xref: /openbmc/qemu/hw/net/vmxnet3.c (revision 80adf54e)
1 /*
2  * QEMU VMWARE VMXNET3 paravirtual NIC
3  *
4  * Copyright (c) 2012 Ravello Systems LTD (http://ravellosystems.com)
5  *
6  * Developed by Daynix Computing LTD (http://www.daynix.com)
7  *
8  * Authors:
9  * Dmitry Fleytman <dmitry@daynix.com>
10  * Tamir Shomer <tamirs@daynix.com>
11  * Yan Vugenfirer <yan@daynix.com>
12  *
13  * This work is licensed under the terms of the GNU GPL, version 2.
14  * See the COPYING file in the top-level directory.
15  *
16  */
17 
18 #include "qemu/osdep.h"
19 #include "hw/hw.h"
20 #include "hw/pci/pci.h"
21 #include "net/net.h"
22 #include "net/tap.h"
23 #include "net/checksum.h"
24 #include "sysemu/sysemu.h"
25 #include "qemu-common.h"
26 #include "qemu/bswap.h"
27 #include "hw/pci/msix.h"
28 #include "hw/pci/msi.h"
29 #include "migration/register.h"
30 
31 #include "vmxnet3.h"
32 #include "vmxnet_debug.h"
33 #include "vmware_utils.h"
34 #include "net_tx_pkt.h"
35 #include "net_rx_pkt.h"
36 
37 #define PCI_DEVICE_ID_VMWARE_VMXNET3_REVISION 0x1
38 #define VMXNET3_MSIX_BAR_SIZE 0x2000
39 #define MIN_BUF_SIZE 60
40 
41 /* Compatibility flags for migration */
42 #define VMXNET3_COMPAT_FLAG_OLD_MSI_OFFSETS_BIT 0
43 #define VMXNET3_COMPAT_FLAG_OLD_MSI_OFFSETS \
44     (1 << VMXNET3_COMPAT_FLAG_OLD_MSI_OFFSETS_BIT)
45 #define VMXNET3_COMPAT_FLAG_DISABLE_PCIE_BIT 1
46 #define VMXNET3_COMPAT_FLAG_DISABLE_PCIE \
47     (1 << VMXNET3_COMPAT_FLAG_DISABLE_PCIE_BIT)
48 
49 #define VMXNET3_EXP_EP_OFFSET (0x48)
50 #define VMXNET3_MSI_OFFSET(s) \
51     ((s)->compat_flags & VMXNET3_COMPAT_FLAG_OLD_MSI_OFFSETS ? 0x50 : 0x84)
52 #define VMXNET3_MSIX_OFFSET(s) \
53     ((s)->compat_flags & VMXNET3_COMPAT_FLAG_OLD_MSI_OFFSETS ? 0 : 0x9c)
54 #define VMXNET3_DSN_OFFSET     (0x100)
55 
56 #define VMXNET3_BAR0_IDX      (0)
57 #define VMXNET3_BAR1_IDX      (1)
58 #define VMXNET3_MSIX_BAR_IDX  (2)
59 
60 #define VMXNET3_OFF_MSIX_TABLE (0x000)
61 #define VMXNET3_OFF_MSIX_PBA(s) \
62     ((s)->compat_flags & VMXNET3_COMPAT_FLAG_OLD_MSI_OFFSETS ? 0x800 : 0x1000)
63 
64 /* Link speed in Mbps should be shifted by 16 */
65 #define VMXNET3_LINK_SPEED      (1000 << 16)
66 
67 /* Link status: 1 - up, 0 - down. */
68 #define VMXNET3_LINK_STATUS_UP  0x1
69 
70 /* Least significant bit should be set for revision and version */
71 #define VMXNET3_UPT_REVISION      0x1
72 #define VMXNET3_DEVICE_REVISION   0x1
73 
74 /* Number of interrupt vectors for non-MSIx modes */
75 #define VMXNET3_MAX_NMSIX_INTRS   (1)
76 
77 /* Macros for rings descriptors access */
78 #define VMXNET3_READ_TX_QUEUE_DESCR8(_d, dpa, field) \
79     (vmw_shmem_ld8(_d, dpa + offsetof(struct Vmxnet3_TxQueueDesc, field)))
80 
81 #define VMXNET3_WRITE_TX_QUEUE_DESCR8(_d, dpa, field, value) \
82     (vmw_shmem_st8(_d, dpa + offsetof(struct Vmxnet3_TxQueueDesc, field, value)))
83 
84 #define VMXNET3_READ_TX_QUEUE_DESCR32(_d, dpa, field) \
85     (vmw_shmem_ld32(_d, dpa + offsetof(struct Vmxnet3_TxQueueDesc, field)))
86 
87 #define VMXNET3_WRITE_TX_QUEUE_DESCR32(_d, dpa, field, value) \
88     (vmw_shmem_st32(_d, dpa + offsetof(struct Vmxnet3_TxQueueDesc, field), value))
89 
90 #define VMXNET3_READ_TX_QUEUE_DESCR64(_d, dpa, field) \
91     (vmw_shmem_ld64(_d, dpa + offsetof(struct Vmxnet3_TxQueueDesc, field)))
92 
93 #define VMXNET3_WRITE_TX_QUEUE_DESCR64(_d, dpa, field, value) \
94     (vmw_shmem_st64(_d, dpa + offsetof(struct Vmxnet3_TxQueueDesc, field), value))
95 
96 #define VMXNET3_READ_RX_QUEUE_DESCR64(_d, dpa, field) \
97     (vmw_shmem_ld64(_d, dpa + offsetof(struct Vmxnet3_RxQueueDesc, field)))
98 
99 #define VMXNET3_READ_RX_QUEUE_DESCR32(_d, dpa, field) \
100     (vmw_shmem_ld32(_d, dpa + offsetof(struct Vmxnet3_RxQueueDesc, field)))
101 
102 #define VMXNET3_WRITE_RX_QUEUE_DESCR64(_d, dpa, field, value) \
103     (vmw_shmem_st64(_d, dpa + offsetof(struct Vmxnet3_RxQueueDesc, field), value))
104 
105 #define VMXNET3_WRITE_RX_QUEUE_DESCR8(_d, dpa, field, value) \
106     (vmw_shmem_st8(_d, dpa + offsetof(struct Vmxnet3_RxQueueDesc, field), value))
107 
108 /* Macros for guest driver shared area access */
109 #define VMXNET3_READ_DRV_SHARED64(_d, shpa, field) \
110     (vmw_shmem_ld64(_d, shpa + offsetof(struct Vmxnet3_DriverShared, field)))
111 
112 #define VMXNET3_READ_DRV_SHARED32(_d, shpa, field) \
113     (vmw_shmem_ld32(_d, shpa + offsetof(struct Vmxnet3_DriverShared, field)))
114 
115 #define VMXNET3_WRITE_DRV_SHARED32(_d, shpa, field, val) \
116     (vmw_shmem_st32(_d, shpa + offsetof(struct Vmxnet3_DriverShared, field), val))
117 
118 #define VMXNET3_READ_DRV_SHARED16(_d, shpa, field) \
119     (vmw_shmem_ld16(_d, shpa + offsetof(struct Vmxnet3_DriverShared, field)))
120 
121 #define VMXNET3_READ_DRV_SHARED8(_d, shpa, field) \
122     (vmw_shmem_ld8(_d, shpa + offsetof(struct Vmxnet3_DriverShared, field)))
123 
124 #define VMXNET3_READ_DRV_SHARED(_d, shpa, field, b, l) \
125     (vmw_shmem_read(_d, shpa + offsetof(struct Vmxnet3_DriverShared, field), b, l))
126 
127 #define VMXNET_FLAG_IS_SET(field, flag) (((field) & (flag)) == (flag))
128 
129 typedef struct VMXNET3Class {
130     PCIDeviceClass parent_class;
131     DeviceRealize parent_dc_realize;
132 } VMXNET3Class;
133 
134 #define TYPE_VMXNET3 "vmxnet3"
135 #define VMXNET3(obj) OBJECT_CHECK(VMXNET3State, (obj), TYPE_VMXNET3)
136 
137 #define VMXNET3_DEVICE_CLASS(klass) \
138     OBJECT_CLASS_CHECK(VMXNET3Class, (klass), TYPE_VMXNET3)
139 #define VMXNET3_DEVICE_GET_CLASS(obj) \
140     OBJECT_GET_CLASS(VMXNET3Class, (obj), TYPE_VMXNET3)
141 
142 /* Cyclic ring abstraction */
143 typedef struct {
144     hwaddr pa;
145     uint32_t size;
146     uint32_t cell_size;
147     uint32_t next;
148     uint8_t gen;
149 } Vmxnet3Ring;
150 
151 static inline void vmxnet3_ring_init(PCIDevice *d,
152 				     Vmxnet3Ring *ring,
153                                      hwaddr pa,
154                                      uint32_t size,
155                                      uint32_t cell_size,
156                                      bool zero_region)
157 {
158     ring->pa = pa;
159     ring->size = size;
160     ring->cell_size = cell_size;
161     ring->gen = VMXNET3_INIT_GEN;
162     ring->next = 0;
163 
164     if (zero_region) {
165         vmw_shmem_set(d, pa, 0, size * cell_size);
166     }
167 }
168 
169 #define VMXNET3_RING_DUMP(macro, ring_name, ridx, r)                         \
170     macro("%s#%d: base %" PRIx64 " size %u cell_size %u gen %d next %u",  \
171           (ring_name), (ridx),                                               \
172           (r)->pa, (r)->size, (r)->cell_size, (r)->gen, (r)->next)
173 
174 static inline void vmxnet3_ring_inc(Vmxnet3Ring *ring)
175 {
176     if (++ring->next >= ring->size) {
177         ring->next = 0;
178         ring->gen ^= 1;
179     }
180 }
181 
182 static inline void vmxnet3_ring_dec(Vmxnet3Ring *ring)
183 {
184     if (ring->next-- == 0) {
185         ring->next = ring->size - 1;
186         ring->gen ^= 1;
187     }
188 }
189 
190 static inline hwaddr vmxnet3_ring_curr_cell_pa(Vmxnet3Ring *ring)
191 {
192     return ring->pa + ring->next * ring->cell_size;
193 }
194 
195 static inline void vmxnet3_ring_read_curr_cell(PCIDevice *d, Vmxnet3Ring *ring,
196 					       void *buff)
197 {
198     vmw_shmem_read(d, vmxnet3_ring_curr_cell_pa(ring), buff, ring->cell_size);
199 }
200 
201 static inline void vmxnet3_ring_write_curr_cell(PCIDevice *d, Vmxnet3Ring *ring,
202 						void *buff)
203 {
204     vmw_shmem_write(d, vmxnet3_ring_curr_cell_pa(ring), buff, ring->cell_size);
205 }
206 
207 static inline size_t vmxnet3_ring_curr_cell_idx(Vmxnet3Ring *ring)
208 {
209     return ring->next;
210 }
211 
212 static inline uint8_t vmxnet3_ring_curr_gen(Vmxnet3Ring *ring)
213 {
214     return ring->gen;
215 }
216 
217 /* Debug trace-related functions */
218 static inline void
219 vmxnet3_dump_tx_descr(struct Vmxnet3_TxDesc *descr)
220 {
221     VMW_PKPRN("TX DESCR: "
222               "addr %" PRIx64 ", len: %d, gen: %d, rsvd: %d, "
223               "dtype: %d, ext1: %d, msscof: %d, hlen: %d, om: %d, "
224               "eop: %d, cq: %d, ext2: %d, ti: %d, tci: %d",
225               le64_to_cpu(descr->addr), descr->len, descr->gen, descr->rsvd,
226               descr->dtype, descr->ext1, descr->msscof, descr->hlen, descr->om,
227               descr->eop, descr->cq, descr->ext2, descr->ti, descr->tci);
228 }
229 
230 static inline void
231 vmxnet3_dump_virt_hdr(struct virtio_net_hdr *vhdr)
232 {
233     VMW_PKPRN("VHDR: flags 0x%x, gso_type: 0x%x, hdr_len: %d, gso_size: %d, "
234               "csum_start: %d, csum_offset: %d",
235               vhdr->flags, vhdr->gso_type, vhdr->hdr_len, vhdr->gso_size,
236               vhdr->csum_start, vhdr->csum_offset);
237 }
238 
239 static inline void
240 vmxnet3_dump_rx_descr(struct Vmxnet3_RxDesc *descr)
241 {
242     VMW_PKPRN("RX DESCR: addr %" PRIx64 ", len: %d, gen: %d, rsvd: %d, "
243               "dtype: %d, ext1: %d, btype: %d",
244               le64_to_cpu(descr->addr), descr->len, descr->gen,
245               descr->rsvd, descr->dtype, descr->ext1, descr->btype);
246 }
247 
248 /* Device state and helper functions */
249 #define VMXNET3_RX_RINGS_PER_QUEUE (2)
250 
251 typedef struct {
252     Vmxnet3Ring tx_ring;
253     Vmxnet3Ring comp_ring;
254 
255     uint8_t intr_idx;
256     hwaddr tx_stats_pa;
257     struct UPT1_TxStats txq_stats;
258 } Vmxnet3TxqDescr;
259 
260 typedef struct {
261     Vmxnet3Ring rx_ring[VMXNET3_RX_RINGS_PER_QUEUE];
262     Vmxnet3Ring comp_ring;
263     uint8_t intr_idx;
264     hwaddr rx_stats_pa;
265     struct UPT1_RxStats rxq_stats;
266 } Vmxnet3RxqDescr;
267 
268 typedef struct {
269     bool is_masked;
270     bool is_pending;
271     bool is_asserted;
272 } Vmxnet3IntState;
273 
274 typedef struct {
275         PCIDevice parent_obj;
276         NICState *nic;
277         NICConf conf;
278         MemoryRegion bar0;
279         MemoryRegion bar1;
280         MemoryRegion msix_bar;
281 
282         Vmxnet3RxqDescr rxq_descr[VMXNET3_DEVICE_MAX_RX_QUEUES];
283         Vmxnet3TxqDescr txq_descr[VMXNET3_DEVICE_MAX_TX_QUEUES];
284 
285         /* Whether MSI-X support was installed successfully */
286         bool msix_used;
287         hwaddr drv_shmem;
288         hwaddr temp_shared_guest_driver_memory;
289 
290         uint8_t txq_num;
291 
292         /* This boolean tells whether RX packet being indicated has to */
293         /* be split into head and body chunks from different RX rings  */
294         bool rx_packets_compound;
295 
296         bool rx_vlan_stripping;
297         bool lro_supported;
298 
299         uint8_t rxq_num;
300 
301         /* Network MTU */
302         uint32_t mtu;
303 
304         /* Maximum number of fragments for indicated TX packets */
305         uint32_t max_tx_frags;
306 
307         /* Maximum number of fragments for indicated RX packets */
308         uint16_t max_rx_frags;
309 
310         /* Index for events interrupt */
311         uint8_t event_int_idx;
312 
313         /* Whether automatic interrupts masking enabled */
314         bool auto_int_masking;
315 
316         bool peer_has_vhdr;
317 
318         /* TX packets to QEMU interface */
319         struct NetTxPkt *tx_pkt;
320         uint32_t offload_mode;
321         uint32_t cso_or_gso_size;
322         uint16_t tci;
323         bool needs_vlan;
324 
325         struct NetRxPkt *rx_pkt;
326 
327         bool tx_sop;
328         bool skip_current_tx_pkt;
329 
330         uint32_t device_active;
331         uint32_t last_command;
332 
333         uint32_t link_status_and_speed;
334 
335         Vmxnet3IntState interrupt_states[VMXNET3_MAX_INTRS];
336 
337         uint32_t temp_mac;   /* To store the low part first */
338 
339         MACAddr perm_mac;
340         uint32_t vlan_table[VMXNET3_VFT_SIZE];
341         uint32_t rx_mode;
342         MACAddr *mcast_list;
343         uint32_t mcast_list_len;
344         uint32_t mcast_list_buff_size; /* needed for live migration. */
345 
346         /* Compatibility flags for migration */
347         uint32_t compat_flags;
348 } VMXNET3State;
349 
350 /* Interrupt management */
351 
352 /*
353  * This function returns sign whether interrupt line is in asserted state
354  * This depends on the type of interrupt used. For INTX interrupt line will
355  * be asserted until explicit deassertion, for MSI(X) interrupt line will
356  * be deasserted automatically due to notification semantics of the MSI(X)
357  * interrupts
358  */
359 static bool _vmxnet3_assert_interrupt_line(VMXNET3State *s, uint32_t int_idx)
360 {
361     PCIDevice *d = PCI_DEVICE(s);
362 
363     if (s->msix_used && msix_enabled(d)) {
364         VMW_IRPRN("Sending MSI-X notification for vector %u", int_idx);
365         msix_notify(d, int_idx);
366         return false;
367     }
368     if (msi_enabled(d)) {
369         VMW_IRPRN("Sending MSI notification for vector %u", int_idx);
370         msi_notify(d, int_idx);
371         return false;
372     }
373 
374     VMW_IRPRN("Asserting line for interrupt %u", int_idx);
375     pci_irq_assert(d);
376     return true;
377 }
378 
379 static void _vmxnet3_deassert_interrupt_line(VMXNET3State *s, int lidx)
380 {
381     PCIDevice *d = PCI_DEVICE(s);
382 
383     /*
384      * This function should never be called for MSI(X) interrupts
385      * because deassertion never required for message interrupts
386      */
387     assert(!s->msix_used || !msix_enabled(d));
388     /*
389      * This function should never be called for MSI(X) interrupts
390      * because deassertion never required for message interrupts
391      */
392     assert(!msi_enabled(d));
393 
394     VMW_IRPRN("Deasserting line for interrupt %u", lidx);
395     pci_irq_deassert(d);
396 }
397 
398 static void vmxnet3_update_interrupt_line_state(VMXNET3State *s, int lidx)
399 {
400     if (!s->interrupt_states[lidx].is_pending &&
401        s->interrupt_states[lidx].is_asserted) {
402         VMW_IRPRN("New interrupt line state for index %d is DOWN", lidx);
403         _vmxnet3_deassert_interrupt_line(s, lidx);
404         s->interrupt_states[lidx].is_asserted = false;
405         return;
406     }
407 
408     if (s->interrupt_states[lidx].is_pending &&
409        !s->interrupt_states[lidx].is_masked &&
410        !s->interrupt_states[lidx].is_asserted) {
411         VMW_IRPRN("New interrupt line state for index %d is UP", lidx);
412         s->interrupt_states[lidx].is_asserted =
413             _vmxnet3_assert_interrupt_line(s, lidx);
414         s->interrupt_states[lidx].is_pending = false;
415         return;
416     }
417 }
418 
419 static void vmxnet3_trigger_interrupt(VMXNET3State *s, int lidx)
420 {
421     PCIDevice *d = PCI_DEVICE(s);
422     s->interrupt_states[lidx].is_pending = true;
423     vmxnet3_update_interrupt_line_state(s, lidx);
424 
425     if (s->msix_used && msix_enabled(d) && s->auto_int_masking) {
426         goto do_automask;
427     }
428 
429     if (msi_enabled(d) && s->auto_int_masking) {
430         goto do_automask;
431     }
432 
433     return;
434 
435 do_automask:
436     s->interrupt_states[lidx].is_masked = true;
437     vmxnet3_update_interrupt_line_state(s, lidx);
438 }
439 
440 static bool vmxnet3_interrupt_asserted(VMXNET3State *s, int lidx)
441 {
442     return s->interrupt_states[lidx].is_asserted;
443 }
444 
445 static void vmxnet3_clear_interrupt(VMXNET3State *s, int int_idx)
446 {
447     s->interrupt_states[int_idx].is_pending = false;
448     if (s->auto_int_masking) {
449         s->interrupt_states[int_idx].is_masked = true;
450     }
451     vmxnet3_update_interrupt_line_state(s, int_idx);
452 }
453 
454 static void
455 vmxnet3_on_interrupt_mask_changed(VMXNET3State *s, int lidx, bool is_masked)
456 {
457     s->interrupt_states[lidx].is_masked = is_masked;
458     vmxnet3_update_interrupt_line_state(s, lidx);
459 }
460 
461 static bool vmxnet3_verify_driver_magic(PCIDevice *d, hwaddr dshmem)
462 {
463     return (VMXNET3_READ_DRV_SHARED32(d, dshmem, magic) == VMXNET3_REV1_MAGIC);
464 }
465 
466 #define VMXNET3_GET_BYTE(x, byte_num) (((x) >> (byte_num)*8) & 0xFF)
467 #define VMXNET3_MAKE_BYTE(byte_num, val) \
468     (((uint32_t)((val) & 0xFF)) << (byte_num)*8)
469 
470 static void vmxnet3_set_variable_mac(VMXNET3State *s, uint32_t h, uint32_t l)
471 {
472     s->conf.macaddr.a[0] = VMXNET3_GET_BYTE(l,  0);
473     s->conf.macaddr.a[1] = VMXNET3_GET_BYTE(l,  1);
474     s->conf.macaddr.a[2] = VMXNET3_GET_BYTE(l,  2);
475     s->conf.macaddr.a[3] = VMXNET3_GET_BYTE(l,  3);
476     s->conf.macaddr.a[4] = VMXNET3_GET_BYTE(h, 0);
477     s->conf.macaddr.a[5] = VMXNET3_GET_BYTE(h, 1);
478 
479     VMW_CFPRN("Variable MAC: " MAC_FMT, MAC_ARG(s->conf.macaddr.a));
480 
481     qemu_format_nic_info_str(qemu_get_queue(s->nic), s->conf.macaddr.a);
482 }
483 
484 static uint64_t vmxnet3_get_mac_low(MACAddr *addr)
485 {
486     return VMXNET3_MAKE_BYTE(0, addr->a[0]) |
487            VMXNET3_MAKE_BYTE(1, addr->a[1]) |
488            VMXNET3_MAKE_BYTE(2, addr->a[2]) |
489            VMXNET3_MAKE_BYTE(3, addr->a[3]);
490 }
491 
492 static uint64_t vmxnet3_get_mac_high(MACAddr *addr)
493 {
494     return VMXNET3_MAKE_BYTE(0, addr->a[4]) |
495            VMXNET3_MAKE_BYTE(1, addr->a[5]);
496 }
497 
498 static void
499 vmxnet3_inc_tx_consumption_counter(VMXNET3State *s, int qidx)
500 {
501     vmxnet3_ring_inc(&s->txq_descr[qidx].tx_ring);
502 }
503 
504 static inline void
505 vmxnet3_inc_rx_consumption_counter(VMXNET3State *s, int qidx, int ridx)
506 {
507     vmxnet3_ring_inc(&s->rxq_descr[qidx].rx_ring[ridx]);
508 }
509 
510 static inline void
511 vmxnet3_inc_tx_completion_counter(VMXNET3State *s, int qidx)
512 {
513     vmxnet3_ring_inc(&s->txq_descr[qidx].comp_ring);
514 }
515 
516 static void
517 vmxnet3_inc_rx_completion_counter(VMXNET3State *s, int qidx)
518 {
519     vmxnet3_ring_inc(&s->rxq_descr[qidx].comp_ring);
520 }
521 
522 static void
523 vmxnet3_dec_rx_completion_counter(VMXNET3State *s, int qidx)
524 {
525     vmxnet3_ring_dec(&s->rxq_descr[qidx].comp_ring);
526 }
527 
528 static void vmxnet3_complete_packet(VMXNET3State *s, int qidx, uint32_t tx_ridx)
529 {
530     struct Vmxnet3_TxCompDesc txcq_descr;
531     PCIDevice *d = PCI_DEVICE(s);
532 
533     VMXNET3_RING_DUMP(VMW_RIPRN, "TXC", qidx, &s->txq_descr[qidx].comp_ring);
534 
535     memset(&txcq_descr, 0, sizeof(txcq_descr));
536     txcq_descr.txdIdx = tx_ridx;
537     txcq_descr.gen = vmxnet3_ring_curr_gen(&s->txq_descr[qidx].comp_ring);
538 
539     vmxnet3_ring_write_curr_cell(d, &s->txq_descr[qidx].comp_ring, &txcq_descr);
540 
541     /* Flush changes in TX descriptor before changing the counter value */
542     smp_wmb();
543 
544     vmxnet3_inc_tx_completion_counter(s, qidx);
545     vmxnet3_trigger_interrupt(s, s->txq_descr[qidx].intr_idx);
546 }
547 
548 static bool
549 vmxnet3_setup_tx_offloads(VMXNET3State *s)
550 {
551     switch (s->offload_mode) {
552     case VMXNET3_OM_NONE:
553         net_tx_pkt_build_vheader(s->tx_pkt, false, false, 0);
554         break;
555 
556     case VMXNET3_OM_CSUM:
557         net_tx_pkt_build_vheader(s->tx_pkt, false, true, 0);
558         VMW_PKPRN("L4 CSO requested\n");
559         break;
560 
561     case VMXNET3_OM_TSO:
562         net_tx_pkt_build_vheader(s->tx_pkt, true, true,
563             s->cso_or_gso_size);
564         net_tx_pkt_update_ip_checksums(s->tx_pkt);
565         VMW_PKPRN("GSO offload requested.");
566         break;
567 
568     default:
569         g_assert_not_reached();
570         return false;
571     }
572 
573     return true;
574 }
575 
576 static void
577 vmxnet3_tx_retrieve_metadata(VMXNET3State *s,
578                              const struct Vmxnet3_TxDesc *txd)
579 {
580     s->offload_mode = txd->om;
581     s->cso_or_gso_size = txd->msscof;
582     s->tci = txd->tci;
583     s->needs_vlan = txd->ti;
584 }
585 
586 typedef enum {
587     VMXNET3_PKT_STATUS_OK,
588     VMXNET3_PKT_STATUS_ERROR,
589     VMXNET3_PKT_STATUS_DISCARD,/* only for tx */
590     VMXNET3_PKT_STATUS_OUT_OF_BUF /* only for rx */
591 } Vmxnet3PktStatus;
592 
593 static void
594 vmxnet3_on_tx_done_update_stats(VMXNET3State *s, int qidx,
595     Vmxnet3PktStatus status)
596 {
597     size_t tot_len = net_tx_pkt_get_total_len(s->tx_pkt);
598     struct UPT1_TxStats *stats = &s->txq_descr[qidx].txq_stats;
599 
600     switch (status) {
601     case VMXNET3_PKT_STATUS_OK:
602         switch (net_tx_pkt_get_packet_type(s->tx_pkt)) {
603         case ETH_PKT_BCAST:
604             stats->bcastPktsTxOK++;
605             stats->bcastBytesTxOK += tot_len;
606             break;
607         case ETH_PKT_MCAST:
608             stats->mcastPktsTxOK++;
609             stats->mcastBytesTxOK += tot_len;
610             break;
611         case ETH_PKT_UCAST:
612             stats->ucastPktsTxOK++;
613             stats->ucastBytesTxOK += tot_len;
614             break;
615         default:
616             g_assert_not_reached();
617         }
618 
619         if (s->offload_mode == VMXNET3_OM_TSO) {
620             /*
621              * According to VMWARE headers this statistic is a number
622              * of packets after segmentation but since we don't have
623              * this information in QEMU model, the best we can do is to
624              * provide number of non-segmented packets
625              */
626             stats->TSOPktsTxOK++;
627             stats->TSOBytesTxOK += tot_len;
628         }
629         break;
630 
631     case VMXNET3_PKT_STATUS_DISCARD:
632         stats->pktsTxDiscard++;
633         break;
634 
635     case VMXNET3_PKT_STATUS_ERROR:
636         stats->pktsTxError++;
637         break;
638 
639     default:
640         g_assert_not_reached();
641     }
642 }
643 
644 static void
645 vmxnet3_on_rx_done_update_stats(VMXNET3State *s,
646                                 int qidx,
647                                 Vmxnet3PktStatus status)
648 {
649     struct UPT1_RxStats *stats = &s->rxq_descr[qidx].rxq_stats;
650     size_t tot_len = net_rx_pkt_get_total_len(s->rx_pkt);
651 
652     switch (status) {
653     case VMXNET3_PKT_STATUS_OUT_OF_BUF:
654         stats->pktsRxOutOfBuf++;
655         break;
656 
657     case VMXNET3_PKT_STATUS_ERROR:
658         stats->pktsRxError++;
659         break;
660     case VMXNET3_PKT_STATUS_OK:
661         switch (net_rx_pkt_get_packet_type(s->rx_pkt)) {
662         case ETH_PKT_BCAST:
663             stats->bcastPktsRxOK++;
664             stats->bcastBytesRxOK += tot_len;
665             break;
666         case ETH_PKT_MCAST:
667             stats->mcastPktsRxOK++;
668             stats->mcastBytesRxOK += tot_len;
669             break;
670         case ETH_PKT_UCAST:
671             stats->ucastPktsRxOK++;
672             stats->ucastBytesRxOK += tot_len;
673             break;
674         default:
675             g_assert_not_reached();
676         }
677 
678         if (tot_len > s->mtu) {
679             stats->LROPktsRxOK++;
680             stats->LROBytesRxOK += tot_len;
681         }
682         break;
683     default:
684         g_assert_not_reached();
685     }
686 }
687 
688 static inline bool
689 vmxnet3_pop_next_tx_descr(VMXNET3State *s,
690                           int qidx,
691                           struct Vmxnet3_TxDesc *txd,
692                           uint32_t *descr_idx)
693 {
694     Vmxnet3Ring *ring = &s->txq_descr[qidx].tx_ring;
695     PCIDevice *d = PCI_DEVICE(s);
696 
697     vmxnet3_ring_read_curr_cell(d, ring, txd);
698     if (txd->gen == vmxnet3_ring_curr_gen(ring)) {
699         /* Only read after generation field verification */
700         smp_rmb();
701         /* Re-read to be sure we got the latest version */
702         vmxnet3_ring_read_curr_cell(d, ring, txd);
703         VMXNET3_RING_DUMP(VMW_RIPRN, "TX", qidx, ring);
704         *descr_idx = vmxnet3_ring_curr_cell_idx(ring);
705         vmxnet3_inc_tx_consumption_counter(s, qidx);
706         return true;
707     }
708 
709     return false;
710 }
711 
712 static bool
713 vmxnet3_send_packet(VMXNET3State *s, uint32_t qidx)
714 {
715     Vmxnet3PktStatus status = VMXNET3_PKT_STATUS_OK;
716 
717     if (!vmxnet3_setup_tx_offloads(s)) {
718         status = VMXNET3_PKT_STATUS_ERROR;
719         goto func_exit;
720     }
721 
722     /* debug prints */
723     vmxnet3_dump_virt_hdr(net_tx_pkt_get_vhdr(s->tx_pkt));
724     net_tx_pkt_dump(s->tx_pkt);
725 
726     if (!net_tx_pkt_send(s->tx_pkt, qemu_get_queue(s->nic))) {
727         status = VMXNET3_PKT_STATUS_DISCARD;
728         goto func_exit;
729     }
730 
731 func_exit:
732     vmxnet3_on_tx_done_update_stats(s, qidx, status);
733     return (status == VMXNET3_PKT_STATUS_OK);
734 }
735 
736 static void vmxnet3_process_tx_queue(VMXNET3State *s, int qidx)
737 {
738     struct Vmxnet3_TxDesc txd;
739     uint32_t txd_idx;
740     uint32_t data_len;
741     hwaddr data_pa;
742 
743     for (;;) {
744         if (!vmxnet3_pop_next_tx_descr(s, qidx, &txd, &txd_idx)) {
745             break;
746         }
747 
748         vmxnet3_dump_tx_descr(&txd);
749 
750         if (!s->skip_current_tx_pkt) {
751             data_len = (txd.len > 0) ? txd.len : VMXNET3_MAX_TX_BUF_SIZE;
752             data_pa = le64_to_cpu(txd.addr);
753 
754             if (!net_tx_pkt_add_raw_fragment(s->tx_pkt,
755                                                 data_pa,
756                                                 data_len)) {
757                 s->skip_current_tx_pkt = true;
758             }
759         }
760 
761         if (s->tx_sop) {
762             vmxnet3_tx_retrieve_metadata(s, &txd);
763             s->tx_sop = false;
764         }
765 
766         if (txd.eop) {
767             if (!s->skip_current_tx_pkt && net_tx_pkt_parse(s->tx_pkt)) {
768                 if (s->needs_vlan) {
769                     net_tx_pkt_setup_vlan_header(s->tx_pkt, s->tci);
770                 }
771 
772                 vmxnet3_send_packet(s, qidx);
773             } else {
774                 vmxnet3_on_tx_done_update_stats(s, qidx,
775                                                 VMXNET3_PKT_STATUS_ERROR);
776             }
777 
778             vmxnet3_complete_packet(s, qidx, txd_idx);
779             s->tx_sop = true;
780             s->skip_current_tx_pkt = false;
781             net_tx_pkt_reset(s->tx_pkt);
782         }
783     }
784 }
785 
786 static inline void
787 vmxnet3_read_next_rx_descr(VMXNET3State *s, int qidx, int ridx,
788                            struct Vmxnet3_RxDesc *dbuf, uint32_t *didx)
789 {
790     PCIDevice *d = PCI_DEVICE(s);
791 
792     Vmxnet3Ring *ring = &s->rxq_descr[qidx].rx_ring[ridx];
793     *didx = vmxnet3_ring_curr_cell_idx(ring);
794     vmxnet3_ring_read_curr_cell(d, ring, dbuf);
795 }
796 
797 static inline uint8_t
798 vmxnet3_get_rx_ring_gen(VMXNET3State *s, int qidx, int ridx)
799 {
800     return s->rxq_descr[qidx].rx_ring[ridx].gen;
801 }
802 
803 static inline hwaddr
804 vmxnet3_pop_rxc_descr(VMXNET3State *s, int qidx, uint32_t *descr_gen)
805 {
806     uint8_t ring_gen;
807     struct Vmxnet3_RxCompDesc rxcd;
808 
809     hwaddr daddr =
810         vmxnet3_ring_curr_cell_pa(&s->rxq_descr[qidx].comp_ring);
811 
812     pci_dma_read(PCI_DEVICE(s),
813                  daddr, &rxcd, sizeof(struct Vmxnet3_RxCompDesc));
814     ring_gen = vmxnet3_ring_curr_gen(&s->rxq_descr[qidx].comp_ring);
815 
816     if (rxcd.gen != ring_gen) {
817         *descr_gen = ring_gen;
818         vmxnet3_inc_rx_completion_counter(s, qidx);
819         return daddr;
820     }
821 
822     return 0;
823 }
824 
825 static inline void
826 vmxnet3_revert_rxc_descr(VMXNET3State *s, int qidx)
827 {
828     vmxnet3_dec_rx_completion_counter(s, qidx);
829 }
830 
831 #define RXQ_IDX      (0)
832 #define RX_HEAD_BODY_RING (0)
833 #define RX_BODY_ONLY_RING (1)
834 
835 static bool
836 vmxnet3_get_next_head_rx_descr(VMXNET3State *s,
837                                struct Vmxnet3_RxDesc *descr_buf,
838                                uint32_t *descr_idx,
839                                uint32_t *ridx)
840 {
841     for (;;) {
842         uint32_t ring_gen;
843         vmxnet3_read_next_rx_descr(s, RXQ_IDX, RX_HEAD_BODY_RING,
844                                    descr_buf, descr_idx);
845 
846         /* If no more free descriptors - return */
847         ring_gen = vmxnet3_get_rx_ring_gen(s, RXQ_IDX, RX_HEAD_BODY_RING);
848         if (descr_buf->gen != ring_gen) {
849             return false;
850         }
851 
852         /* Only read after generation field verification */
853         smp_rmb();
854         /* Re-read to be sure we got the latest version */
855         vmxnet3_read_next_rx_descr(s, RXQ_IDX, RX_HEAD_BODY_RING,
856                                    descr_buf, descr_idx);
857 
858         /* Mark current descriptor as used/skipped */
859         vmxnet3_inc_rx_consumption_counter(s, RXQ_IDX, RX_HEAD_BODY_RING);
860 
861         /* If this is what we are looking for - return */
862         if (descr_buf->btype == VMXNET3_RXD_BTYPE_HEAD) {
863             *ridx = RX_HEAD_BODY_RING;
864             return true;
865         }
866     }
867 }
868 
869 static bool
870 vmxnet3_get_next_body_rx_descr(VMXNET3State *s,
871                                struct Vmxnet3_RxDesc *d,
872                                uint32_t *didx,
873                                uint32_t *ridx)
874 {
875     vmxnet3_read_next_rx_descr(s, RXQ_IDX, RX_HEAD_BODY_RING, d, didx);
876 
877     /* Try to find corresponding descriptor in head/body ring */
878     if (d->gen == vmxnet3_get_rx_ring_gen(s, RXQ_IDX, RX_HEAD_BODY_RING)) {
879         /* Only read after generation field verification */
880         smp_rmb();
881         /* Re-read to be sure we got the latest version */
882         vmxnet3_read_next_rx_descr(s, RXQ_IDX, RX_HEAD_BODY_RING, d, didx);
883         if (d->btype == VMXNET3_RXD_BTYPE_BODY) {
884             vmxnet3_inc_rx_consumption_counter(s, RXQ_IDX, RX_HEAD_BODY_RING);
885             *ridx = RX_HEAD_BODY_RING;
886             return true;
887         }
888     }
889 
890     /*
891      * If there is no free descriptors on head/body ring or next free
892      * descriptor is a head descriptor switch to body only ring
893      */
894     vmxnet3_read_next_rx_descr(s, RXQ_IDX, RX_BODY_ONLY_RING, d, didx);
895 
896     /* If no more free descriptors - return */
897     if (d->gen == vmxnet3_get_rx_ring_gen(s, RXQ_IDX, RX_BODY_ONLY_RING)) {
898         /* Only read after generation field verification */
899         smp_rmb();
900         /* Re-read to be sure we got the latest version */
901         vmxnet3_read_next_rx_descr(s, RXQ_IDX, RX_BODY_ONLY_RING, d, didx);
902         assert(d->btype == VMXNET3_RXD_BTYPE_BODY);
903         *ridx = RX_BODY_ONLY_RING;
904         vmxnet3_inc_rx_consumption_counter(s, RXQ_IDX, RX_BODY_ONLY_RING);
905         return true;
906     }
907 
908     return false;
909 }
910 
911 static inline bool
912 vmxnet3_get_next_rx_descr(VMXNET3State *s, bool is_head,
913                           struct Vmxnet3_RxDesc *descr_buf,
914                           uint32_t *descr_idx,
915                           uint32_t *ridx)
916 {
917     if (is_head || !s->rx_packets_compound) {
918         return vmxnet3_get_next_head_rx_descr(s, descr_buf, descr_idx, ridx);
919     } else {
920         return vmxnet3_get_next_body_rx_descr(s, descr_buf, descr_idx, ridx);
921     }
922 }
923 
924 /* In case packet was csum offloaded (either NEEDS_CSUM or DATA_VALID),
925  * the implementation always passes an RxCompDesc with a "Checksum
926  * calculated and found correct" to the OS (cnc=0 and tuc=1, see
927  * vmxnet3_rx_update_descr). This emulates the observed ESXi behavior.
928  *
929  * Therefore, if packet has the NEEDS_CSUM set, we must calculate
930  * and place a fully computed checksum into the tcp/udp header.
931  * Otherwise, the OS driver will receive a checksum-correct indication
932  * (CHECKSUM_UNNECESSARY), but with the actual tcp/udp checksum field
933  * having just the pseudo header csum value.
934  *
935  * While this is not a problem if packet is destined for local delivery,
936  * in the case the host OS performs forwarding, it will forward an
937  * incorrectly checksummed packet.
938  */
939 static void vmxnet3_rx_need_csum_calculate(struct NetRxPkt *pkt,
940                                            const void *pkt_data,
941                                            size_t pkt_len)
942 {
943     struct virtio_net_hdr *vhdr;
944     bool isip4, isip6, istcp, isudp;
945     uint8_t *data;
946     int len;
947 
948     if (!net_rx_pkt_has_virt_hdr(pkt)) {
949         return;
950     }
951 
952     vhdr = net_rx_pkt_get_vhdr(pkt);
953     if (!VMXNET_FLAG_IS_SET(vhdr->flags, VIRTIO_NET_HDR_F_NEEDS_CSUM)) {
954         return;
955     }
956 
957     net_rx_pkt_get_protocols(pkt, &isip4, &isip6, &isudp, &istcp);
958     if (!(isip4 || isip6) || !(istcp || isudp)) {
959         return;
960     }
961 
962     vmxnet3_dump_virt_hdr(vhdr);
963 
964     /* Validate packet len: csum_start + scum_offset + length of csum field */
965     if (pkt_len < (vhdr->csum_start + vhdr->csum_offset + 2)) {
966         VMW_PKPRN("packet len:%zu < csum_start(%d) + csum_offset(%d) + 2, "
967                   "cannot calculate checksum",
968                   pkt_len, vhdr->csum_start, vhdr->csum_offset);
969         return;
970     }
971 
972     data = (uint8_t *)pkt_data + vhdr->csum_start;
973     len = pkt_len - vhdr->csum_start;
974     /* Put the checksum obtained into the packet */
975     stw_be_p(data + vhdr->csum_offset, net_raw_checksum(data, len));
976 
977     vhdr->flags &= ~VIRTIO_NET_HDR_F_NEEDS_CSUM;
978     vhdr->flags |= VIRTIO_NET_HDR_F_DATA_VALID;
979 }
980 
981 static void vmxnet3_rx_update_descr(struct NetRxPkt *pkt,
982     struct Vmxnet3_RxCompDesc *rxcd)
983 {
984     int csum_ok, is_gso;
985     bool isip4, isip6, istcp, isudp;
986     struct virtio_net_hdr *vhdr;
987     uint8_t offload_type;
988 
989     if (net_rx_pkt_is_vlan_stripped(pkt)) {
990         rxcd->ts = 1;
991         rxcd->tci = net_rx_pkt_get_vlan_tag(pkt);
992     }
993 
994     if (!net_rx_pkt_has_virt_hdr(pkt)) {
995         goto nocsum;
996     }
997 
998     vhdr = net_rx_pkt_get_vhdr(pkt);
999     /*
1000      * Checksum is valid when lower level tell so or when lower level
1001      * requires checksum offload telling that packet produced/bridged
1002      * locally and did travel over network after last checksum calculation
1003      * or production
1004      */
1005     csum_ok = VMXNET_FLAG_IS_SET(vhdr->flags, VIRTIO_NET_HDR_F_DATA_VALID) ||
1006               VMXNET_FLAG_IS_SET(vhdr->flags, VIRTIO_NET_HDR_F_NEEDS_CSUM);
1007 
1008     offload_type = vhdr->gso_type & ~VIRTIO_NET_HDR_GSO_ECN;
1009     is_gso = (offload_type != VIRTIO_NET_HDR_GSO_NONE) ? 1 : 0;
1010 
1011     if (!csum_ok && !is_gso) {
1012         goto nocsum;
1013     }
1014 
1015     net_rx_pkt_get_protocols(pkt, &isip4, &isip6, &isudp, &istcp);
1016     if ((!istcp && !isudp) || (!isip4 && !isip6)) {
1017         goto nocsum;
1018     }
1019 
1020     rxcd->cnc = 0;
1021     rxcd->v4 = isip4 ? 1 : 0;
1022     rxcd->v6 = isip6 ? 1 : 0;
1023     rxcd->tcp = istcp ? 1 : 0;
1024     rxcd->udp = isudp ? 1 : 0;
1025     rxcd->fcs = rxcd->tuc = rxcd->ipc = 1;
1026     return;
1027 
1028 nocsum:
1029     rxcd->cnc = 1;
1030     return;
1031 }
1032 
1033 static void
1034 vmxnet3_pci_dma_writev(PCIDevice *pci_dev,
1035                        const struct iovec *iov,
1036                        size_t start_iov_off,
1037                        hwaddr target_addr,
1038                        size_t bytes_to_copy)
1039 {
1040     size_t curr_off = 0;
1041     size_t copied = 0;
1042 
1043     while (bytes_to_copy) {
1044         if (start_iov_off < (curr_off + iov->iov_len)) {
1045             size_t chunk_len =
1046                 MIN((curr_off + iov->iov_len) - start_iov_off, bytes_to_copy);
1047 
1048             pci_dma_write(pci_dev, target_addr + copied,
1049                           iov->iov_base + start_iov_off - curr_off,
1050                           chunk_len);
1051 
1052             copied += chunk_len;
1053             start_iov_off += chunk_len;
1054             curr_off = start_iov_off;
1055             bytes_to_copy -= chunk_len;
1056         } else {
1057             curr_off += iov->iov_len;
1058         }
1059         iov++;
1060     }
1061 }
1062 
1063 static bool
1064 vmxnet3_indicate_packet(VMXNET3State *s)
1065 {
1066     struct Vmxnet3_RxDesc rxd;
1067     PCIDevice *d = PCI_DEVICE(s);
1068     bool is_head = true;
1069     uint32_t rxd_idx;
1070     uint32_t rx_ridx = 0;
1071 
1072     struct Vmxnet3_RxCompDesc rxcd;
1073     uint32_t new_rxcd_gen = VMXNET3_INIT_GEN;
1074     hwaddr new_rxcd_pa = 0;
1075     hwaddr ready_rxcd_pa = 0;
1076     struct iovec *data = net_rx_pkt_get_iovec(s->rx_pkt);
1077     size_t bytes_copied = 0;
1078     size_t bytes_left = net_rx_pkt_get_total_len(s->rx_pkt);
1079     uint16_t num_frags = 0;
1080     size_t chunk_size;
1081 
1082     net_rx_pkt_dump(s->rx_pkt);
1083 
1084     while (bytes_left > 0) {
1085 
1086         /* cannot add more frags to packet */
1087         if (num_frags == s->max_rx_frags) {
1088             break;
1089         }
1090 
1091         new_rxcd_pa = vmxnet3_pop_rxc_descr(s, RXQ_IDX, &new_rxcd_gen);
1092         if (!new_rxcd_pa) {
1093             break;
1094         }
1095 
1096         if (!vmxnet3_get_next_rx_descr(s, is_head, &rxd, &rxd_idx, &rx_ridx)) {
1097             break;
1098         }
1099 
1100         chunk_size = MIN(bytes_left, rxd.len);
1101         vmxnet3_pci_dma_writev(d, data, bytes_copied,
1102                                le64_to_cpu(rxd.addr), chunk_size);
1103         bytes_copied += chunk_size;
1104         bytes_left -= chunk_size;
1105 
1106         vmxnet3_dump_rx_descr(&rxd);
1107 
1108         if (ready_rxcd_pa != 0) {
1109             pci_dma_write(d, ready_rxcd_pa, &rxcd, sizeof(rxcd));
1110         }
1111 
1112         memset(&rxcd, 0, sizeof(struct Vmxnet3_RxCompDesc));
1113         rxcd.rxdIdx = rxd_idx;
1114         rxcd.len = chunk_size;
1115         rxcd.sop = is_head;
1116         rxcd.gen = new_rxcd_gen;
1117         rxcd.rqID = RXQ_IDX + rx_ridx * s->rxq_num;
1118 
1119         if (bytes_left == 0) {
1120             vmxnet3_rx_update_descr(s->rx_pkt, &rxcd);
1121         }
1122 
1123         VMW_RIPRN("RX Completion descriptor: rxRing: %lu rxIdx %lu len %lu "
1124                   "sop %d csum_correct %lu",
1125                   (unsigned long) rx_ridx,
1126                   (unsigned long) rxcd.rxdIdx,
1127                   (unsigned long) rxcd.len,
1128                   (int) rxcd.sop,
1129                   (unsigned long) rxcd.tuc);
1130 
1131         is_head = false;
1132         ready_rxcd_pa = new_rxcd_pa;
1133         new_rxcd_pa = 0;
1134         num_frags++;
1135     }
1136 
1137     if (ready_rxcd_pa != 0) {
1138         rxcd.eop = 1;
1139         rxcd.err = (bytes_left != 0);
1140 
1141         pci_dma_write(d, ready_rxcd_pa, &rxcd, sizeof(rxcd));
1142 
1143         /* Flush RX descriptor changes */
1144         smp_wmb();
1145     }
1146 
1147     if (new_rxcd_pa != 0) {
1148         vmxnet3_revert_rxc_descr(s, RXQ_IDX);
1149     }
1150 
1151     vmxnet3_trigger_interrupt(s, s->rxq_descr[RXQ_IDX].intr_idx);
1152 
1153     if (bytes_left == 0) {
1154         vmxnet3_on_rx_done_update_stats(s, RXQ_IDX, VMXNET3_PKT_STATUS_OK);
1155         return true;
1156     } else if (num_frags == s->max_rx_frags) {
1157         vmxnet3_on_rx_done_update_stats(s, RXQ_IDX, VMXNET3_PKT_STATUS_ERROR);
1158         return false;
1159     } else {
1160         vmxnet3_on_rx_done_update_stats(s, RXQ_IDX,
1161                                         VMXNET3_PKT_STATUS_OUT_OF_BUF);
1162         return false;
1163     }
1164 }
1165 
1166 static void
1167 vmxnet3_io_bar0_write(void *opaque, hwaddr addr,
1168                       uint64_t val, unsigned size)
1169 {
1170     VMXNET3State *s = opaque;
1171 
1172     if (!s->device_active) {
1173         return;
1174     }
1175 
1176     if (VMW_IS_MULTIREG_ADDR(addr, VMXNET3_REG_TXPROD,
1177                         VMXNET3_DEVICE_MAX_TX_QUEUES, VMXNET3_REG_ALIGN)) {
1178         int tx_queue_idx =
1179             VMW_MULTIREG_IDX_BY_ADDR(addr, VMXNET3_REG_TXPROD,
1180                                      VMXNET3_REG_ALIGN);
1181         assert(tx_queue_idx <= s->txq_num);
1182         vmxnet3_process_tx_queue(s, tx_queue_idx);
1183         return;
1184     }
1185 
1186     if (VMW_IS_MULTIREG_ADDR(addr, VMXNET3_REG_IMR,
1187                         VMXNET3_MAX_INTRS, VMXNET3_REG_ALIGN)) {
1188         int l = VMW_MULTIREG_IDX_BY_ADDR(addr, VMXNET3_REG_IMR,
1189                                          VMXNET3_REG_ALIGN);
1190 
1191         VMW_CBPRN("Interrupt mask for line %d written: 0x%" PRIx64, l, val);
1192 
1193         vmxnet3_on_interrupt_mask_changed(s, l, val);
1194         return;
1195     }
1196 
1197     if (VMW_IS_MULTIREG_ADDR(addr, VMXNET3_REG_RXPROD,
1198                         VMXNET3_DEVICE_MAX_RX_QUEUES, VMXNET3_REG_ALIGN) ||
1199        VMW_IS_MULTIREG_ADDR(addr, VMXNET3_REG_RXPROD2,
1200                         VMXNET3_DEVICE_MAX_RX_QUEUES, VMXNET3_REG_ALIGN)) {
1201         return;
1202     }
1203 
1204     VMW_WRPRN("BAR0 unknown write [%" PRIx64 "] = %" PRIx64 ", size %d",
1205               (uint64_t) addr, val, size);
1206 }
1207 
1208 static uint64_t
1209 vmxnet3_io_bar0_read(void *opaque, hwaddr addr, unsigned size)
1210 {
1211     VMXNET3State *s = opaque;
1212 
1213     if (VMW_IS_MULTIREG_ADDR(addr, VMXNET3_REG_IMR,
1214                         VMXNET3_MAX_INTRS, VMXNET3_REG_ALIGN)) {
1215         int l = VMW_MULTIREG_IDX_BY_ADDR(addr, VMXNET3_REG_IMR,
1216                                          VMXNET3_REG_ALIGN);
1217         return s->interrupt_states[l].is_masked;
1218     }
1219 
1220     VMW_CBPRN("BAR0 unknown read [%" PRIx64 "], size %d", addr, size);
1221     return 0;
1222 }
1223 
1224 static void vmxnet3_reset_interrupt_states(VMXNET3State *s)
1225 {
1226     int i;
1227     for (i = 0; i < ARRAY_SIZE(s->interrupt_states); i++) {
1228         s->interrupt_states[i].is_asserted = false;
1229         s->interrupt_states[i].is_pending = false;
1230         s->interrupt_states[i].is_masked = true;
1231     }
1232 }
1233 
1234 static void vmxnet3_reset_mac(VMXNET3State *s)
1235 {
1236     memcpy(&s->conf.macaddr.a, &s->perm_mac.a, sizeof(s->perm_mac.a));
1237     VMW_CFPRN("MAC address set to: " MAC_FMT, MAC_ARG(s->conf.macaddr.a));
1238 }
1239 
1240 static void vmxnet3_deactivate_device(VMXNET3State *s)
1241 {
1242     if (s->device_active) {
1243         VMW_CBPRN("Deactivating vmxnet3...");
1244         net_tx_pkt_reset(s->tx_pkt);
1245         net_tx_pkt_uninit(s->tx_pkt);
1246         net_rx_pkt_uninit(s->rx_pkt);
1247         s->device_active = false;
1248     }
1249 }
1250 
1251 static void vmxnet3_reset(VMXNET3State *s)
1252 {
1253     VMW_CBPRN("Resetting vmxnet3...");
1254 
1255     vmxnet3_deactivate_device(s);
1256     vmxnet3_reset_interrupt_states(s);
1257     s->drv_shmem = 0;
1258     s->tx_sop = true;
1259     s->skip_current_tx_pkt = false;
1260 }
1261 
1262 static void vmxnet3_update_rx_mode(VMXNET3State *s)
1263 {
1264     PCIDevice *d = PCI_DEVICE(s);
1265 
1266     s->rx_mode = VMXNET3_READ_DRV_SHARED32(d, s->drv_shmem,
1267                                            devRead.rxFilterConf.rxMode);
1268     VMW_CFPRN("RX mode: 0x%08X", s->rx_mode);
1269 }
1270 
1271 static void vmxnet3_update_vlan_filters(VMXNET3State *s)
1272 {
1273     int i;
1274     PCIDevice *d = PCI_DEVICE(s);
1275 
1276     /* Copy configuration from shared memory */
1277     VMXNET3_READ_DRV_SHARED(d, s->drv_shmem,
1278                             devRead.rxFilterConf.vfTable,
1279                             s->vlan_table,
1280                             sizeof(s->vlan_table));
1281 
1282     /* Invert byte order when needed */
1283     for (i = 0; i < ARRAY_SIZE(s->vlan_table); i++) {
1284         s->vlan_table[i] = le32_to_cpu(s->vlan_table[i]);
1285     }
1286 
1287     /* Dump configuration for debugging purposes */
1288     VMW_CFPRN("Configured VLANs:");
1289     for (i = 0; i < sizeof(s->vlan_table) * 8; i++) {
1290         if (VMXNET3_VFTABLE_ENTRY_IS_SET(s->vlan_table, i)) {
1291             VMW_CFPRN("\tVLAN %d is present", i);
1292         }
1293     }
1294 }
1295 
1296 static void vmxnet3_update_mcast_filters(VMXNET3State *s)
1297 {
1298     PCIDevice *d = PCI_DEVICE(s);
1299 
1300     uint16_t list_bytes =
1301         VMXNET3_READ_DRV_SHARED16(d, s->drv_shmem,
1302                                   devRead.rxFilterConf.mfTableLen);
1303 
1304     s->mcast_list_len = list_bytes / sizeof(s->mcast_list[0]);
1305 
1306     s->mcast_list = g_realloc(s->mcast_list, list_bytes);
1307     if (!s->mcast_list) {
1308         if (s->mcast_list_len == 0) {
1309             VMW_CFPRN("Current multicast list is empty");
1310         } else {
1311             VMW_ERPRN("Failed to allocate multicast list of %d elements",
1312                       s->mcast_list_len);
1313         }
1314         s->mcast_list_len = 0;
1315     } else {
1316         int i;
1317         hwaddr mcast_list_pa =
1318             VMXNET3_READ_DRV_SHARED64(d, s->drv_shmem,
1319                                       devRead.rxFilterConf.mfTablePA);
1320 
1321         pci_dma_read(d, mcast_list_pa, s->mcast_list, list_bytes);
1322 
1323         VMW_CFPRN("Current multicast list len is %d:", s->mcast_list_len);
1324         for (i = 0; i < s->mcast_list_len; i++) {
1325             VMW_CFPRN("\t" MAC_FMT, MAC_ARG(s->mcast_list[i].a));
1326         }
1327     }
1328 }
1329 
1330 static void vmxnet3_setup_rx_filtering(VMXNET3State *s)
1331 {
1332     vmxnet3_update_rx_mode(s);
1333     vmxnet3_update_vlan_filters(s);
1334     vmxnet3_update_mcast_filters(s);
1335 }
1336 
1337 static uint32_t vmxnet3_get_interrupt_config(VMXNET3State *s)
1338 {
1339     uint32_t interrupt_mode = VMXNET3_IT_AUTO | (VMXNET3_IMM_AUTO << 2);
1340     VMW_CFPRN("Interrupt config is 0x%X", interrupt_mode);
1341     return interrupt_mode;
1342 }
1343 
1344 static void vmxnet3_fill_stats(VMXNET3State *s)
1345 {
1346     int i;
1347     PCIDevice *d = PCI_DEVICE(s);
1348 
1349     if (!s->device_active)
1350         return;
1351 
1352     for (i = 0; i < s->txq_num; i++) {
1353         pci_dma_write(d,
1354                       s->txq_descr[i].tx_stats_pa,
1355                       &s->txq_descr[i].txq_stats,
1356                       sizeof(s->txq_descr[i].txq_stats));
1357     }
1358 
1359     for (i = 0; i < s->rxq_num; i++) {
1360         pci_dma_write(d,
1361                       s->rxq_descr[i].rx_stats_pa,
1362                       &s->rxq_descr[i].rxq_stats,
1363                       sizeof(s->rxq_descr[i].rxq_stats));
1364     }
1365 }
1366 
1367 static void vmxnet3_adjust_by_guest_type(VMXNET3State *s)
1368 {
1369     struct Vmxnet3_GOSInfo gos;
1370     PCIDevice *d = PCI_DEVICE(s);
1371 
1372     VMXNET3_READ_DRV_SHARED(d, s->drv_shmem, devRead.misc.driverInfo.gos,
1373                             &gos, sizeof(gos));
1374     s->rx_packets_compound =
1375         (gos.gosType == VMXNET3_GOS_TYPE_WIN) ? false : true;
1376 
1377     VMW_CFPRN("Guest type specifics: RXCOMPOUND: %d", s->rx_packets_compound);
1378 }
1379 
1380 static void
1381 vmxnet3_dump_conf_descr(const char *name,
1382                         struct Vmxnet3_VariableLenConfDesc *pm_descr)
1383 {
1384     VMW_CFPRN("%s descriptor dump: Version %u, Length %u",
1385               name, pm_descr->confVer, pm_descr->confLen);
1386 
1387 };
1388 
1389 static void vmxnet3_update_pm_state(VMXNET3State *s)
1390 {
1391     struct Vmxnet3_VariableLenConfDesc pm_descr;
1392     PCIDevice *d = PCI_DEVICE(s);
1393 
1394     pm_descr.confLen =
1395         VMXNET3_READ_DRV_SHARED32(d, s->drv_shmem, devRead.pmConfDesc.confLen);
1396     pm_descr.confVer =
1397         VMXNET3_READ_DRV_SHARED32(d, s->drv_shmem, devRead.pmConfDesc.confVer);
1398     pm_descr.confPA =
1399         VMXNET3_READ_DRV_SHARED64(d, s->drv_shmem, devRead.pmConfDesc.confPA);
1400 
1401     vmxnet3_dump_conf_descr("PM State", &pm_descr);
1402 }
1403 
1404 static void vmxnet3_update_features(VMXNET3State *s)
1405 {
1406     uint32_t guest_features;
1407     int rxcso_supported;
1408     PCIDevice *d = PCI_DEVICE(s);
1409 
1410     guest_features = VMXNET3_READ_DRV_SHARED32(d, s->drv_shmem,
1411                                                devRead.misc.uptFeatures);
1412 
1413     rxcso_supported = VMXNET_FLAG_IS_SET(guest_features, UPT1_F_RXCSUM);
1414     s->rx_vlan_stripping = VMXNET_FLAG_IS_SET(guest_features, UPT1_F_RXVLAN);
1415     s->lro_supported = VMXNET_FLAG_IS_SET(guest_features, UPT1_F_LRO);
1416 
1417     VMW_CFPRN("Features configuration: LRO: %d, RXCSUM: %d, VLANSTRIP: %d",
1418               s->lro_supported, rxcso_supported,
1419               s->rx_vlan_stripping);
1420     if (s->peer_has_vhdr) {
1421         qemu_set_offload(qemu_get_queue(s->nic)->peer,
1422                          rxcso_supported,
1423                          s->lro_supported,
1424                          s->lro_supported,
1425                          0,
1426                          0);
1427     }
1428 }
1429 
1430 static bool vmxnet3_verify_intx(VMXNET3State *s, int intx)
1431 {
1432     return s->msix_used || msi_enabled(PCI_DEVICE(s))
1433         || intx == pci_get_byte(s->parent_obj.config + PCI_INTERRUPT_PIN) - 1;
1434 }
1435 
1436 static void vmxnet3_validate_interrupt_idx(bool is_msix, int idx)
1437 {
1438     int max_ints = is_msix ? VMXNET3_MAX_INTRS : VMXNET3_MAX_NMSIX_INTRS;
1439     if (idx >= max_ints) {
1440         hw_error("Bad interrupt index: %d\n", idx);
1441     }
1442 }
1443 
1444 static void vmxnet3_validate_interrupts(VMXNET3State *s)
1445 {
1446     int i;
1447 
1448     VMW_CFPRN("Verifying event interrupt index (%d)", s->event_int_idx);
1449     vmxnet3_validate_interrupt_idx(s->msix_used, s->event_int_idx);
1450 
1451     for (i = 0; i < s->txq_num; i++) {
1452         int idx = s->txq_descr[i].intr_idx;
1453         VMW_CFPRN("Verifying TX queue %d interrupt index (%d)", i, idx);
1454         vmxnet3_validate_interrupt_idx(s->msix_used, idx);
1455     }
1456 
1457     for (i = 0; i < s->rxq_num; i++) {
1458         int idx = s->rxq_descr[i].intr_idx;
1459         VMW_CFPRN("Verifying RX queue %d interrupt index (%d)", i, idx);
1460         vmxnet3_validate_interrupt_idx(s->msix_used, idx);
1461     }
1462 }
1463 
1464 static void vmxnet3_validate_queues(VMXNET3State *s)
1465 {
1466     /*
1467     * txq_num and rxq_num are total number of queues
1468     * configured by guest. These numbers must not
1469     * exceed corresponding maximal values.
1470     */
1471 
1472     if (s->txq_num > VMXNET3_DEVICE_MAX_TX_QUEUES) {
1473         hw_error("Bad TX queues number: %d\n", s->txq_num);
1474     }
1475 
1476     if (s->rxq_num > VMXNET3_DEVICE_MAX_RX_QUEUES) {
1477         hw_error("Bad RX queues number: %d\n", s->rxq_num);
1478     }
1479 }
1480 
1481 static void vmxnet3_activate_device(VMXNET3State *s)
1482 {
1483     int i;
1484     static const uint32_t VMXNET3_DEF_TX_THRESHOLD = 1;
1485     PCIDevice *d = PCI_DEVICE(s);
1486     hwaddr qdescr_table_pa;
1487     uint64_t pa;
1488     uint32_t size;
1489 
1490     /* Verify configuration consistency */
1491     if (!vmxnet3_verify_driver_magic(d, s->drv_shmem)) {
1492         VMW_ERPRN("Device configuration received from driver is invalid");
1493         return;
1494     }
1495 
1496     /* Verify if device is active */
1497     if (s->device_active) {
1498         VMW_CFPRN("Vmxnet3 device is active");
1499         return;
1500     }
1501 
1502     vmxnet3_adjust_by_guest_type(s);
1503     vmxnet3_update_features(s);
1504     vmxnet3_update_pm_state(s);
1505     vmxnet3_setup_rx_filtering(s);
1506     /* Cache fields from shared memory */
1507     s->mtu = VMXNET3_READ_DRV_SHARED32(d, s->drv_shmem, devRead.misc.mtu);
1508     VMW_CFPRN("MTU is %u", s->mtu);
1509 
1510     s->max_rx_frags =
1511         VMXNET3_READ_DRV_SHARED16(d, s->drv_shmem, devRead.misc.maxNumRxSG);
1512 
1513     if (s->max_rx_frags == 0) {
1514         s->max_rx_frags = 1;
1515     }
1516 
1517     VMW_CFPRN("Max RX fragments is %u", s->max_rx_frags);
1518 
1519     s->event_int_idx =
1520         VMXNET3_READ_DRV_SHARED8(d, s->drv_shmem, devRead.intrConf.eventIntrIdx);
1521     assert(vmxnet3_verify_intx(s, s->event_int_idx));
1522     VMW_CFPRN("Events interrupt line is %u", s->event_int_idx);
1523 
1524     s->auto_int_masking =
1525         VMXNET3_READ_DRV_SHARED8(d, s->drv_shmem, devRead.intrConf.autoMask);
1526     VMW_CFPRN("Automatic interrupt masking is %d", (int)s->auto_int_masking);
1527 
1528     s->txq_num =
1529         VMXNET3_READ_DRV_SHARED8(d, s->drv_shmem, devRead.misc.numTxQueues);
1530     s->rxq_num =
1531         VMXNET3_READ_DRV_SHARED8(d, s->drv_shmem, devRead.misc.numRxQueues);
1532 
1533     VMW_CFPRN("Number of TX/RX queues %u/%u", s->txq_num, s->rxq_num);
1534     vmxnet3_validate_queues(s);
1535 
1536     qdescr_table_pa =
1537         VMXNET3_READ_DRV_SHARED64(d, s->drv_shmem, devRead.misc.queueDescPA);
1538     VMW_CFPRN("TX queues descriptors table is at 0x%" PRIx64, qdescr_table_pa);
1539 
1540     /*
1541      * Worst-case scenario is a packet that holds all TX rings space so
1542      * we calculate total size of all TX rings for max TX fragments number
1543      */
1544     s->max_tx_frags = 0;
1545 
1546     /* TX queues */
1547     for (i = 0; i < s->txq_num; i++) {
1548         hwaddr qdescr_pa =
1549             qdescr_table_pa + i * sizeof(struct Vmxnet3_TxQueueDesc);
1550 
1551         /* Read interrupt number for this TX queue */
1552         s->txq_descr[i].intr_idx =
1553             VMXNET3_READ_TX_QUEUE_DESCR8(d, qdescr_pa, conf.intrIdx);
1554         assert(vmxnet3_verify_intx(s, s->txq_descr[i].intr_idx));
1555 
1556         VMW_CFPRN("TX Queue %d interrupt: %d", i, s->txq_descr[i].intr_idx);
1557 
1558         /* Read rings memory locations for TX queues */
1559         pa = VMXNET3_READ_TX_QUEUE_DESCR64(d, qdescr_pa, conf.txRingBasePA);
1560         size = VMXNET3_READ_TX_QUEUE_DESCR32(d, qdescr_pa, conf.txRingSize);
1561 
1562         vmxnet3_ring_init(d, &s->txq_descr[i].tx_ring, pa, size,
1563                           sizeof(struct Vmxnet3_TxDesc), false);
1564         VMXNET3_RING_DUMP(VMW_CFPRN, "TX", i, &s->txq_descr[i].tx_ring);
1565 
1566         s->max_tx_frags += size;
1567 
1568         /* TXC ring */
1569         pa = VMXNET3_READ_TX_QUEUE_DESCR64(d, qdescr_pa, conf.compRingBasePA);
1570         size = VMXNET3_READ_TX_QUEUE_DESCR32(d, qdescr_pa, conf.compRingSize);
1571         vmxnet3_ring_init(d, &s->txq_descr[i].comp_ring, pa, size,
1572                           sizeof(struct Vmxnet3_TxCompDesc), true);
1573         VMXNET3_RING_DUMP(VMW_CFPRN, "TXC", i, &s->txq_descr[i].comp_ring);
1574 
1575         s->txq_descr[i].tx_stats_pa =
1576             qdescr_pa + offsetof(struct Vmxnet3_TxQueueDesc, stats);
1577 
1578         memset(&s->txq_descr[i].txq_stats, 0,
1579                sizeof(s->txq_descr[i].txq_stats));
1580 
1581         /* Fill device-managed parameters for queues */
1582         VMXNET3_WRITE_TX_QUEUE_DESCR32(d, qdescr_pa,
1583                                        ctrl.txThreshold,
1584                                        VMXNET3_DEF_TX_THRESHOLD);
1585     }
1586 
1587     /* Preallocate TX packet wrapper */
1588     VMW_CFPRN("Max TX fragments is %u", s->max_tx_frags);
1589     net_tx_pkt_init(&s->tx_pkt, PCI_DEVICE(s),
1590                     s->max_tx_frags, s->peer_has_vhdr);
1591     net_rx_pkt_init(&s->rx_pkt, s->peer_has_vhdr);
1592 
1593     /* Read rings memory locations for RX queues */
1594     for (i = 0; i < s->rxq_num; i++) {
1595         int j;
1596         hwaddr qd_pa =
1597             qdescr_table_pa + s->txq_num * sizeof(struct Vmxnet3_TxQueueDesc) +
1598             i * sizeof(struct Vmxnet3_RxQueueDesc);
1599 
1600         /* Read interrupt number for this RX queue */
1601         s->rxq_descr[i].intr_idx =
1602             VMXNET3_READ_TX_QUEUE_DESCR8(d, qd_pa, conf.intrIdx);
1603         assert(vmxnet3_verify_intx(s, s->rxq_descr[i].intr_idx));
1604 
1605         VMW_CFPRN("RX Queue %d interrupt: %d", i, s->rxq_descr[i].intr_idx);
1606 
1607         /* Read rings memory locations */
1608         for (j = 0; j < VMXNET3_RX_RINGS_PER_QUEUE; j++) {
1609             /* RX rings */
1610             pa = VMXNET3_READ_RX_QUEUE_DESCR64(d, qd_pa, conf.rxRingBasePA[j]);
1611             size = VMXNET3_READ_RX_QUEUE_DESCR32(d, qd_pa, conf.rxRingSize[j]);
1612             vmxnet3_ring_init(d, &s->rxq_descr[i].rx_ring[j], pa, size,
1613                               sizeof(struct Vmxnet3_RxDesc), false);
1614             VMW_CFPRN("RX queue %d:%d: Base: %" PRIx64 ", Size: %d",
1615                       i, j, pa, size);
1616         }
1617 
1618         /* RXC ring */
1619         pa = VMXNET3_READ_RX_QUEUE_DESCR64(d, qd_pa, conf.compRingBasePA);
1620         size = VMXNET3_READ_RX_QUEUE_DESCR32(d, qd_pa, conf.compRingSize);
1621         vmxnet3_ring_init(d, &s->rxq_descr[i].comp_ring, pa, size,
1622                           sizeof(struct Vmxnet3_RxCompDesc), true);
1623         VMW_CFPRN("RXC queue %d: Base: %" PRIx64 ", Size: %d", i, pa, size);
1624 
1625         s->rxq_descr[i].rx_stats_pa =
1626             qd_pa + offsetof(struct Vmxnet3_RxQueueDesc, stats);
1627         memset(&s->rxq_descr[i].rxq_stats, 0,
1628                sizeof(s->rxq_descr[i].rxq_stats));
1629     }
1630 
1631     vmxnet3_validate_interrupts(s);
1632 
1633     /* Make sure everything is in place before device activation */
1634     smp_wmb();
1635 
1636     vmxnet3_reset_mac(s);
1637 
1638     s->device_active = true;
1639 }
1640 
1641 static void vmxnet3_handle_command(VMXNET3State *s, uint64_t cmd)
1642 {
1643     s->last_command = cmd;
1644 
1645     switch (cmd) {
1646     case VMXNET3_CMD_GET_PERM_MAC_HI:
1647         VMW_CBPRN("Set: Get upper part of permanent MAC");
1648         break;
1649 
1650     case VMXNET3_CMD_GET_PERM_MAC_LO:
1651         VMW_CBPRN("Set: Get lower part of permanent MAC");
1652         break;
1653 
1654     case VMXNET3_CMD_GET_STATS:
1655         VMW_CBPRN("Set: Get device statistics");
1656         vmxnet3_fill_stats(s);
1657         break;
1658 
1659     case VMXNET3_CMD_ACTIVATE_DEV:
1660         VMW_CBPRN("Set: Activating vmxnet3 device");
1661         vmxnet3_activate_device(s);
1662         break;
1663 
1664     case VMXNET3_CMD_UPDATE_RX_MODE:
1665         VMW_CBPRN("Set: Update rx mode");
1666         vmxnet3_update_rx_mode(s);
1667         break;
1668 
1669     case VMXNET3_CMD_UPDATE_VLAN_FILTERS:
1670         VMW_CBPRN("Set: Update VLAN filters");
1671         vmxnet3_update_vlan_filters(s);
1672         break;
1673 
1674     case VMXNET3_CMD_UPDATE_MAC_FILTERS:
1675         VMW_CBPRN("Set: Update MAC filters");
1676         vmxnet3_update_mcast_filters(s);
1677         break;
1678 
1679     case VMXNET3_CMD_UPDATE_FEATURE:
1680         VMW_CBPRN("Set: Update features");
1681         vmxnet3_update_features(s);
1682         break;
1683 
1684     case VMXNET3_CMD_UPDATE_PMCFG:
1685         VMW_CBPRN("Set: Update power management config");
1686         vmxnet3_update_pm_state(s);
1687         break;
1688 
1689     case VMXNET3_CMD_GET_LINK:
1690         VMW_CBPRN("Set: Get link");
1691         break;
1692 
1693     case VMXNET3_CMD_RESET_DEV:
1694         VMW_CBPRN("Set: Reset device");
1695         vmxnet3_reset(s);
1696         break;
1697 
1698     case VMXNET3_CMD_QUIESCE_DEV:
1699         VMW_CBPRN("Set: VMXNET3_CMD_QUIESCE_DEV - deactivate the device");
1700         vmxnet3_deactivate_device(s);
1701         break;
1702 
1703     case VMXNET3_CMD_GET_CONF_INTR:
1704         VMW_CBPRN("Set: VMXNET3_CMD_GET_CONF_INTR - interrupt configuration");
1705         break;
1706 
1707     case VMXNET3_CMD_GET_ADAPTIVE_RING_INFO:
1708         VMW_CBPRN("Set: VMXNET3_CMD_GET_ADAPTIVE_RING_INFO - "
1709                   "adaptive ring info flags");
1710         break;
1711 
1712     case VMXNET3_CMD_GET_DID_LO:
1713         VMW_CBPRN("Set: Get lower part of device ID");
1714         break;
1715 
1716     case VMXNET3_CMD_GET_DID_HI:
1717         VMW_CBPRN("Set: Get upper part of device ID");
1718         break;
1719 
1720     case VMXNET3_CMD_GET_DEV_EXTRA_INFO:
1721         VMW_CBPRN("Set: Get device extra info");
1722         break;
1723 
1724     default:
1725         VMW_CBPRN("Received unknown command: %" PRIx64, cmd);
1726         break;
1727     }
1728 }
1729 
1730 static uint64_t vmxnet3_get_command_status(VMXNET3State *s)
1731 {
1732     uint64_t ret;
1733 
1734     switch (s->last_command) {
1735     case VMXNET3_CMD_ACTIVATE_DEV:
1736         ret = (s->device_active) ? 0 : 1;
1737         VMW_CFPRN("Device active: %" PRIx64, ret);
1738         break;
1739 
1740     case VMXNET3_CMD_RESET_DEV:
1741     case VMXNET3_CMD_QUIESCE_DEV:
1742     case VMXNET3_CMD_GET_QUEUE_STATUS:
1743     case VMXNET3_CMD_GET_DEV_EXTRA_INFO:
1744         ret = 0;
1745         break;
1746 
1747     case VMXNET3_CMD_GET_LINK:
1748         ret = s->link_status_and_speed;
1749         VMW_CFPRN("Link and speed: %" PRIx64, ret);
1750         break;
1751 
1752     case VMXNET3_CMD_GET_PERM_MAC_LO:
1753         ret = vmxnet3_get_mac_low(&s->perm_mac);
1754         break;
1755 
1756     case VMXNET3_CMD_GET_PERM_MAC_HI:
1757         ret = vmxnet3_get_mac_high(&s->perm_mac);
1758         break;
1759 
1760     case VMXNET3_CMD_GET_CONF_INTR:
1761         ret = vmxnet3_get_interrupt_config(s);
1762         break;
1763 
1764     case VMXNET3_CMD_GET_ADAPTIVE_RING_INFO:
1765         ret = VMXNET3_DISABLE_ADAPTIVE_RING;
1766         break;
1767 
1768     case VMXNET3_CMD_GET_DID_LO:
1769         ret = PCI_DEVICE_ID_VMWARE_VMXNET3;
1770         break;
1771 
1772     case VMXNET3_CMD_GET_DID_HI:
1773         ret = VMXNET3_DEVICE_REVISION;
1774         break;
1775 
1776     default:
1777         VMW_WRPRN("Received request for unknown command: %x", s->last_command);
1778         ret = 0;
1779         break;
1780     }
1781 
1782     return ret;
1783 }
1784 
1785 static void vmxnet3_set_events(VMXNET3State *s, uint32_t val)
1786 {
1787     uint32_t events;
1788     PCIDevice *d = PCI_DEVICE(s);
1789 
1790     VMW_CBPRN("Setting events: 0x%x", val);
1791     events = VMXNET3_READ_DRV_SHARED32(d, s->drv_shmem, ecr) | val;
1792     VMXNET3_WRITE_DRV_SHARED32(d, s->drv_shmem, ecr, events);
1793 }
1794 
1795 static void vmxnet3_ack_events(VMXNET3State *s, uint32_t val)
1796 {
1797     PCIDevice *d = PCI_DEVICE(s);
1798     uint32_t events;
1799 
1800     VMW_CBPRN("Clearing events: 0x%x", val);
1801     events = VMXNET3_READ_DRV_SHARED32(d, s->drv_shmem, ecr) & ~val;
1802     VMXNET3_WRITE_DRV_SHARED32(d, s->drv_shmem, ecr, events);
1803 }
1804 
1805 static void
1806 vmxnet3_io_bar1_write(void *opaque,
1807                       hwaddr addr,
1808                       uint64_t val,
1809                       unsigned size)
1810 {
1811     VMXNET3State *s = opaque;
1812 
1813     switch (addr) {
1814     /* Vmxnet3 Revision Report Selection */
1815     case VMXNET3_REG_VRRS:
1816         VMW_CBPRN("Write BAR1 [VMXNET3_REG_VRRS] = %" PRIx64 ", size %d",
1817                   val, size);
1818         break;
1819 
1820     /* UPT Version Report Selection */
1821     case VMXNET3_REG_UVRS:
1822         VMW_CBPRN("Write BAR1 [VMXNET3_REG_UVRS] = %" PRIx64 ", size %d",
1823                   val, size);
1824         break;
1825 
1826     /* Driver Shared Address Low */
1827     case VMXNET3_REG_DSAL:
1828         VMW_CBPRN("Write BAR1 [VMXNET3_REG_DSAL] = %" PRIx64 ", size %d",
1829                   val, size);
1830         /*
1831          * Guest driver will first write the low part of the shared
1832          * memory address. We save it to temp variable and set the
1833          * shared address only after we get the high part
1834          */
1835         if (val == 0) {
1836             vmxnet3_deactivate_device(s);
1837         }
1838         s->temp_shared_guest_driver_memory = val;
1839         s->drv_shmem = 0;
1840         break;
1841 
1842     /* Driver Shared Address High */
1843     case VMXNET3_REG_DSAH:
1844         VMW_CBPRN("Write BAR1 [VMXNET3_REG_DSAH] = %" PRIx64 ", size %d",
1845                   val, size);
1846         /*
1847          * Set the shared memory between guest driver and device.
1848          * We already should have low address part.
1849          */
1850         s->drv_shmem = s->temp_shared_guest_driver_memory | (val << 32);
1851         break;
1852 
1853     /* Command */
1854     case VMXNET3_REG_CMD:
1855         VMW_CBPRN("Write BAR1 [VMXNET3_REG_CMD] = %" PRIx64 ", size %d",
1856                   val, size);
1857         vmxnet3_handle_command(s, val);
1858         break;
1859 
1860     /* MAC Address Low */
1861     case VMXNET3_REG_MACL:
1862         VMW_CBPRN("Write BAR1 [VMXNET3_REG_MACL] = %" PRIx64 ", size %d",
1863                   val, size);
1864         s->temp_mac = val;
1865         break;
1866 
1867     /* MAC Address High */
1868     case VMXNET3_REG_MACH:
1869         VMW_CBPRN("Write BAR1 [VMXNET3_REG_MACH] = %" PRIx64 ", size %d",
1870                   val, size);
1871         vmxnet3_set_variable_mac(s, val, s->temp_mac);
1872         break;
1873 
1874     /* Interrupt Cause Register */
1875     case VMXNET3_REG_ICR:
1876         VMW_CBPRN("Write BAR1 [VMXNET3_REG_ICR] = %" PRIx64 ", size %d",
1877                   val, size);
1878         g_assert_not_reached();
1879         break;
1880 
1881     /* Event Cause Register */
1882     case VMXNET3_REG_ECR:
1883         VMW_CBPRN("Write BAR1 [VMXNET3_REG_ECR] = %" PRIx64 ", size %d",
1884                   val, size);
1885         vmxnet3_ack_events(s, val);
1886         break;
1887 
1888     default:
1889         VMW_CBPRN("Unknown Write to BAR1 [%" PRIx64 "] = %" PRIx64 ", size %d",
1890                   addr, val, size);
1891         break;
1892     }
1893 }
1894 
1895 static uint64_t
1896 vmxnet3_io_bar1_read(void *opaque, hwaddr addr, unsigned size)
1897 {
1898         VMXNET3State *s = opaque;
1899         uint64_t ret = 0;
1900 
1901         switch (addr) {
1902         /* Vmxnet3 Revision Report Selection */
1903         case VMXNET3_REG_VRRS:
1904             VMW_CBPRN("Read BAR1 [VMXNET3_REG_VRRS], size %d", size);
1905             ret = VMXNET3_DEVICE_REVISION;
1906             break;
1907 
1908         /* UPT Version Report Selection */
1909         case VMXNET3_REG_UVRS:
1910             VMW_CBPRN("Read BAR1 [VMXNET3_REG_UVRS], size %d", size);
1911             ret = VMXNET3_UPT_REVISION;
1912             break;
1913 
1914         /* Command */
1915         case VMXNET3_REG_CMD:
1916             VMW_CBPRN("Read BAR1 [VMXNET3_REG_CMD], size %d", size);
1917             ret = vmxnet3_get_command_status(s);
1918             break;
1919 
1920         /* MAC Address Low */
1921         case VMXNET3_REG_MACL:
1922             VMW_CBPRN("Read BAR1 [VMXNET3_REG_MACL], size %d", size);
1923             ret = vmxnet3_get_mac_low(&s->conf.macaddr);
1924             break;
1925 
1926         /* MAC Address High */
1927         case VMXNET3_REG_MACH:
1928             VMW_CBPRN("Read BAR1 [VMXNET3_REG_MACH], size %d", size);
1929             ret = vmxnet3_get_mac_high(&s->conf.macaddr);
1930             break;
1931 
1932         /*
1933          * Interrupt Cause Register
1934          * Used for legacy interrupts only so interrupt index always 0
1935          */
1936         case VMXNET3_REG_ICR:
1937             VMW_CBPRN("Read BAR1 [VMXNET3_REG_ICR], size %d", size);
1938             if (vmxnet3_interrupt_asserted(s, 0)) {
1939                 vmxnet3_clear_interrupt(s, 0);
1940                 ret = true;
1941             } else {
1942                 ret = false;
1943             }
1944             break;
1945 
1946         default:
1947             VMW_CBPRN("Unknow read BAR1[%" PRIx64 "], %d bytes", addr, size);
1948             break;
1949         }
1950 
1951         return ret;
1952 }
1953 
1954 static int
1955 vmxnet3_can_receive(NetClientState *nc)
1956 {
1957     VMXNET3State *s = qemu_get_nic_opaque(nc);
1958     return s->device_active &&
1959            VMXNET_FLAG_IS_SET(s->link_status_and_speed, VMXNET3_LINK_STATUS_UP);
1960 }
1961 
1962 static inline bool
1963 vmxnet3_is_registered_vlan(VMXNET3State *s, const void *data)
1964 {
1965     uint16_t vlan_tag = eth_get_pkt_tci(data) & VLAN_VID_MASK;
1966     if (IS_SPECIAL_VLAN_ID(vlan_tag)) {
1967         return true;
1968     }
1969 
1970     return VMXNET3_VFTABLE_ENTRY_IS_SET(s->vlan_table, vlan_tag);
1971 }
1972 
1973 static bool
1974 vmxnet3_is_allowed_mcast_group(VMXNET3State *s, const uint8_t *group_mac)
1975 {
1976     int i;
1977     for (i = 0; i < s->mcast_list_len; i++) {
1978         if (!memcmp(group_mac, s->mcast_list[i].a, sizeof(s->mcast_list[i]))) {
1979             return true;
1980         }
1981     }
1982     return false;
1983 }
1984 
1985 static bool
1986 vmxnet3_rx_filter_may_indicate(VMXNET3State *s, const void *data,
1987     size_t size)
1988 {
1989     struct eth_header *ehdr = PKT_GET_ETH_HDR(data);
1990 
1991     if (VMXNET_FLAG_IS_SET(s->rx_mode, VMXNET3_RXM_PROMISC)) {
1992         return true;
1993     }
1994 
1995     if (!vmxnet3_is_registered_vlan(s, data)) {
1996         return false;
1997     }
1998 
1999     switch (net_rx_pkt_get_packet_type(s->rx_pkt)) {
2000     case ETH_PKT_UCAST:
2001         if (!VMXNET_FLAG_IS_SET(s->rx_mode, VMXNET3_RXM_UCAST)) {
2002             return false;
2003         }
2004         if (memcmp(s->conf.macaddr.a, ehdr->h_dest, ETH_ALEN)) {
2005             return false;
2006         }
2007         break;
2008 
2009     case ETH_PKT_BCAST:
2010         if (!VMXNET_FLAG_IS_SET(s->rx_mode, VMXNET3_RXM_BCAST)) {
2011             return false;
2012         }
2013         break;
2014 
2015     case ETH_PKT_MCAST:
2016         if (VMXNET_FLAG_IS_SET(s->rx_mode, VMXNET3_RXM_ALL_MULTI)) {
2017             return true;
2018         }
2019         if (!VMXNET_FLAG_IS_SET(s->rx_mode, VMXNET3_RXM_MCAST)) {
2020             return false;
2021         }
2022         if (!vmxnet3_is_allowed_mcast_group(s, ehdr->h_dest)) {
2023             return false;
2024         }
2025         break;
2026 
2027     default:
2028         g_assert_not_reached();
2029     }
2030 
2031     return true;
2032 }
2033 
2034 static ssize_t
2035 vmxnet3_receive(NetClientState *nc, const uint8_t *buf, size_t size)
2036 {
2037     VMXNET3State *s = qemu_get_nic_opaque(nc);
2038     size_t bytes_indicated;
2039     uint8_t min_buf[MIN_BUF_SIZE];
2040 
2041     if (!vmxnet3_can_receive(nc)) {
2042         VMW_PKPRN("Cannot receive now");
2043         return -1;
2044     }
2045 
2046     if (s->peer_has_vhdr) {
2047         net_rx_pkt_set_vhdr(s->rx_pkt, (struct virtio_net_hdr *)buf);
2048         buf += sizeof(struct virtio_net_hdr);
2049         size -= sizeof(struct virtio_net_hdr);
2050     }
2051 
2052     /* Pad to minimum Ethernet frame length */
2053     if (size < sizeof(min_buf)) {
2054         memcpy(min_buf, buf, size);
2055         memset(&min_buf[size], 0, sizeof(min_buf) - size);
2056         buf = min_buf;
2057         size = sizeof(min_buf);
2058     }
2059 
2060     net_rx_pkt_set_packet_type(s->rx_pkt,
2061         get_eth_packet_type(PKT_GET_ETH_HDR(buf)));
2062 
2063     if (vmxnet3_rx_filter_may_indicate(s, buf, size)) {
2064         net_rx_pkt_set_protocols(s->rx_pkt, buf, size);
2065         vmxnet3_rx_need_csum_calculate(s->rx_pkt, buf, size);
2066         net_rx_pkt_attach_data(s->rx_pkt, buf, size, s->rx_vlan_stripping);
2067         bytes_indicated = vmxnet3_indicate_packet(s) ? size : -1;
2068         if (bytes_indicated < size) {
2069             VMW_PKPRN("RX: %zu of %zu bytes indicated", bytes_indicated, size);
2070         }
2071     } else {
2072         VMW_PKPRN("Packet dropped by RX filter");
2073         bytes_indicated = size;
2074     }
2075 
2076     assert(size > 0);
2077     assert(bytes_indicated != 0);
2078     return bytes_indicated;
2079 }
2080 
2081 static void vmxnet3_set_link_status(NetClientState *nc)
2082 {
2083     VMXNET3State *s = qemu_get_nic_opaque(nc);
2084 
2085     if (nc->link_down) {
2086         s->link_status_and_speed &= ~VMXNET3_LINK_STATUS_UP;
2087     } else {
2088         s->link_status_and_speed |= VMXNET3_LINK_STATUS_UP;
2089     }
2090 
2091     vmxnet3_set_events(s, VMXNET3_ECR_LINK);
2092     vmxnet3_trigger_interrupt(s, s->event_int_idx);
2093 }
2094 
2095 static NetClientInfo net_vmxnet3_info = {
2096         .type = NET_CLIENT_DRIVER_NIC,
2097         .size = sizeof(NICState),
2098         .receive = vmxnet3_receive,
2099         .link_status_changed = vmxnet3_set_link_status,
2100 };
2101 
2102 static bool vmxnet3_peer_has_vnet_hdr(VMXNET3State *s)
2103 {
2104     NetClientState *nc = qemu_get_queue(s->nic);
2105 
2106     if (qemu_has_vnet_hdr(nc->peer)) {
2107         return true;
2108     }
2109 
2110     return false;
2111 }
2112 
2113 static void vmxnet3_net_uninit(VMXNET3State *s)
2114 {
2115     g_free(s->mcast_list);
2116     vmxnet3_deactivate_device(s);
2117     qemu_del_nic(s->nic);
2118 }
2119 
2120 static void vmxnet3_net_init(VMXNET3State *s)
2121 {
2122     DeviceState *d = DEVICE(s);
2123 
2124     VMW_CBPRN("vmxnet3_net_init called...");
2125 
2126     qemu_macaddr_default_if_unset(&s->conf.macaddr);
2127 
2128     /* Windows guest will query the address that was set on init */
2129     memcpy(&s->perm_mac.a, &s->conf.macaddr.a, sizeof(s->perm_mac.a));
2130 
2131     s->mcast_list = NULL;
2132     s->mcast_list_len = 0;
2133 
2134     s->link_status_and_speed = VMXNET3_LINK_SPEED | VMXNET3_LINK_STATUS_UP;
2135 
2136     VMW_CFPRN("Permanent MAC: " MAC_FMT, MAC_ARG(s->perm_mac.a));
2137 
2138     s->nic = qemu_new_nic(&net_vmxnet3_info, &s->conf,
2139                           object_get_typename(OBJECT(s)),
2140                           d->id, s);
2141 
2142     s->peer_has_vhdr = vmxnet3_peer_has_vnet_hdr(s);
2143     s->tx_sop = true;
2144     s->skip_current_tx_pkt = false;
2145     s->tx_pkt = NULL;
2146     s->rx_pkt = NULL;
2147     s->rx_vlan_stripping = false;
2148     s->lro_supported = false;
2149 
2150     if (s->peer_has_vhdr) {
2151         qemu_set_vnet_hdr_len(qemu_get_queue(s->nic)->peer,
2152             sizeof(struct virtio_net_hdr));
2153 
2154         qemu_using_vnet_hdr(qemu_get_queue(s->nic)->peer, 1);
2155     }
2156 
2157     qemu_format_nic_info_str(qemu_get_queue(s->nic), s->conf.macaddr.a);
2158 }
2159 
2160 static void
2161 vmxnet3_unuse_msix_vectors(VMXNET3State *s, int num_vectors)
2162 {
2163     PCIDevice *d = PCI_DEVICE(s);
2164     int i;
2165     for (i = 0; i < num_vectors; i++) {
2166         msix_vector_unuse(d, i);
2167     }
2168 }
2169 
2170 static bool
2171 vmxnet3_use_msix_vectors(VMXNET3State *s, int num_vectors)
2172 {
2173     PCIDevice *d = PCI_DEVICE(s);
2174     int i;
2175     for (i = 0; i < num_vectors; i++) {
2176         int res = msix_vector_use(d, i);
2177         if (0 > res) {
2178             VMW_WRPRN("Failed to use MSI-X vector %d, error %d", i, res);
2179             vmxnet3_unuse_msix_vectors(s, i);
2180             return false;
2181         }
2182     }
2183     return true;
2184 }
2185 
2186 static bool
2187 vmxnet3_init_msix(VMXNET3State *s)
2188 {
2189     PCIDevice *d = PCI_DEVICE(s);
2190     int res = msix_init(d, VMXNET3_MAX_INTRS,
2191                         &s->msix_bar,
2192                         VMXNET3_MSIX_BAR_IDX, VMXNET3_OFF_MSIX_TABLE,
2193                         &s->msix_bar,
2194                         VMXNET3_MSIX_BAR_IDX, VMXNET3_OFF_MSIX_PBA(s),
2195                         VMXNET3_MSIX_OFFSET(s), NULL);
2196 
2197     if (0 > res) {
2198         VMW_WRPRN("Failed to initialize MSI-X, error %d", res);
2199         s->msix_used = false;
2200     } else {
2201         if (!vmxnet3_use_msix_vectors(s, VMXNET3_MAX_INTRS)) {
2202             VMW_WRPRN("Failed to use MSI-X vectors, error %d", res);
2203             msix_uninit(d, &s->msix_bar, &s->msix_bar);
2204             s->msix_used = false;
2205         } else {
2206             s->msix_used = true;
2207         }
2208     }
2209     return s->msix_used;
2210 }
2211 
2212 static void
2213 vmxnet3_cleanup_msix(VMXNET3State *s)
2214 {
2215     PCIDevice *d = PCI_DEVICE(s);
2216 
2217     if (s->msix_used) {
2218         vmxnet3_unuse_msix_vectors(s, VMXNET3_MAX_INTRS);
2219         msix_uninit(d, &s->msix_bar, &s->msix_bar);
2220     }
2221 }
2222 
2223 static void
2224 vmxnet3_cleanup_msi(VMXNET3State *s)
2225 {
2226     PCIDevice *d = PCI_DEVICE(s);
2227 
2228     msi_uninit(d);
2229 }
2230 
2231 static void
2232 vmxnet3_msix_save(QEMUFile *f, void *opaque)
2233 {
2234     PCIDevice *d = PCI_DEVICE(opaque);
2235     msix_save(d, f);
2236 }
2237 
2238 static int
2239 vmxnet3_msix_load(QEMUFile *f, void *opaque, int version_id)
2240 {
2241     PCIDevice *d = PCI_DEVICE(opaque);
2242     msix_load(d, f);
2243     return 0;
2244 }
2245 
2246 static const MemoryRegionOps b0_ops = {
2247     .read = vmxnet3_io_bar0_read,
2248     .write = vmxnet3_io_bar0_write,
2249     .endianness = DEVICE_LITTLE_ENDIAN,
2250     .impl = {
2251             .min_access_size = 4,
2252             .max_access_size = 4,
2253     },
2254 };
2255 
2256 static const MemoryRegionOps b1_ops = {
2257     .read = vmxnet3_io_bar1_read,
2258     .write = vmxnet3_io_bar1_write,
2259     .endianness = DEVICE_LITTLE_ENDIAN,
2260     .impl = {
2261             .min_access_size = 4,
2262             .max_access_size = 4,
2263     },
2264 };
2265 
2266 static SaveVMHandlers savevm_vmxnet3_msix = {
2267     .save_state = vmxnet3_msix_save,
2268     .load_state = vmxnet3_msix_load,
2269 };
2270 
2271 static uint64_t vmxnet3_device_serial_num(VMXNET3State *s)
2272 {
2273     uint64_t dsn_payload;
2274     uint8_t *dsnp = (uint8_t *)&dsn_payload;
2275 
2276     dsnp[0] = 0xfe;
2277     dsnp[1] = s->conf.macaddr.a[3];
2278     dsnp[2] = s->conf.macaddr.a[4];
2279     dsnp[3] = s->conf.macaddr.a[5];
2280     dsnp[4] = s->conf.macaddr.a[0];
2281     dsnp[5] = s->conf.macaddr.a[1];
2282     dsnp[6] = s->conf.macaddr.a[2];
2283     dsnp[7] = 0xff;
2284     return dsn_payload;
2285 }
2286 
2287 
2288 #define VMXNET3_USE_64BIT         (true)
2289 #define VMXNET3_PER_VECTOR_MASK   (false)
2290 
2291 static void vmxnet3_pci_realize(PCIDevice *pci_dev, Error **errp)
2292 {
2293     DeviceState *dev = DEVICE(pci_dev);
2294     VMXNET3State *s = VMXNET3(pci_dev);
2295     int ret;
2296 
2297     VMW_CBPRN("Starting init...");
2298 
2299     memory_region_init_io(&s->bar0, OBJECT(s), &b0_ops, s,
2300                           "vmxnet3-b0", VMXNET3_PT_REG_SIZE);
2301     pci_register_bar(pci_dev, VMXNET3_BAR0_IDX,
2302                      PCI_BASE_ADDRESS_SPACE_MEMORY, &s->bar0);
2303 
2304     memory_region_init_io(&s->bar1, OBJECT(s), &b1_ops, s,
2305                           "vmxnet3-b1", VMXNET3_VD_REG_SIZE);
2306     pci_register_bar(pci_dev, VMXNET3_BAR1_IDX,
2307                      PCI_BASE_ADDRESS_SPACE_MEMORY, &s->bar1);
2308 
2309     memory_region_init(&s->msix_bar, OBJECT(s), "vmxnet3-msix-bar",
2310                        VMXNET3_MSIX_BAR_SIZE);
2311     pci_register_bar(pci_dev, VMXNET3_MSIX_BAR_IDX,
2312                      PCI_BASE_ADDRESS_SPACE_MEMORY, &s->msix_bar);
2313 
2314     vmxnet3_reset_interrupt_states(s);
2315 
2316     /* Interrupt pin A */
2317     pci_dev->config[PCI_INTERRUPT_PIN] = 0x01;
2318 
2319     ret = msi_init(pci_dev, VMXNET3_MSI_OFFSET(s), VMXNET3_MAX_NMSIX_INTRS,
2320                    VMXNET3_USE_64BIT, VMXNET3_PER_VECTOR_MASK, NULL);
2321     /* Any error other than -ENOTSUP(board's MSI support is broken)
2322      * is a programming error. Fall back to INTx silently on -ENOTSUP */
2323     assert(!ret || ret == -ENOTSUP);
2324 
2325     if (!vmxnet3_init_msix(s)) {
2326         VMW_WRPRN("Failed to initialize MSI-X, configuration is inconsistent.");
2327     }
2328 
2329     vmxnet3_net_init(s);
2330 
2331     if (pci_is_express(pci_dev)) {
2332         if (pci_bus_is_express(pci_dev->bus)) {
2333             pcie_endpoint_cap_init(pci_dev, VMXNET3_EXP_EP_OFFSET);
2334         }
2335 
2336         pcie_dev_ser_num_init(pci_dev, VMXNET3_DSN_OFFSET,
2337                               vmxnet3_device_serial_num(s));
2338     }
2339 
2340     register_savevm_live(dev, "vmxnet3-msix", -1, 1, &savevm_vmxnet3_msix, s);
2341 }
2342 
2343 static void vmxnet3_instance_init(Object *obj)
2344 {
2345     VMXNET3State *s = VMXNET3(obj);
2346     device_add_bootindex_property(obj, &s->conf.bootindex,
2347                                   "bootindex", "/ethernet-phy@0",
2348                                   DEVICE(obj), NULL);
2349 }
2350 
2351 static void vmxnet3_pci_uninit(PCIDevice *pci_dev)
2352 {
2353     DeviceState *dev = DEVICE(pci_dev);
2354     VMXNET3State *s = VMXNET3(pci_dev);
2355 
2356     VMW_CBPRN("Starting uninit...");
2357 
2358     unregister_savevm(dev, "vmxnet3-msix", s);
2359 
2360     vmxnet3_net_uninit(s);
2361 
2362     vmxnet3_cleanup_msix(s);
2363 
2364     vmxnet3_cleanup_msi(s);
2365 }
2366 
2367 static void vmxnet3_qdev_reset(DeviceState *dev)
2368 {
2369     PCIDevice *d = PCI_DEVICE(dev);
2370     VMXNET3State *s = VMXNET3(d);
2371 
2372     VMW_CBPRN("Starting QDEV reset...");
2373     vmxnet3_reset(s);
2374 }
2375 
2376 static bool vmxnet3_mc_list_needed(void *opaque)
2377 {
2378     return true;
2379 }
2380 
2381 static int vmxnet3_mcast_list_pre_load(void *opaque)
2382 {
2383     VMXNET3State *s = opaque;
2384 
2385     s->mcast_list = g_malloc(s->mcast_list_buff_size);
2386 
2387     return 0;
2388 }
2389 
2390 
2391 static void vmxnet3_pre_save(void *opaque)
2392 {
2393     VMXNET3State *s = opaque;
2394 
2395     s->mcast_list_buff_size = s->mcast_list_len * sizeof(MACAddr);
2396 }
2397 
2398 static const VMStateDescription vmxstate_vmxnet3_mcast_list = {
2399     .name = "vmxnet3/mcast_list",
2400     .version_id = 1,
2401     .minimum_version_id = 1,
2402     .pre_load = vmxnet3_mcast_list_pre_load,
2403     .needed = vmxnet3_mc_list_needed,
2404     .fields = (VMStateField[]) {
2405         VMSTATE_VBUFFER_UINT32(mcast_list, VMXNET3State, 0, NULL,
2406             mcast_list_buff_size),
2407         VMSTATE_END_OF_LIST()
2408     }
2409 };
2410 
2411 static const VMStateDescription vmstate_vmxnet3_ring = {
2412     .name = "vmxnet3-ring",
2413     .version_id = 0,
2414     .fields = (VMStateField[]) {
2415         VMSTATE_UINT64(pa, Vmxnet3Ring),
2416         VMSTATE_UINT32(size, Vmxnet3Ring),
2417         VMSTATE_UINT32(cell_size, Vmxnet3Ring),
2418         VMSTATE_UINT32(next, Vmxnet3Ring),
2419         VMSTATE_UINT8(gen, Vmxnet3Ring),
2420         VMSTATE_END_OF_LIST()
2421     }
2422 };
2423 
2424 static const VMStateDescription vmstate_vmxnet3_tx_stats = {
2425     .name = "vmxnet3-tx-stats",
2426     .version_id = 0,
2427     .fields = (VMStateField[]) {
2428         VMSTATE_UINT64(TSOPktsTxOK, struct UPT1_TxStats),
2429         VMSTATE_UINT64(TSOBytesTxOK, struct UPT1_TxStats),
2430         VMSTATE_UINT64(ucastPktsTxOK, struct UPT1_TxStats),
2431         VMSTATE_UINT64(ucastBytesTxOK, struct UPT1_TxStats),
2432         VMSTATE_UINT64(mcastPktsTxOK, struct UPT1_TxStats),
2433         VMSTATE_UINT64(mcastBytesTxOK, struct UPT1_TxStats),
2434         VMSTATE_UINT64(bcastPktsTxOK, struct UPT1_TxStats),
2435         VMSTATE_UINT64(bcastBytesTxOK, struct UPT1_TxStats),
2436         VMSTATE_UINT64(pktsTxError, struct UPT1_TxStats),
2437         VMSTATE_UINT64(pktsTxDiscard, struct UPT1_TxStats),
2438         VMSTATE_END_OF_LIST()
2439     }
2440 };
2441 
2442 static const VMStateDescription vmstate_vmxnet3_txq_descr = {
2443     .name = "vmxnet3-txq-descr",
2444     .version_id = 0,
2445     .fields = (VMStateField[]) {
2446         VMSTATE_STRUCT(tx_ring, Vmxnet3TxqDescr, 0, vmstate_vmxnet3_ring,
2447                        Vmxnet3Ring),
2448         VMSTATE_STRUCT(comp_ring, Vmxnet3TxqDescr, 0, vmstate_vmxnet3_ring,
2449                        Vmxnet3Ring),
2450         VMSTATE_UINT8(intr_idx, Vmxnet3TxqDescr),
2451         VMSTATE_UINT64(tx_stats_pa, Vmxnet3TxqDescr),
2452         VMSTATE_STRUCT(txq_stats, Vmxnet3TxqDescr, 0, vmstate_vmxnet3_tx_stats,
2453                        struct UPT1_TxStats),
2454         VMSTATE_END_OF_LIST()
2455     }
2456 };
2457 
2458 static const VMStateDescription vmstate_vmxnet3_rx_stats = {
2459     .name = "vmxnet3-rx-stats",
2460     .version_id = 0,
2461     .fields = (VMStateField[]) {
2462         VMSTATE_UINT64(LROPktsRxOK, struct UPT1_RxStats),
2463         VMSTATE_UINT64(LROBytesRxOK, struct UPT1_RxStats),
2464         VMSTATE_UINT64(ucastPktsRxOK, struct UPT1_RxStats),
2465         VMSTATE_UINT64(ucastBytesRxOK, struct UPT1_RxStats),
2466         VMSTATE_UINT64(mcastPktsRxOK, struct UPT1_RxStats),
2467         VMSTATE_UINT64(mcastBytesRxOK, struct UPT1_RxStats),
2468         VMSTATE_UINT64(bcastPktsRxOK, struct UPT1_RxStats),
2469         VMSTATE_UINT64(bcastBytesRxOK, struct UPT1_RxStats),
2470         VMSTATE_UINT64(pktsRxOutOfBuf, struct UPT1_RxStats),
2471         VMSTATE_UINT64(pktsRxError, struct UPT1_RxStats),
2472         VMSTATE_END_OF_LIST()
2473     }
2474 };
2475 
2476 static const VMStateDescription vmstate_vmxnet3_rxq_descr = {
2477     .name = "vmxnet3-rxq-descr",
2478     .version_id = 0,
2479     .fields = (VMStateField[]) {
2480         VMSTATE_STRUCT_ARRAY(rx_ring, Vmxnet3RxqDescr,
2481                              VMXNET3_RX_RINGS_PER_QUEUE, 0,
2482                              vmstate_vmxnet3_ring, Vmxnet3Ring),
2483         VMSTATE_STRUCT(comp_ring, Vmxnet3RxqDescr, 0, vmstate_vmxnet3_ring,
2484                        Vmxnet3Ring),
2485         VMSTATE_UINT8(intr_idx, Vmxnet3RxqDescr),
2486         VMSTATE_UINT64(rx_stats_pa, Vmxnet3RxqDescr),
2487         VMSTATE_STRUCT(rxq_stats, Vmxnet3RxqDescr, 0, vmstate_vmxnet3_rx_stats,
2488                        struct UPT1_RxStats),
2489         VMSTATE_END_OF_LIST()
2490     }
2491 };
2492 
2493 static int vmxnet3_post_load(void *opaque, int version_id)
2494 {
2495     VMXNET3State *s = opaque;
2496     PCIDevice *d = PCI_DEVICE(s);
2497 
2498     net_tx_pkt_init(&s->tx_pkt, PCI_DEVICE(s),
2499                     s->max_tx_frags, s->peer_has_vhdr);
2500     net_rx_pkt_init(&s->rx_pkt, s->peer_has_vhdr);
2501 
2502     if (s->msix_used) {
2503         if  (!vmxnet3_use_msix_vectors(s, VMXNET3_MAX_INTRS)) {
2504             VMW_WRPRN("Failed to re-use MSI-X vectors");
2505             msix_uninit(d, &s->msix_bar, &s->msix_bar);
2506             s->msix_used = false;
2507             return -1;
2508         }
2509     }
2510 
2511     vmxnet3_validate_queues(s);
2512     vmxnet3_validate_interrupts(s);
2513 
2514     return 0;
2515 }
2516 
2517 static const VMStateDescription vmstate_vmxnet3_int_state = {
2518     .name = "vmxnet3-int-state",
2519     .version_id = 0,
2520     .fields = (VMStateField[]) {
2521         VMSTATE_BOOL(is_masked, Vmxnet3IntState),
2522         VMSTATE_BOOL(is_pending, Vmxnet3IntState),
2523         VMSTATE_BOOL(is_asserted, Vmxnet3IntState),
2524         VMSTATE_END_OF_LIST()
2525     }
2526 };
2527 
2528 static bool vmxnet3_vmstate_need_pcie_device(void *opaque)
2529 {
2530     VMXNET3State *s = VMXNET3(opaque);
2531 
2532     return !(s->compat_flags & VMXNET3_COMPAT_FLAG_DISABLE_PCIE);
2533 }
2534 
2535 static bool vmxnet3_vmstate_test_pci_device(void *opaque, int version_id)
2536 {
2537     return !vmxnet3_vmstate_need_pcie_device(opaque);
2538 }
2539 
2540 static const VMStateDescription vmstate_vmxnet3_pcie_device = {
2541     .name = "vmxnet3/pcie",
2542     .version_id = 1,
2543     .minimum_version_id = 1,
2544     .needed = vmxnet3_vmstate_need_pcie_device,
2545     .fields = (VMStateField[]) {
2546         VMSTATE_PCI_DEVICE(parent_obj, VMXNET3State),
2547         VMSTATE_END_OF_LIST()
2548     }
2549 };
2550 
2551 static const VMStateDescription vmstate_vmxnet3 = {
2552     .name = "vmxnet3",
2553     .version_id = 1,
2554     .minimum_version_id = 1,
2555     .pre_save = vmxnet3_pre_save,
2556     .post_load = vmxnet3_post_load,
2557     .fields = (VMStateField[]) {
2558             VMSTATE_STRUCT_TEST(parent_obj, VMXNET3State,
2559                                 vmxnet3_vmstate_test_pci_device, 0,
2560                                 vmstate_pci_device, PCIDevice),
2561             VMSTATE_BOOL(rx_packets_compound, VMXNET3State),
2562             VMSTATE_BOOL(rx_vlan_stripping, VMXNET3State),
2563             VMSTATE_BOOL(lro_supported, VMXNET3State),
2564             VMSTATE_UINT32(rx_mode, VMXNET3State),
2565             VMSTATE_UINT32(mcast_list_len, VMXNET3State),
2566             VMSTATE_UINT32(mcast_list_buff_size, VMXNET3State),
2567             VMSTATE_UINT32_ARRAY(vlan_table, VMXNET3State, VMXNET3_VFT_SIZE),
2568             VMSTATE_UINT32(mtu, VMXNET3State),
2569             VMSTATE_UINT16(max_rx_frags, VMXNET3State),
2570             VMSTATE_UINT32(max_tx_frags, VMXNET3State),
2571             VMSTATE_UINT8(event_int_idx, VMXNET3State),
2572             VMSTATE_BOOL(auto_int_masking, VMXNET3State),
2573             VMSTATE_UINT8(txq_num, VMXNET3State),
2574             VMSTATE_UINT8(rxq_num, VMXNET3State),
2575             VMSTATE_UINT32(device_active, VMXNET3State),
2576             VMSTATE_UINT32(last_command, VMXNET3State),
2577             VMSTATE_UINT32(link_status_and_speed, VMXNET3State),
2578             VMSTATE_UINT32(temp_mac, VMXNET3State),
2579             VMSTATE_UINT64(drv_shmem, VMXNET3State),
2580             VMSTATE_UINT64(temp_shared_guest_driver_memory, VMXNET3State),
2581 
2582             VMSTATE_STRUCT_ARRAY(txq_descr, VMXNET3State,
2583                 VMXNET3_DEVICE_MAX_TX_QUEUES, 0, vmstate_vmxnet3_txq_descr,
2584                 Vmxnet3TxqDescr),
2585             VMSTATE_STRUCT_ARRAY(rxq_descr, VMXNET3State,
2586                 VMXNET3_DEVICE_MAX_RX_QUEUES, 0, vmstate_vmxnet3_rxq_descr,
2587                 Vmxnet3RxqDescr),
2588             VMSTATE_STRUCT_ARRAY(interrupt_states, VMXNET3State,
2589                 VMXNET3_MAX_INTRS, 0, vmstate_vmxnet3_int_state,
2590                 Vmxnet3IntState),
2591 
2592             VMSTATE_END_OF_LIST()
2593     },
2594     .subsections = (const VMStateDescription*[]) {
2595         &vmxstate_vmxnet3_mcast_list,
2596         &vmstate_vmxnet3_pcie_device,
2597         NULL
2598     }
2599 };
2600 
2601 static Property vmxnet3_properties[] = {
2602     DEFINE_NIC_PROPERTIES(VMXNET3State, conf),
2603     DEFINE_PROP_BIT("x-old-msi-offsets", VMXNET3State, compat_flags,
2604                     VMXNET3_COMPAT_FLAG_OLD_MSI_OFFSETS_BIT, false),
2605     DEFINE_PROP_BIT("x-disable-pcie", VMXNET3State, compat_flags,
2606                     VMXNET3_COMPAT_FLAG_DISABLE_PCIE_BIT, false),
2607     DEFINE_PROP_END_OF_LIST(),
2608 };
2609 
2610 static void vmxnet3_realize(DeviceState *qdev, Error **errp)
2611 {
2612     VMXNET3Class *vc = VMXNET3_DEVICE_GET_CLASS(qdev);
2613     PCIDevice *pci_dev = PCI_DEVICE(qdev);
2614     VMXNET3State *s = VMXNET3(qdev);
2615 
2616     if (!(s->compat_flags & VMXNET3_COMPAT_FLAG_DISABLE_PCIE)) {
2617         pci_dev->cap_present |= QEMU_PCI_CAP_EXPRESS;
2618     }
2619 
2620     vc->parent_dc_realize(qdev, errp);
2621 }
2622 
2623 static void vmxnet3_class_init(ObjectClass *class, void *data)
2624 {
2625     DeviceClass *dc = DEVICE_CLASS(class);
2626     PCIDeviceClass *c = PCI_DEVICE_CLASS(class);
2627     VMXNET3Class *vc = VMXNET3_DEVICE_CLASS(class);
2628 
2629     c->realize = vmxnet3_pci_realize;
2630     c->exit = vmxnet3_pci_uninit;
2631     c->vendor_id = PCI_VENDOR_ID_VMWARE;
2632     c->device_id = PCI_DEVICE_ID_VMWARE_VMXNET3;
2633     c->revision = PCI_DEVICE_ID_VMWARE_VMXNET3_REVISION;
2634     c->romfile = "efi-vmxnet3.rom";
2635     c->class_id = PCI_CLASS_NETWORK_ETHERNET;
2636     c->subsystem_vendor_id = PCI_VENDOR_ID_VMWARE;
2637     c->subsystem_id = PCI_DEVICE_ID_VMWARE_VMXNET3;
2638     vc->parent_dc_realize = dc->realize;
2639     dc->realize = vmxnet3_realize;
2640     dc->desc = "VMWare Paravirtualized Ethernet v3";
2641     dc->reset = vmxnet3_qdev_reset;
2642     dc->vmsd = &vmstate_vmxnet3;
2643     dc->props = vmxnet3_properties;
2644     set_bit(DEVICE_CATEGORY_NETWORK, dc->categories);
2645 }
2646 
2647 static const TypeInfo vmxnet3_info = {
2648     .name          = TYPE_VMXNET3,
2649     .parent        = TYPE_PCI_DEVICE,
2650     .class_size    = sizeof(VMXNET3Class),
2651     .instance_size = sizeof(VMXNET3State),
2652     .class_init    = vmxnet3_class_init,
2653     .instance_init = vmxnet3_instance_init,
2654 };
2655 
2656 static void vmxnet3_register_types(void)
2657 {
2658     VMW_CBPRN("vmxnet3_register_types called...");
2659     type_register_static(&vmxnet3_info);
2660 }
2661 
2662 type_init(vmxnet3_register_types)
2663