);
1507 
1508         s->txq_descr[i].tx_stats_pa =
1509             qdescr_pa + offsetof(struct Vmxnet3_TxQueueDesc, stats);
1510 
1511         memset(&s->txq_descr[i].txq_stats, 0,
1512                sizeof(s->txq_descr[i].txq_stats));
1513 
1514         /* Fill device-managed parameters for queues */
1515         VMXNET3_WRITE_TX_QUEUE_DESCR32(d, qdescr_pa,
1516                                        ctrl.txThreshold,
1517                                        VMXNET3_DEF_TX_THRESHOLD);
1518     }
1519 
1520     /* Preallocate TX packet wrapper */
1521     VMW_CFPRN("Max TX fragments is %u", s->max_tx_frags);
1522     net_tx_pkt_init(&s->tx_pkt, PCI_DEVICE(s), s->max_tx_frags);
1523     net_rx_pkt_init(&s->rx_pkt);
1524 
1525     /* Read rings memory locations for RX queues */
1526     for (i = 0; i < s->rxq_num; i++) {
1527         int j;
1528         hwaddr qd_pa =
1529             qdescr_table_pa + s->txq_num * sizeof(struct Vmxnet3_TxQueueDesc) +
1530             i * sizeof(struct Vmxnet3_RxQueueDesc);
1531 
1532         /* Read interrupt number for this RX queue */
1533         s->rxq_descr[i].intr_idx =
1534             VMXNET3_READ_TX_QUEUE_DESCR8(d, qd_pa, conf.intrIdx);
1535         assert(vmxnet3_verify_intx(s, s->rxq_descr[i].intr_idx));
1536 
1537         VMW_CFPRN("RX Queue %d interrupt: %d", i, s->rxq_descr[i].intr_idx);
1538 
1539         /* Read rings memory locations */
1540         for (j = 0; j < VMXNET3_RX_RINGS_PER_QUEUE; j++) {
1541             /* RX rings */
1542             pa = VMXNET3_READ_RX_QUEUE_DESCR64(d, qd_pa, conf.rxRingBasePA[j]);
1543             size = VMXNET3_READ_RX_QUEUE_DESCR32(d, qd_pa, conf.rxRingSize[j]);
1544             if (size > VMXNET3_RX_RING_MAX_SIZE) {
1545                 size = VMXNET3_RX_RING_MAX_SIZE;
1546             }
1547             vmxnet3_ring_init(d, &s->rxq_descr[i].rx_ring[j], pa, size,
1548                               sizeof(struct Vmxnet3_RxDesc), false);
1549             VMW_CFPRN("RX queue %d:%d: Base: %" PRIx64 ", Size: %d",
1550                       i, j, pa, size);
1551         }
1552 
1553         /* RXC ring */
1554         pa = VMXNET3_READ_RX_QUEUE_DESCR64(d, qd_pa, conf.compRingBasePA);
1555         size = VMXNET3_READ_RX_QUEUE_DESCR32(d, qd_pa, conf.compRingSize);
1556         if (size > VMXNET3_RC_RING_MAX_SIZE) {
1557             size = VMXNET3_RC_RING_MAX_SIZE;
1558         }
1559         vmxnet3_ring_init(d, &s->rxq_descr[i].comp_ring, pa, size,
1560                           sizeof(struct Vmxnet3_RxCompDesc), true);
1561         VMW_CFPRN("RXC queue %d: Base: %" PRIx64 ", Size: %d", i, pa, size);
1562 
1563         s->rxq_descr[i].rx_stats_pa =
1564             qd_pa + offsetof(struct Vmxnet3_RxQueueDesc, stats);
1565         memset(&s->rxq_descr[i].rxq_stats, 0,
1566                sizeof(s->rxq_descr[i].rxq_stats));
1567     }
1568 
1569     vmxnet3_validate_interrupts(s);
1570 
1571     /* Make sure everything is in place before device activation */
1572     smp_wmb();
1573 
1574     vmxnet3_reset_mac(s);
1575 
1576     s->device_active = true;
1577 }
1578 
1579 static void vmxnet3_handle_command(VMXNET3State *s, uint64_t cmd)
1580 {
1581     s->last_command = cmd;
1582 
1583     switch (cmd) {
1584     case VMXNET3_CMD_GET_PERM_MAC_HI:
1585         VMW_CBPRN("Set: Get upper part of permanent MAC");
1586         break;
1587 
1588     case VMXNET3_CMD_GET_PERM_MAC_LO:
1589         VMW_CBPRN("Set: Get lower part of permanent MAC");
1590         break;
1591 
1592     case VMXNET3_CMD_GET_STATS:
1593         VMW_CBPRN("Set: Get device statistics");
1594         vmxnet3_fill_stats(s);
1595         break;
1596 
1597     case VMXNET3_CMD_ACTIVATE_DEV:
1598         VMW_CBPRN("Set: Activating vmxnet3 device");
1599         vmxnet3_activate_device(s);
1600         break;
1601 
1602     case VMXNET3_CMD_UPDATE_RX_MODE:
1603         VMW_CBPRN("Set: Update rx mode");
1604         vmxnet3_update_rx_mode(s);
1605         break;
1606 
1607     case VMXNET3_CMD_UPDATE_VLAN_FILTERS:
1608         VMW_CBPRN("Set: Update VLAN filters");
1609         vmxnet3_update_vlan_filters(s);
1610         break;
1611 
1612     case VMXNET3_CMD_UPDATE_MAC_FILTERS:
1613         VMW_CBPRN("Set: Update MAC filters");
1614         vmxnet3_update_mcast_filters(s);
1615         break;
1616 
1617     case VMXNET3_CMD_UPDATE_FEATURE:
1618         VMW_CBPRN("Set: Update features");
1619         vmxnet3_update_features(s);
1620         break;
1621 
1622     case VMXNET3_CMD_UPDATE_PMCFG:
1623         VMW_CBPRN("Set: Update power management config");
1624         vmxnet3_update_pm_state(s);
1625         break;
1626 
1627     case VMXNET3_CMD_GET_LINK:
1628         VMW_CBPRN("Set: Get link");
1629         break;
1630 
1631     case VMXNET3_CMD_RESET_DEV:
1632         VMW_CBPRN("Set: Reset device");
1633         vmxnet3_reset(s);
1634         break;
1635 
1636     case VMXNET3_CMD_QUIESCE_DEV:
1637         VMW_CBPRN("Set: VMXNET3_CMD_QUIESCE_DEV - deactivate the device");
1638         vmxnet3_deactivate_device(s);
1639         break;
1640 
1641     case VMXNET3_CMD_GET_CONF_INTR:
1642         VMW_CBPRN("Set: VMXNET3_CMD_GET_CONF_INTR - interrupt configuration");
1643         break;
1644 
1645     case VMXNET3_CMD_GET_ADAPTIVE_RING_INFO:
1646         VMW_CBPRN("Set: VMXNET3_CMD_GET_ADAPTIVE_RING_INFO - "
1647                   "adaptive ring info flags");
1648         break;
1649 
1650     case VMXNET3_CMD_GET_DID_LO:
1651         VMW_CBPRN("Set: Get lower part of device ID");
1652         break;
1653 
1654     case VMXNET3_CMD_GET_DID_HI:
1655         VMW_CBPRN("Set: Get upper part of device ID");
1656         break;
1657 
1658     case VMXNET3_CMD_GET_DEV_EXTRA_INFO:
1659         VMW_CBPRN("Set: Get device extra info");
1660         break;
1661 
1662     default:
1663         VMW_CBPRN("Received unknown command: %" PRIx64, cmd);
1664         break;
1665     }
1666 }
1667 
1668 static uint64_t vmxnet3_get_command_status(VMXNET3State *s)
1669 {
1670     uint64_t ret;
1671 
1672     switch (s->last_command) {
1673     case VMXNET3_CMD_ACTIVATE_DEV:
1674         ret = (s->device_active) ? 0 : 1;
1675         VMW_CFPRN("Device active: %" PRIx64, ret);
1676         break;
1677 
1678     case VMXNET3_CMD_RESET_DEV:
1679     case VMXNET3_CMD_QUIESCE_DEV:
1680     case VMXNET3_CMD_GET_QUEUE_STATUS:
1681     case VMXNET3_CMD_GET_DEV_EXTRA_INFO:
1682         ret = 0;
1683         break;
1684 
1685     case VMXNET3_CMD_GET_LINK:
1686         ret = s->link_status_and_speed;
1687         VMW_CFPRN("Link and speed: %" PRIx64, ret);
1688         break;
1689 
1690     case VMXNET3_CMD_GET_PERM_MAC_LO:
1691         ret = vmxnet3_get_mac_low(&s->perm_mac);
1692         break;
1693 
1694     case VMXNET3_CMD_GET_PERM_MAC_HI:
1695         ret = vmxnet3_get_mac_high(&s->perm_mac);
1696         break;
1697 
1698     case VMXNET3_CMD_GET_CONF_INTR:
1699         ret = vmxnet3_get_interrupt_config(s);
1700         break;
1701 
1702     case VMXNET3_CMD_GET_ADAPTIVE_RING_INFO:
1703         ret = VMXNET3_DISABLE_ADAPTIVE_RING;
1704         break;
1705 
1706     case VMXNET3_CMD_GET_DID_LO:
1707         ret = PCI_DEVICE_ID_VMWARE_VMXNET3;
1708         break;
1709 
1710     case VMXNET3_CMD_GET_DID_HI:
1711         ret = VMXNET3_DEVICE_REVISION;
1712         break;
1713 
1714     default:
1715         VMW_WRPRN("Received request for unknown command: %x", s->last_command);
1716         ret = 0;
1717         break;
1718     }
1719 
1720     return ret;
1721 }
1722 
1723 static void vmxnet3_set_events(VMXNET3State *s, uint32_t val)
1724 {
1725     uint32_t events;
1726     PCIDevice *d = PCI_DEVICE(s);
1727 
1728     VMW_CBPRN("Setting events: 0x%x", val);
1729     events = VMXNET3_READ_DRV_SHARED32(d, s->drv_shmem, ecr) | val;
1730     VMXNET3_WRITE_DRV_SHARED32(d, s->drv_shmem, ecr, events);
1731 }
1732 
1733 static void vmxnet3_ack_events(VMXNET3State *s, uint32_t val)
1734 {
1735     PCIDevice *d = PCI_DEVICE(s);
1736     uint32_t events;
1737 
1738     VMW_CBPRN("Clearing events: 0x%x", val);
1739     events = VMXNET3_READ_DRV_SHARED32(d, s->drv_shmem, ecr) & ~val;
1740     VMXNET3_WRITE_DRV_SHARED32(d, s->drv_shmem, ecr, events);
1741 }
1742 
1743 static void
1744 vmxnet3_io_bar1_write(void *opaque,
1745                       hwaddr addr,
1746                       uint64_t val,
1747                       unsigned size)
1748 {
1749     VMXNET3State *s = opaque;
1750 
1751     switch (addr) {
1752     /* Vmxnet3 Revision Report Selection */
1753     case VMXNET3_REG_VRRS:
1754         VMW_CBPRN("Write BAR1 [VMXNET3_REG_VRRS] = %" PRIx64 ", size %d",
1755                   val, size);
1756         break;
1757 
1758     /* UPT Version Report Selection */
1759     case VMXNET3_REG_UVRS:
1760         VMW_CBPRN("Write BAR1 [VMXNET3_REG_UVRS] = %" PRIx64 ", size %d",
1761                   val, size);
1762         break;
1763 
1764     /* Driver Shared Address Low */
1765     case VMXNET3_REG_DSAL:
1766         VMW_CBPRN("Write BAR1 [VMXNET3_REG_DSAL] = %" PRIx64 ", size %d",
1767                   val, size);
1768         /*
1769          * Guest driver will first write the low part of the shared
1770          * memory address. We save it to temp variable and set the
1771          * shared address only after we get the high part
1772          */
1773         if (val == 0) {
1774             vmxnet3_deactivate_device(s);
1775         }
1776         s->temp_shared_guest_driver_memory = val;
1777         s->drv_shmem = 0;
1778         break;
1779 
1780     /* Driver Shared Address High */
1781     case VMXNET3_REG_DSAH:
1782         VMW_CBPRN("Write BAR1 [VMXNET3_REG_DSAH] = %" PRIx64 ", size %d",
1783                   val, size);
1784         /*
1785          * Set the shared memory between guest driver and device.
1786          * We already should have low address part.
1787          */
1788         s->drv_shmem = s->temp_shared_guest_driver_memory | (val << 32);
1789         break;
1790 
1791     /* Command */
1792     case VMXNET3_REG_CMD:
1793         VMW_CBPRN("Write BAR1 [VMXNET3_REG_CMD] = %" PRIx64 ", size %d",
1794                   val, size);
1795         vmxnet3_handle_command(s, val);
1796         break;
1797 
1798     /* MAC Address Low */
1799     case VMXNET3_REG_MACL:
1800         VMW_CBPRN("Write BAR1 [VMXNET3_REG_MACL] = %" PRIx64 ", size %d",
1801                   val, size);
1802         s->temp_mac = val;
1803         break;
1804 
1805     /* MAC Address High */
1806     case VMXNET3_REG_MACH:
1807         VMW_CBPRN("Write BAR1 [VMXNET3_REG_MACH] = %" PRIx64 ", size %d",
1808                   val, size);
1809         vmxnet3_set_variable_mac(s, val, s->temp_mac);
1810         break;
1811 
1812     /* Interrupt Cause Register */
1813     case VMXNET3_REG_ICR:
1814         VMW_CBPRN("Write BAR1 [VMXNET3_REG_ICR] = %" PRIx64 ", size %d",
1815                   val, size);
1816         qemu_log_mask(LOG_GUEST_ERROR,
1817                       "%s: write to read-only register VMXNET3_REG_ICR\n",
1818                       TYPE_VMXNET3);
1819         break;
1820 
1821     /* Event Cause Register */
1822     case VMXNET3_REG_ECR:
1823         VMW_CBPRN("Write BAR1 [VMXNET3_REG_ECR] = %" PRIx64 ", size %d",
1824                   val, size);
1825         vmxnet3_ack_events(s, val);
1826         break;
1827 
1828     default:
1829         VMW_CBPRN("Unknown Write to BAR1 [%" PRIx64 "] = %" PRIx64 ", size %d",
1830                   addr, val, size);
1831         break;
1832     }
1833 }
1834 
1835 static uint64_t
1836 vmxnet3_io_bar1_read(void *opaque, hwaddr addr, unsigned size)
1837 {
1838         VMXNET3State *s = opaque;
1839         uint64_t ret = 0;
1840 
1841         switch (addr) {
1842         /* Vmxnet3 Revision Report Selection */
1843         case VMXNET3_REG_VRRS:
1844             VMW_CBPRN("Read BAR1 [VMXNET3_REG_VRRS], size %d", size);
1845             ret = VMXNET3_DEVICE_REVISION;
1846             break;
1847 
1848         /* UPT Version Report Selection */
1849         case VMXNET3_REG_UVRS:
1850             VMW_CBPRN("Read BAR1 [VMXNET3_REG_UVRS], size %d", size);
1851             ret = VMXNET3_UPT_REVISION;
1852             break;
1853 
1854         /* Command */
1855         case VMXNET3_REG_CMD:
1856             VMW_CBPRN("Read BAR1 [VMXNET3_REG_CMD], size %d", size);
1857             ret = vmxnet3_get_command_status(s);
1858             break;
1859 
1860         /* MAC Address Low */
1861         case VMXNET3_REG_MACL:
1862             VMW_CBPRN("Read BAR1 [VMXNET3_REG_MACL], size %d", size);
1863             ret = vmxnet3_get_mac_low(&s->conf.macaddr);
1864             break;
1865 
1866         /* MAC Address High */
1867         case VMXNET3_REG_MACH:
1868             VMW_CBPRN("Read BAR1 [VMXNET3_REG_MACH], size %d", size);
1869             ret = vmxnet3_get_mac_high(&s->conf.macaddr);
1870             break;
1871 
1872         /*
1873          * Interrupt Cause Register
1874          * Used for legacy interrupts only so interrupt index always 0
1875          */
1876         case VMXNET3_REG_ICR:
1877             VMW_CBPRN("Read BAR1 [VMXNET3_REG_ICR], size %d", size);
1878             if (vmxnet3_interrupt_asserted(s, 0)) {
1879                 vmxnet3_clear_interrupt(s, 0);
1880                 ret = true;
1881             } else {
1882                 ret = false;
1883             }
1884             break;
1885 
1886         default:
1887             VMW_CBPRN("Unknow read BAR1[%" PRIx64 "], %d bytes", addr, size);
1888             break;
1889         }
1890 
1891         return ret;
1892 }
1893 
1894 static int
1895 vmxnet3_can_receive(NetClientState *nc)
1896 {
1897     VMXNET3State *s = qemu_get_nic_opaque(nc);
1898     return s->device_active &&
1899            VMXNET_FLAG_IS_SET(s->link_status_and_speed, VMXNET3_LINK_STATUS_UP);
1900 }
1901 
1902 static inline bool
1903 vmxnet3_is_registered_vlan(VMXNET3State *s, const void *data)
1904 {
1905     uint16_t vlan_tag = eth_get_pkt_tci(data) & VLAN_VID_MASK;
1906     if (IS_SPECIAL_VLAN_ID(vlan_tag)) {
1907         return true;
1908     }
1909 
1910     return VMXNET3_VFTABLE_ENTRY_IS_SET(s->vlan_table, vlan_tag);
1911 }
1912 
1913 static bool
1914 vmxnet3_is_allowed_mcast_group(VMXNET3State *s, const uint8_t *group_mac)
1915 {
1916     int i;
1917     for (i = 0; i < s->mcast_list_len; i++) {
1918         if (!memcmp(group_mac, s->mcast_list[i].a, sizeof(s->mcast_list[i]))) {
1919             return true;
1920         }
1921     }
1922     return false;
1923 }
1924 
1925 static bool
1926 vmxnet3_rx_filter_may_indicate(VMXNET3State *s, const void *data,
1927     size_t size)
1928 {
1929     struct eth_header *ehdr = PKT_GET_ETH_HDR(data);
1930 
1931     if (VMXNET_FLAG_IS_SET(s->rx_mode, VMXNET3_RXM_PROMISC)) {
1932         return true;
1933     }
1934 
1935     if (!vmxnet3_is_registered_vlan(s, data)) {
1936         return false;
1937     }
1938 
1939     switch (net_rx_pkt_get_packet_type(s->rx_pkt)) {
1940     case ETH_PKT_UCAST:
1941         if (!VMXNET_FLAG_IS_SET(s->rx_mode, VMXNET3_RXM_UCAST)) {
1942             return false;
1943         }
1944         if (memcmp(s->conf.macaddr.a, ehdr->h_dest, ETH_ALEN)) {
1945             return false;
1946         }
1947         break;
1948 
1949     case ETH_PKT_BCAST:
1950         if (!VMXNET_FLAG_IS_SET(s->rx_mode, VMXNET3_RXM_BCAST)) {
1951             return false;
1952         }
1953         break;
1954 
1955     case ETH_PKT_MCAST:
1956         if (VMXNET_FLAG_IS_SET(s->rx_mode, VMXNET3_RXM_ALL_MULTI)) {
1957             return true;
1958         }
1959         if (!VMXNET_FLAG_IS_SET(s->rx_mode, VMXNET3_RXM_MCAST)) {
1960             return false;
1961         }
1962         if (!vmxnet3_is_allowed_mcast_group(s, ehdr->h_dest)) {
1963             return false;
1964         }
1965         break;
1966 
1967     default:
1968         g_assert_not_reached();
1969     }
1970 
1971     return true;
1972 }
1973 
1974 static ssize_t
1975 vmxnet3_receive(NetClientState *nc, const uint8_t *buf, size_t size)
1976 {
1977     VMXNET3State *s = qemu_get_nic_opaque(nc);
1978     size_t bytes_indicated;
1979     uint8_t min_buf[MIN_BUF_SIZE];
1980 
1981     if (!vmxnet3_can_receive(nc)) {
1982         VMW_PKPRN("Cannot receive now");
1983         return -1;
1984     }
1985 
1986     if (s->peer_has_vhdr) {
1987         net_rx_pkt_set_vhdr(s->rx_pkt, (struct virtio_net_hdr *)buf);
1988         buf += sizeof(struct virtio_net_hdr);
1989         size -= sizeof(struct virtio_net_hdr);
1990     }
1991 
1992     /* Pad to minimum Ethernet frame length */
1993     if (size < sizeof(min_buf)) {
1994         memcpy(min_buf, buf, size);
1995         memset(&min_buf[size], 0, sizeof(min_buf) - size);
1996         buf = min_buf;
1997         size = sizeof(min_buf);
1998     }
1999 
2000     net_rx_pkt_set_packet_type(s->rx_pkt,
2001         get_eth_packet_type(PKT_GET_ETH_HDR(buf)));
2002 
2003     if (vmxnet3_rx_filter_may_indicate(s, buf, size)) {
2004         net_rx_pkt_set_protocols(s->rx_pkt, buf, size);
2005         vmxnet3_rx_need_csum_calculate(s->rx_pkt, buf, size);
2006         net_rx_pkt_attach_data(s->rx_pkt, buf, size, s->rx_vlan_stripping);
2007         bytes_indicated = vmxnet3_indicate_packet(s) ? size : -1;
2008         if (bytes_indicated < size) {
2009             VMW_PKPRN("RX: %zu of %zu bytes indicated", bytes_indicated, size);
2010         }
2011     } else {
2012         VMW_PKPRN("Packet dropped by RX filter");
2013         bytes_indicated = size;
2014     }
2015 
2016     assert(size > 0);
2017     assert(bytes_indicated != 0);
2018     return bytes_indicated;
2019 }
2020 
2021 static void vmxnet3_set_link_status(NetClientState *nc)
2022 {
2023     VMXNET3State *s = qemu_get_nic_opaque(nc);
2024 
2025     if (nc->link_down) {
2026         s->link_status_and_speed &= ~VMXNET3_LINK_STATUS_UP;
2027     } else {
2028         s->link_status_and_speed |= VMXNET3_LINK_STATUS_UP;
2029     }
2030 
2031     vmxnet3_set_events(s, VMXNET3_ECR_LINK);
2032     vmxnet3_trigger_interrupt(s, s->event_int_idx);
2033 }
2034 
2035 static NetClientInfo net_vmxnet3_info = {
2036         .type = NET_CLIENT_DRIVER_NIC,
2037         .size = sizeof(NICState),
2038         .receive = vmxnet3_receive,
2039         .link_status_changed = vmxnet3_set_link_status,
2040 };
2041 
2042 static bool vmxnet3_peer_has_vnet_hdr(VMXNET3State *s)
2043 {
2044     NetClientState *nc = qemu_get_queue(s->nic);
2045 
2046     if (qemu_has_vnet_hdr(nc->peer)) {
2047         return true;
2048     }
2049 
2050     return false;
2051 }
2052 
2053 static void vmxnet3_net_uninit(VMXNET3State *s)
2054 {
2055     g_free(s->mcast_list);
2056     vmxnet3_deactivate_device(s);
2057     qemu_del_nic(s->nic);
2058 }
2059 
2060 static void vmxnet3_net_init(VMXNET3State *s)
2061 {
2062     DeviceState *d = DEVICE(s);
2063 
2064     VMW_CBPRN("vmxnet3_net_init called...");
2065 
2066     qemu_macaddr_default_if_unset(&s->conf.macaddr);
2067 
2068     /* Windows guest will query the address that was set on init */
2069     memcpy(&s->perm_mac.a, &s->conf.macaddr.a, sizeof(s->perm_mac.a));
2070 
2071     s->mcast_list = NULL;
2072     s->mcast_list_len = 0;
2073 
2074     s->link_status_and_speed = VMXNET3_LINK_SPEED | VMXNET3_LINK_STATUS_UP;
2075 
2076     VMW_CFPRN("Permanent MAC: " MAC_FMT, MAC_ARG(s->perm_mac.a));
2077 
2078     s->nic = qemu_new_nic(&net_vmxnet3_info, &s->conf,
2079                           object_get_typename(OBJECT(s)),
2080                           d->id, s);
2081 
2082     s->peer_has_vhdr = vmxnet3_peer_has_vnet_hdr(s);
2083     s->tx_sop = true;
2084     s->skip_current_tx_pkt = false;
2085     s->tx_pkt = NULL;
2086     s->rx_pkt = NULL;
2087     s->rx_vlan_stripping = false;
2088     s->lro_supported = false;
2089 
2090     if (s->peer_has_vhdr) {
2091         qemu_set_vnet_hdr_len(qemu_get_queue(s->nic)->peer,
2092             sizeof(struct virtio_net_hdr));
2093 
2094         qemu_using_vnet_hdr(qemu_get_queue(s->nic)->peer, 1);
2095     }
2096 
2097     qemu_format_nic_info_str(qemu_get_queue(s->nic), s->conf.macaddr.a);
2098 }
2099 
2100 static void
2101 vmxnet3_unuse_msix_vectors(VMXNET3State *s, int num_vectors)
2102 {
2103     PCIDevice *d = PCI_DEVICE(s);
2104     int i;
2105     for (i = 0; i < num_vectors; i++) {
2106         msix_vector_unuse(d, i);
2107     }
2108 }
2109 
2110 static void
2111 vmxnet3_use_msix_vectors(VMXNET3State *s, int num_vectors)
2112 {
2113     PCIDevice *d = PCI_DEVICE(s);
2114     int i;
2115     for (i = 0; i < num_vectors; i++) {
2116         msix_vector_use(d, i);
2117     }
2118 }
2119 
2120 static bool
2121 vmxnet3_init_msix(VMXNET3State *s)
2122 {
2123     PCIDevice *d = PCI_DEVICE(s);
2124     int res = msix_init(d, VMXNET3_MAX_INTRS,
2125                         &s->msix_bar,
2126                         VMXNET3_MSIX_BAR_IDX, VMXNET3_OFF_MSIX_TABLE,
2127                         &s->msix_bar,
2128                         VMXNET3_MSIX_BAR_IDX, VMXNET3_OFF_MSIX_PBA(s),
2129                         VMXNET3_MSIX_OFFSET(s), NULL);
2130 
2131     if (0 > res) {
2132         VMW_WRPRN("Failed to initialize MSI-X, error %d", res);
2133         s->msix_used = false;
2134     } else {
2135         vmxnet3_use_msix_vectors(s, VMXNET3_MAX_INTRS);
2136         s->msix_used = true;
2137     }
2138     return s->msix_used;
2139 }
2140 
2141 static void
2142 vmxnet3_cleanup_msix(VMXNET3State *s)
2143 {
2144     PCIDevice *d = PCI_DEVICE(s);
2145 
2146     if (s->msix_used) {
2147         vmxnet3_unuse_msix_vectors(s, VMXNET3_MAX_INTRS);
2148         msix_uninit(d, &s->msix_bar, &s->msix_bar);
2149     }
2150 }
2151 
2152 static void
2153 vmxnet3_cleanup_msi(VMXNET3State *s)
2154 {
2155     PCIDevice *d = PCI_DEVICE(s);
2156 
2157     msi_uninit(d);
2158 }
2159 
2160 static const MemoryRegionOps b0_ops = {
2161     .read = vmxnet3_io_bar0_read,
2162     .write = vmxnet3_io_bar0_write,
2163     .endianness = DEVICE_LITTLE_ENDIAN,
2164     .impl = {
2165             .min_access_size = 4,
2166             .max_access_size = 4,
2167     },
2168 };
2169 
2170 static const MemoryRegionOps b1_ops = {
2171     .read = vmxnet3_io_bar1_read,
2172     .write = vmxnet3_io_bar1_write,
2173     .endianness = DEVICE_LITTLE_ENDIAN,
2174     .impl = {
2175             .min_access_size = 4,
2176             .max_access_size = 4,
2177     },
2178 };
2179 
2180 static uint64_t vmxnet3_device_serial_num(VMXNET3State *s)
2181 {
2182     uint64_t dsn_payload;
2183     uint8_t *dsnp = (uint8_t *)&dsn_payload;
2184 
2185     dsnp[0] = 0xfe;
2186     dsnp[1] = s->conf.macaddr.a[3];
2187     dsnp[2] = s->conf.macaddr.a[4];
2188     dsnp[3] = s->conf.macaddr.a[5];
2189     dsnp[4] = s->conf.macaddr.a[0];
2190     dsnp[5] = s->conf.macaddr.a[1];
2191     dsnp[6] = s->conf.macaddr.a[2];
2192     dsnp[7] = 0xff;
2193     return dsn_payload;
2194 }
2195 
2196 
2197 #define VMXNET3_USE_64BIT         (true)
2198 #define VMXNET3_PER_VECTOR_MASK   (false)
2199 
2200 static void vmxnet3_pci_realize(PCIDevice *pci_dev, Error **errp)
2201 {
2202     VMXNET3State *s = VMXNET3(pci_dev);
2203     int ret;
2204 
2205     VMW_CBPRN("Starting init...");
2206 
2207     memory_region_init_io(&s->bar0, OBJECT(s), &b0_ops, s,
2208                           "vmxnet3-b0", VMXNET3_PT_REG_SIZE);
2209     pci_register_bar(pci_dev, VMXNET3_BAR0_IDX,
2210                      PCI_BASE_ADDRESS_SPACE_MEMORY, &s->bar0);
2211 
2212     memory_region_init_io(&s->bar1, OBJECT(s), &b1_ops, s,
2213                           "vmxnet3-b1", VMXNET3_VD_REG_SIZE);
2214     pci_register_bar(pci_dev, VMXNET3_BAR1_IDX,
2215                      PCI_BASE_ADDRESS_SPACE_MEMORY, &s->bar1);
2216 
2217     memory_region_init(&s->msix_bar, OBJECT(s), "vmxnet3-msix-bar",
2218                        VMXNET3_MSIX_BAR_SIZE);
2219     pci_register_bar(pci_dev, VMXNET3_MSIX_BAR_IDX,
2220                      PCI_BASE_ADDRESS_SPACE_MEMORY, &s->msix_bar);
2221 
2222     vmxnet3_reset_interrupt_states(s);
2223 
2224     /* Interrupt pin A */
2225     pci_dev->config[PCI_INTERRUPT_PIN] = 0x01;
2226 
2227     ret = msi_init(pci_dev, VMXNET3_MSI_OFFSET(s), VMXNET3_MAX_NMSIX_INTRS,
2228                    VMXNET3_USE_64BIT, VMXNET3_PER_VECTOR_MASK, NULL);
2229     /* Any error other than -ENOTSUP(board's MSI support is broken)
2230      * is a programming error. Fall back to INTx silently on -ENOTSUP */
2231     assert(!ret || ret == -ENOTSUP);
2232 
2233     if (!vmxnet3_init_msix(s)) {
2234         VMW_WRPRN("Failed to initialize MSI-X, configuration is inconsistent.");
2235     }
2236 
2237     vmxnet3_net_init(s);
2238 
2239     if (pci_is_express(pci_dev)) {
2240         if (pci_bus_is_express(pci_get_bus(pci_dev))) {
2241             pcie_endpoint_cap_init(pci_dev, VMXNET3_EXP_EP_OFFSET);
2242         }
2243 
2244         pcie_dev_ser_num_init(pci_dev, VMXNET3_DSN_OFFSET,
2245                               vmxnet3_device_serial_num(s));
2246     }
2247 }
2248 
2249 static void vmxnet3_instance_init(Object *obj)
2250 {
2251     VMXNET3State *s = VMXNET3(obj);
2252     device_add_bootindex_property(obj, &s->conf.bootindex,
2253                                   "bootindex", "/ethernet-phy@0",
2254                                   DEVICE(obj));
2255 }
2256 
2257 static void vmxnet3_pci_uninit(PCIDevice *pci_dev)
2258 {
2259     VMXNET3State *s = VMXNET3(pci_dev);
2260 
2261     VMW_CBPRN("Starting uninit...");
2262 
2263     vmxnet3_net_uninit(s);
2264 
2265     vmxnet3_cleanup_msix(s);
2266 
2267     vmxnet3_cleanup_msi(s);
2268 }
2269 
2270 static void vmxnet3_qdev_reset(DeviceState *dev)
2271 {
2272     PCIDevice *d = PCI_DEVICE(dev);
2273     VMXNET3State *s = VMXNET3(d);
2274 
2275     VMW_CBPRN("Starting QDEV reset...");
2276     vmxnet3_reset(s);
2277 }
2278 
2279 static bool vmxnet3_mc_list_needed(void *opaque)
2280 {
2281     return true;
2282 }
2283 
2284 static int vmxnet3_mcast_list_pre_load(void *opaque)
2285 {
2286     VMXNET3State *s = opaque;
2287 
2288     s->mcast_list = g_malloc(s->mcast_list_buff_size);
2289 
2290     return 0;
2291 }
2292 
2293 
2294 static int vmxnet3_pre_save(void *opaque)
2295 {
2296     VMXNET3State *s = opaque;
2297 
2298     s->mcast_list_buff_size = s->mcast_list_len * sizeof(MACAddr);
2299 
2300     return 0;
2301 }
2302 
2303 static const VMStateDescription vmxstate_vmxnet3_mcast_list = {
2304     .name = "vmxnet3/mcast_list",
2305     .version_id = 1,
2306     .minimum_version_id = 1,
2307     .pre_load = vmxnet3_mcast_list_pre_load,
2308     .needed = vmxnet3_mc_list_needed,
2309     .fields = (VMStateField[]) {
2310         VMSTATE_VBUFFER_UINT32(mcast_list, VMXNET3State, 0, NULL,
2311             mcast_list_buff_size),
2312         VMSTATE_END_OF_LIST()
2313     }
2314 };
2315 
2316 static const VMStateDescription vmstate_vmxnet3_ring = {
2317     .name = "vmxnet3-ring",
2318     .version_id = 0,
2319     .fields = (VMStateField[]) {
2320         VMSTATE_UINT64(pa, Vmxnet3Ring),
2321         VMSTATE_UINT32(size, Vmxnet3Ring),
2322         VMSTATE_UINT32(cell_size, Vmxnet3Ring),
2323         VMSTATE_UINT32(next, Vmxnet3Ring),
2324         VMSTATE_UINT8(gen, Vmxnet3Ring),
2325         VMSTATE_END_OF_LIST()
2326     }
2327 };
2328 
2329 static const VMStateDescription vmstate_vmxnet3_tx_stats = {
2330     .name = "vmxnet3-tx-stats",
2331     .version_id = 0,
2332     .fields = (VMStateField[]) {
2333         VMSTATE_UINT64(TSOPktsTxOK, struct UPT1_TxStats),
2334         VMSTATE_UINT64(TSOBytesTxOK, struct UPT1_TxStats),
2335         VMSTATE_UINT64(ucastPktsTxOK, struct UPT1_TxStats),
2336         VMSTATE_UINT64(ucastBytesTxOK, struct UPT1_TxStats),
2337         VMSTATE_UINT64(mcastPktsTxOK, struct UPT1_TxStats),
2338         VMSTATE_UINT64(mcastBytesTxOK, struct UPT1_TxStats),
2339         VMSTATE_UINT64(bcastPktsTxOK, struct UPT1_TxStats),
2340         VMSTATE_UINT64(bcastBytesTxOK, struct UPT1_TxStats),
2341         VMSTATE_UINT64(pktsTxError, struct UPT1_TxStats),
2342         VMSTATE_UINT64(pktsTxDiscard, struct UPT1_TxStats),
2343         VMSTATE_END_OF_LIST()
2344     }
2345 };
2346 
2347 static const VMStateDescription vmstate_vmxnet3_txq_descr = {
2348     .name = "vmxnet3-txq-descr",
2349     .version_id = 0,
2350     .fields = (VMStateField[]) {
2351         VMSTATE_STRUCT(tx_ring, Vmxnet3TxqDescr, 0, vmstate_vmxnet3_ring,
2352                        Vmxnet3Ring),
2353         VMSTATE_STRUCT(comp_ring, Vmxnet3TxqDescr, 0, vmstate_vmxnet3_ring,
2354                        Vmxnet3Ring),
2355         VMSTATE_UINT8(intr_idx, Vmxnet3TxqDescr),
2356         VMSTATE_UINT64(tx_stats_pa, Vmxnet3TxqDescr),
2357         VMSTATE_STRUCT(txq_stats, Vmxnet3TxqDescr, 0, vmstate_vmxnet3_tx_stats,
2358                        struct UPT1_TxStats),
2359         VMSTATE_END_OF_LIST()
2360     }
2361 };
2362 
2363 static const VMStateDescription vmstate_vmxnet3_rx_stats = {
2364     .name = "vmxnet3-rx-stats",
2365     .version_id = 0,
2366     .fields = (VMStateField[]) {
2367         VMSTATE_UINT64(LROPktsRxOK, struct UPT1_RxStats),
2368         VMSTATE_UINT64(LROBytesRxOK, struct UPT1_RxStats),
2369         VMSTATE_UINT64(ucastPktsRxOK, struct UPT1_RxStats),
2370         VMSTATE_UINT64(ucastBytesRxOK, struct UPT1_RxStats),
2371         VMSTATE_UINT64(mcastPktsRxOK, struct UPT1_RxStats),
2372         VMSTATE_UINT64(mcastBytesRxOK, struct UPT1_RxStats),
2373         VMSTATE_UINT64(bcastPktsRxOK, struct UPT1_RxStats),
2374         VMSTATE_UINT64(bcastBytesRxOK, struct UPT1_RxStats),
2375         VMSTATE_UINT64(pktsRxOutOfBuf, struct UPT1_RxStats),
2376         VMSTATE_UINT64(pktsRxError, struct UPT1_RxStats),
2377         VMSTATE_END_OF_LIST()
2378     }
2379 };
2380 
2381 static const VMStateDescription vmstate_vmxnet3_rxq_descr = {
2382     .name = "vmxnet3-rxq-descr",
2383     .version_id = 0,
2384     .fields = (VMStateField[]) {
2385         VMSTATE_STRUCT_ARRAY(rx_ring, Vmxnet3RxqDescr,
2386                              VMXNET3_RX_RINGS_PER_QUEUE, 0,
2387                              vmstate_vmxnet3_ring, Vmxnet3Ring),
2388         VMSTATE_STRUCT(comp_ring, Vmxnet3RxqDescr, 0, vmstate_vmxnet3_ring,
2389                        Vmxnet3Ring),
2390         VMSTATE_UINT8(intr_idx, Vmxnet3RxqDescr),
2391         VMSTATE_UINT64(rx_stats_pa, Vmxnet3RxqDescr),
2392         VMSTATE_STRUCT(rxq_stats, Vmxnet3RxqDescr, 0, vmstate_vmxnet3_rx_stats,
2393                        struct UPT1_RxStats),
2394         VMSTATE_END_OF_LIST()
2395     }
2396 };
2397 
2398 static int vmxnet3_post_load(void *opaque, int version_id)
2399 {
2400     VMXNET3State *s = opaque;
2401 
2402     net_tx_pkt_init(&s->tx_pkt, PCI_DEVICE(s), s->max_tx_frags);
2403     net_rx_pkt_init(&s->rx_pkt);
2404 
2405     if (s->msix_used) {
2406         vmxnet3_use_msix_vectors(s, VMXNET3_MAX_INTRS);
2407     }
2408 
2409     if (!vmxnet3_validate_queues(s)) {
2410         return -1;
2411     }
2412     vmxnet3_validate_interrupts(s);
2413 
2414     return 0;
2415 }
2416 
2417 static const VMStateDescription vmstate_vmxnet3_int_state = {
2418     .name = "vmxnet3-int-state",
2419     .version_id = 0,
2420     .fields = (VMStateField[]) {
2421         VMSTATE_BOOL(is_masked, Vmxnet3IntState),
2422         VMSTATE_BOOL(is_pending, Vmxnet3IntState),
2423         VMSTATE_BOOL(is_asserted, Vmxnet3IntState),
2424         VMSTATE_END_OF_LIST()
2425     }
2426 };
2427 
2428 static const VMStateDescription vmstate_vmxnet3 = {
2429     .name = "vmxnet3",
2430     .version_id = 1,
2431     .minimum_version_id = 1,
2432     .pre_save = vmxnet3_pre_save,
2433     .post_load = vmxnet3_post_load,
2434     .fields = (VMStateField[]) {
2435             VMSTATE_PCI_DEVICE(parent_obj, VMXNET3State),
2436             VMSTATE_MSIX(parent_obj, VMXNET3State),
2437             VMSTATE_BOOL(rx_packets_compound, VMXNET3State),
2438             VMSTATE_BOOL(rx_vlan_stripping, VMXNET3State),
2439             VMSTATE_BOOL(lro_supported, VMXNET3State),
2440             VMSTATE_UINT32(rx_mode, VMXNET3State),
2441             VMSTATE_UINT32(mcast_list_len, VMXNET3State),
2442             VMSTATE_UINT32(mcast_list_buff_size, VMXNET3State),
2443             VMSTATE_UINT32_ARRAY(vlan_table, VMXNET3State, VMXNET3_VFT_SIZE),
2444             VMSTATE_UINT32(mtu, VMXNET3State),
2445             VMSTATE_UINT16(max_rx_frags, VMXNET3State),
2446             VMSTATE_UINT32(max_tx_frags, VMXNET3State),
2447             VMSTATE_UINT8(event_int_idx, VMXNET3State),
2448             VMSTATE_BOOL(auto_int_masking, VMXNET3State),
2449             VMSTATE_UINT8(txq_num, VMXNET3State),
2450             VMSTATE_UINT8(rxq_num, VMXNET3State),
2451             VMSTATE_UINT32(device_active, VMXNET3State),
2452             VMSTATE_UINT32(last_command, VMXNET3State),
2453             VMSTATE_UINT32(link_status_and_speed, VMXNET3State),
2454             VMSTATE_UINT32(temp_mac, VMXNET3State),
2455             VMSTATE_UINT64(drv_shmem, VMXNET3State),
2456             VMSTATE_UINT64(temp_shared_guest_driver_memory, VMXNET3State),
2457 
2458             VMSTATE_STRUCT_ARRAY(txq_descr, VMXNET3State,
2459                 VMXNET3_DEVICE_MAX_TX_QUEUES, 0, vmstate_vmxnet3_txq_descr,
2460                 Vmxnet3TxqDescr),
2461             VMSTATE_STRUCT_ARRAY(rxq_descr, VMXNET3State,
2462                 VMXNET3_DEVICE_MAX_RX_QUEUES, 0, vmstate_vmxnet3_rxq_descr,
2463                 Vmxnet3RxqDescr),
2464             VMSTATE_STRUCT_ARRAY(interrupt_states, VMXNET3State,
2465                 VMXNET3_MAX_INTRS, 0, vmstate_vmxnet3_int_state,
2466                 Vmxnet3IntState),
2467 
2468             VMSTATE_END_OF_LIST()
2469     },
2470     .subsections = (const VMStateDescription*[]) {
2471         &vmxstate_vmxnet3_mcast_list,
2472         NULL
2473     }
2474 };
2475 
2476 static Property vmxnet3_properties[] = {
2477     DEFINE_NIC_PROPERTIES(VMXNET3State, conf),
2478     DEFINE_PROP_BIT("x-old-msi-offsets", VMXNET3State, compat_flags,
2479                     VMXNET3_COMPAT_FLAG_OLD_MSI_OFFSETS_BIT, false),
2480     DEFINE_PROP_BIT("x-disable-pcie", VMXNET3State, compat_flags,
2481                     VMXNET3_COMPAT_FLAG_DISABLE_PCIE_BIT, false),
2482     DEFINE_PROP_END_OF_LIST(),
2483 };
2484 
2485 static void vmxnet3_realize(DeviceState *qdev, Error **errp)
2486 {
2487     VMXNET3Class *vc = VMXNET3_DEVICE_GET_CLASS(qdev);
2488     PCIDevice *pci_dev = PCI_DEVICE(qdev);
2489     VMXNET3State *s = VMXNET3(qdev);
2490 
2491     if (!(s->compat_flags & VMXNET3_COMPAT_FLAG_DISABLE_PCIE)) {
2492         pci_dev->cap_present |= QEMU_PCI_CAP_EXPRESS;
2493     }
2494 
2495     vc->parent_dc_realize(qdev, errp);
2496 }
2497 
2498 static void vmxnet3_class_init(ObjectClass *class, void *data)
2499 {
2500     DeviceClass *dc = DEVICE_CLASS(class);
2501     PCIDeviceClass *c = PCI_DEVICE_CLASS(class);
2502     VMXNET3Class *vc = VMXNET3_DEVICE_CLASS(class);
2503 
2504     c->realize = vmxnet3_pci_realize;
2505     c->exit = vmxnet3_pci_uninit;
2506     c->vendor_id = PCI_VENDOR_ID_VMWARE;
2507     c->device_id = PCI_DEVICE_ID_VMWARE_VMXNET3;
2508     c->revision = PCI_DEVICE_ID_VMWARE_VMXNET3_REVISION;
2509     c->romfile = "efi-vmxnet3.rom";
2510     c->class_id = PCI_CLASS_NETWORK_ETHERNET;
2511     c->subsystem_vendor_id = PCI_VENDOR_ID_VMWARE;
2512     c->subsystem_id = PCI_DEVICE_ID_VMWARE_VMXNET3;
2513     device_class_set_parent_realize(dc, vmxnet3_realize,
2514                                     &vc->parent_dc_realize);
2515     dc->desc = "VMWare Paravirtualized Ethernet v3";
2516     dc->reset = vmxnet3_qdev_reset;
2517     dc->vmsd = &vmstate_vmxnet3;
2518     device_class_set_props(dc, vmxnet3_properties);
2519     set_bit(DEVICE_CATEGORY_NETWORK, dc->categories);
2520 }
2521 
2522 static const TypeInfo vmxnet3_info = {
2523     .name          = TYPE_VMXNET3,
2524     .parent        = TYPE_PCI_DEVICE,
2525     .class_size    = sizeof(VMXNET3Class),
2526     .instance_size = sizeof(VMXNET3State),
2527     .class_init    = vmxnet3_class_init,
2528     .instance_init = vmxnet3_instance_init,
2529     .interfaces = (InterfaceInfo[]) {
2530         { INTERFACE_PCIE_DEVICE },
2531         { INTERFACE_CONVENTIONAL_PCI_DEVICE },
2532         { }
2533     },
2534 };
2535 
2536 static void vmxnet3_register_types(void)
2537 {
2538     VMW_CBPRN("vmxnet3_register_types called...");
2539     type_register_static(&vmxnet3_info);
2540 }
2541 
2542 type_init(vmxnet3_register_types)
2543