?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
usb: use local path for local headers
[qemu.git] / hw / net / vmxnet3.c
blob364863038637d0a6579284b8ed066582d65bb8ad
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 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 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 txcq_descr.val1 = cpu_to_le32(txcq_descr.val1);
539 txcq_descr.val2 = cpu_to_le32(txcq_descr.val2);
540 vmxnet3_ring_write_curr_cell(d, &s->txq_descr[qidx].comp_ring, &txcq_descr);
541
542 /* Flush changes in TX descriptor before changing the counter value */
543 smp_wmb();
544
545 vmxnet3_inc_tx_completion_counter(s, qidx);
546 vmxnet3_trigger_interrupt(s, s->txq_descr[qidx].intr_idx);
547 }
548
549 static bool
550 vmxnet3_setup_tx_offloads(VMXNET3State *s)
551 {
552 switch (s->offload_mode) {
553 case VMXNET3_OM_NONE:
554 net_tx_pkt_build_vheader(s->tx_pkt, false, false, 0);
555 break;
556
557 case VMXNET3_OM_CSUM:
558 net_tx_pkt_build_vheader(s->tx_pkt, false, true, 0);
559 VMW_PKPRN("L4 CSO requested\n");
560 break;
561
562 case VMXNET3_OM_TSO:
563 net_tx_pkt_build_vheader(s->tx_pkt, true, true,
564 s->cso_or_gso_size);
565 net_tx_pkt_update_ip_checksums(s->tx_pkt);
566 VMW_PKPRN("GSO offload requested.");
567 break;
568
569 default:
570 g_assert_not_reached();
571 return false;
572 }
573
574 return true;
575 }
576
577 static void
578 vmxnet3_tx_retrieve_metadata(VMXNET3State *s,
579 const struct Vmxnet3_TxDesc *txd)
580 {
581 s->offload_mode = txd->om;
582 s->cso_or_gso_size = txd->msscof;
583 s->tci = txd->tci;
584 s->needs_vlan = txd->ti;
585 }
586
587 typedef enum {
588 VMXNET3_PKT_STATUS_OK,
589 VMXNET3_PKT_STATUS_ERROR,
590 VMXNET3_PKT_STATUS_DISCARD,/* only for tx */
591 VMXNET3_PKT_STATUS_OUT_OF_BUF /* only for rx */
592 } Vmxnet3PktStatus;
593
594 static void
595 vmxnet3_on_tx_done_update_stats(VMXNET3State *s, int qidx,
596 Vmxnet3PktStatus status)
597 {
598 size_t tot_len = net_tx_pkt_get_total_len(s->tx_pkt);
599 struct UPT1_TxStats *stats = &s->txq_descr[qidx].txq_stats;
600
601 switch (status) {
602 case VMXNET3_PKT_STATUS_OK:
603 switch (net_tx_pkt_get_packet_type(s->tx_pkt)) {
604 case ETH_PKT_BCAST:
605 stats->bcastPktsTxOK++;
606 stats->bcastBytesTxOK += tot_len;
607 break;
608 case ETH_PKT_MCAST:
609 stats->mcastPktsTxOK++;
610 stats->mcastBytesTxOK += tot_len;
611 break;
612 case ETH_PKT_UCAST:
613 stats->ucastPktsTxOK++;
614 stats->ucastBytesTxOK += tot_len;
615 break;
616 default:
617 g_assert_not_reached();
618 }
619
620 if (s->offload_mode == VMXNET3_OM_TSO) {
621 /*
622 * According to VMWARE headers this statistic is a number
623 * of packets after segmentation but since we don't have
624 * this information in QEMU model, the best we can do is to
625 * provide number of non-segmented packets
626 */
627 stats->TSOPktsTxOK++;
628 stats->TSOBytesTxOK += tot_len;
629 }
630 break;
631
632 case VMXNET3_PKT_STATUS_DISCARD:
633 stats->pktsTxDiscard++;
634 break;
635
636 case VMXNET3_PKT_STATUS_ERROR:
637 stats->pktsTxError++;
638 break;
639
640 default:
641 g_assert_not_reached();
642 }
643 }
644
645 static void
646 vmxnet3_on_rx_done_update_stats(VMXNET3State *s,
647 int qidx,
648 Vmxnet3PktStatus status)
649 {
650 struct UPT1_RxStats *stats = &s->rxq_descr[qidx].rxq_stats;
651 size_t tot_len = net_rx_pkt_get_total_len(s->rx_pkt);
652
653 switch (status) {
654 case VMXNET3_PKT_STATUS_OUT_OF_BUF:
655 stats->pktsRxOutOfBuf++;
656 break;
657
658 case VMXNET3_PKT_STATUS_ERROR:
659 stats->pktsRxError++;
660 break;
661 case VMXNET3_PKT_STATUS_OK:
662 switch (net_rx_pkt_get_packet_type(s->rx_pkt)) {
663 case ETH_PKT_BCAST:
664 stats->bcastPktsRxOK++;
665 stats->bcastBytesRxOK += tot_len;
666 break;
667 case ETH_PKT_MCAST:
668 stats->mcastPktsRxOK++;
669 stats->mcastBytesRxOK += tot_len;
670 break;
671 case ETH_PKT_UCAST:
672 stats->ucastPktsRxOK++;
673 stats->ucastBytesRxOK += tot_len;
674 break;
675 default:
676 g_assert_not_reached();
677 }
678
679 if (tot_len > s->mtu) {
680 stats->LROPktsRxOK++;
681 stats->LROBytesRxOK += tot_len;
682 }
683 break;
684 default:
685 g_assert_not_reached();
686 }
687 }
688
689 static inline void
690 vmxnet3_ring_read_curr_txdesc(PCIDevice *pcidev, Vmxnet3Ring *ring,
691 struct Vmxnet3_TxDesc *txd)
692 {
693 vmxnet3_ring_read_curr_cell(pcidev, ring, txd);
694 txd->addr = le64_to_cpu(txd->addr);
695 txd->val1 = le32_to_cpu(txd->val1);
696 txd->val2 = le32_to_cpu(txd->val2);
697 }
698
699 static inline bool
700 vmxnet3_pop_next_tx_descr(VMXNET3State *s,
701 int qidx,
702 struct Vmxnet3_TxDesc *txd,
703 uint32_t *descr_idx)
704 {
705 Vmxnet3Ring *ring = &s->txq_descr[qidx].tx_ring;
706 PCIDevice *d = PCI_DEVICE(s);
707
708 vmxnet3_ring_read_curr_txdesc(d, ring, txd);
709 if (txd->gen == vmxnet3_ring_curr_gen(ring)) {
710 /* Only read after generation field verification */
711 smp_rmb();
712 /* Re-read to be sure we got the latest version */
713 vmxnet3_ring_read_curr_txdesc(d, ring, txd);
714 VMXNET3_RING_DUMP(VMW_RIPRN, "TX", qidx, ring);
715 *descr_idx = vmxnet3_ring_curr_cell_idx(ring);
716 vmxnet3_inc_tx_consumption_counter(s, qidx);
717 return true;
718 }
719
720 return false;
721 }
722
723 static bool
724 vmxnet3_send_packet(VMXNET3State *s, uint32_t qidx)
725 {
726 Vmxnet3PktStatus status = VMXNET3_PKT_STATUS_OK;
727
728 if (!vmxnet3_setup_tx_offloads(s)) {
729 status = VMXNET3_PKT_STATUS_ERROR;
730 goto func_exit;
731 }
732
733 /* debug prints */
734 vmxnet3_dump_virt_hdr(net_tx_pkt_get_vhdr(s->tx_pkt));
735 net_tx_pkt_dump(s->tx_pkt);
736
737 if (!net_tx_pkt_send(s->tx_pkt, qemu_get_queue(s->nic))) {
738 status = VMXNET3_PKT_STATUS_DISCARD;
739 goto func_exit;
740 }
741
742 func_exit:
743 vmxnet3_on_tx_done_update_stats(s, qidx, status);
744 return (status == VMXNET3_PKT_STATUS_OK);
745 }
746
747 static void vmxnet3_process_tx_queue(VMXNET3State *s, int qidx)
748 {
749 struct Vmxnet3_TxDesc txd;
750 uint32_t txd_idx;
751 uint32_t data_len;
752 hwaddr data_pa;
753
754 for (;;) {
755 if (!vmxnet3_pop_next_tx_descr(s, qidx, &txd, &txd_idx)) {
756 break;
757 }
758
759 vmxnet3_dump_tx_descr(&txd);
760
761 if (!s->skip_current_tx_pkt) {
762 data_len = (txd.len > 0) ? txd.len : VMXNET3_MAX_TX_BUF_SIZE;
763 data_pa = txd.addr;
764
765 if (!net_tx_pkt_add_raw_fragment(s->tx_pkt,
766 data_pa,
767 data_len)) {
768 s->skip_current_tx_pkt = true;
769 }
770 }
771
772 if (s->tx_sop) {
773 vmxnet3_tx_retrieve_metadata(s, &txd);
774 s->tx_sop = false;
775 }
776
777 if (txd.eop) {
778 if (!s->skip_current_tx_pkt && net_tx_pkt_parse(s->tx_pkt)) {
779 if (s->needs_vlan) {
780 net_tx_pkt_setup_vlan_header(s->tx_pkt, s->tci);
781 }
782
783 vmxnet3_send_packet(s, qidx);
784 } else {
785 vmxnet3_on_tx_done_update_stats(s, qidx,
786 VMXNET3_PKT_STATUS_ERROR);
787 }
788
789 vmxnet3_complete_packet(s, qidx, txd_idx);
790 s->tx_sop = true;
791 s->skip_current_tx_pkt = false;
792 net_tx_pkt_reset(s->tx_pkt);
793 }
794 }
795 }
796
797 static inline void
798 vmxnet3_read_next_rx_descr(VMXNET3State *s, int qidx, int ridx,
799 struct Vmxnet3_RxDesc *dbuf, uint32_t *didx)
800 {
801 PCIDevice *d = PCI_DEVICE(s);
802
803 Vmxnet3Ring *ring = &s->rxq_descr[qidx].rx_ring[ridx];
804 *didx = vmxnet3_ring_curr_cell_idx(ring);
805 vmxnet3_ring_read_curr_cell(d, ring, dbuf);
806 dbuf->addr = le64_to_cpu(dbuf->addr);
807 dbuf->val1 = le32_to_cpu(dbuf->val1);
808 dbuf->ext1 = le32_to_cpu(dbuf->ext1);
809 }
810
811 static inline uint8_t
812 vmxnet3_get_rx_ring_gen(VMXNET3State *s, int qidx, int ridx)
813 {
814 return s->rxq_descr[qidx].rx_ring[ridx].gen;
815 }
816
817 static inline hwaddr
818 vmxnet3_pop_rxc_descr(VMXNET3State *s, int qidx, uint32_t *descr_gen)
819 {
820 uint8_t ring_gen;
821 struct Vmxnet3_RxCompDesc rxcd;
822
823 hwaddr daddr =
824 vmxnet3_ring_curr_cell_pa(&s->rxq_descr[qidx].comp_ring);
825
826 pci_dma_read(PCI_DEVICE(s),
827 daddr, &rxcd, sizeof(struct Vmxnet3_RxCompDesc));
828 rxcd.val1 = le32_to_cpu(rxcd.val1);
829 rxcd.val2 = le32_to_cpu(rxcd.val2);
830 rxcd.val3 = le32_to_cpu(rxcd.val3);
831 ring_gen = vmxnet3_ring_curr_gen(&s->rxq_descr[qidx].comp_ring);
832
833 if (rxcd.gen != ring_gen) {
834 *descr_gen = ring_gen;
835 vmxnet3_inc_rx_completion_counter(s, qidx);
836 return daddr;
837 }
838
839 return 0;
840 }
841
842 static inline void
843 vmxnet3_revert_rxc_descr(VMXNET3State *s, int qidx)
844 {
845 vmxnet3_dec_rx_completion_counter(s, qidx);
846 }
847
848 #define RXQ_IDX (0)
849 #define RX_HEAD_BODY_RING (0)
850 #define RX_BODY_ONLY_RING (1)
851
852 static bool
853 vmxnet3_get_next_head_rx_descr(VMXNET3State *s,
854 struct Vmxnet3_RxDesc *descr_buf,
855 uint32_t *descr_idx,
856 uint32_t *ridx)
857 {
858 for (;;) {
859 uint32_t ring_gen;
860 vmxnet3_read_next_rx_descr(s, RXQ_IDX, RX_HEAD_BODY_RING,
861 descr_buf, descr_idx);
862
863 /* If no more free descriptors - return */
864 ring_gen = vmxnet3_get_rx_ring_gen(s, RXQ_IDX, RX_HEAD_BODY_RING);
865 if (descr_buf->gen != ring_gen) {
866 return false;
867 }
868
869 /* Only read after generation field verification */
870 smp_rmb();
871 /* Re-read to be sure we got the latest version */
872 vmxnet3_read_next_rx_descr(s, RXQ_IDX, RX_HEAD_BODY_RING,
873 descr_buf, descr_idx);
874
875 /* Mark current descriptor as used/skipped */
876 vmxnet3_inc_rx_consumption_counter(s, RXQ_IDX, RX_HEAD_BODY_RING);
877
878 /* If this is what we are looking for - return */
879 if (descr_buf->btype == VMXNET3_RXD_BTYPE_HEAD) {
880 *ridx = RX_HEAD_BODY_RING;
881 return true;
882 }
883 }
884 }
885
886 static bool
887 vmxnet3_get_next_body_rx_descr(VMXNET3State *s,
888 struct Vmxnet3_RxDesc *d,
889 uint32_t *didx,
890 uint32_t *ridx)
891 {
892 vmxnet3_read_next_rx_descr(s, RXQ_IDX, RX_HEAD_BODY_RING, d, didx);
893
894 /* Try to find corresponding descriptor in head/body ring */
895 if (d->gen == vmxnet3_get_rx_ring_gen(s, RXQ_IDX, RX_HEAD_BODY_RING)) {
896 /* Only read after generation field verification */
897 smp_rmb();
898 /* Re-read to be sure we got the latest version */
899 vmxnet3_read_next_rx_descr(s, RXQ_IDX, RX_HEAD_BODY_RING, d, didx);
900 if (d->btype == VMXNET3_RXD_BTYPE_BODY) {
901 vmxnet3_inc_rx_consumption_counter(s, RXQ_IDX, RX_HEAD_BODY_RING);
902 *ridx = RX_HEAD_BODY_RING;
903 return true;
904 }
905 }
906
907 /*
908 * If there is no free descriptors on head/body ring or next free
909 * descriptor is a head descriptor switch to body only ring
910 */
911 vmxnet3_read_next_rx_descr(s, RXQ_IDX, RX_BODY_ONLY_RING, d, didx);
912
913 /* If no more free descriptors - return */
914 if (d->gen == vmxnet3_get_rx_ring_gen(s, RXQ_IDX, RX_BODY_ONLY_RING)) {
915 /* Only read after generation field verification */
916 smp_rmb();
917 /* Re-read to be sure we got the latest version */
918 vmxnet3_read_next_rx_descr(s, RXQ_IDX, RX_BODY_ONLY_RING, d, didx);
919 assert(d->btype == VMXNET3_RXD_BTYPE_BODY);
920 *ridx = RX_BODY_ONLY_RING;
921 vmxnet3_inc_rx_consumption_counter(s, RXQ_IDX, RX_BODY_ONLY_RING);
922 return true;
923 }
924
925 return false;
926 }
927
928 static inline bool
929 vmxnet3_get_next_rx_descr(VMXNET3State *s, bool is_head,
930 struct Vmxnet3_RxDesc *descr_buf,
931 uint32_t *descr_idx,
932 uint32_t *ridx)
933 {
934 if (is_head || !s->rx_packets_compound) {
935 return vmxnet3_get_next_head_rx_descr(s, descr_buf, descr_idx, ridx);
936 } else {
937 return vmxnet3_get_next_body_rx_descr(s, descr_buf, descr_idx, ridx);
938 }
939 }
940
941 /* In case packet was csum offloaded (either NEEDS_CSUM or DATA_VALID),
942 * the implementation always passes an RxCompDesc with a "Checksum
943 * calculated and found correct" to the OS (cnc=0 and tuc=1, see
944 * vmxnet3_rx_update_descr). This emulates the observed ESXi behavior.
945 *
946 * Therefore, if packet has the NEEDS_CSUM set, we must calculate
947 * and place a fully computed checksum into the tcp/udp header.
948 * Otherwise, the OS driver will receive a checksum-correct indication
949 * (CHECKSUM_UNNECESSARY), but with the actual tcp/udp checksum field
950 * having just the pseudo header csum value.
951 *
952 * While this is not a problem if packet is destined for local delivery,
953 * in the case the host OS performs forwarding, it will forward an
954 * incorrectly checksummed packet.
955 */
956 static void vmxnet3_rx_need_csum_calculate(struct NetRxPkt *pkt,
957 const void *pkt_data,
958 size_t pkt_len)
959 {
960 struct virtio_net_hdr *vhdr;
961 bool isip4, isip6, istcp, isudp;
962 uint8_t *data;
963 int len;
964
965 if (!net_rx_pkt_has_virt_hdr(pkt)) {
966 return;
967 }
968
969 vhdr = net_rx_pkt_get_vhdr(pkt);
970 if (!VMXNET_FLAG_IS_SET(vhdr->flags, VIRTIO_NET_HDR_F_NEEDS_CSUM)) {
971 return;
972 }
973
974 net_rx_pkt_get_protocols(pkt, &isip4, &isip6, &isudp, &istcp);
975 if (!(isip4 || isip6) || !(istcp || isudp)) {
976 return;
977 }
978
979 vmxnet3_dump_virt_hdr(vhdr);
980
981 /* Validate packet len: csum_start + scum_offset + length of csum field */
982 if (pkt_len < (vhdr->csum_start + vhdr->csum_offset + 2)) {
983 VMW_PKPRN("packet len:%zu < csum_start(%d) + csum_offset(%d) + 2, "
984 "cannot calculate checksum",
985 pkt_len, vhdr->csum_start, vhdr->csum_offset);
986 return;
987 }
988
989 data = (uint8_t *)pkt_data + vhdr->csum_start;
990 len = pkt_len - vhdr->csum_start;
991 /* Put the checksum obtained into the packet */
992 stw_be_p(data + vhdr->csum_offset,
993 net_checksum_finish_nozero(net_checksum_add(len, data)));
994
995 vhdr->flags &= ~VIRTIO_NET_HDR_F_NEEDS_CSUM;
996 vhdr->flags |= VIRTIO_NET_HDR_F_DATA_VALID;
997 }
998
999 static void vmxnet3_rx_update_descr(struct NetRxPkt *pkt,
1000 struct Vmxnet3_RxCompDesc *rxcd)
1001 {
1002 int csum_ok, is_gso;
1003 bool isip4, isip6, istcp, isudp;
1004 struct virtio_net_hdr *vhdr;
1005 uint8_t offload_type;
1006
1007 if (net_rx_pkt_is_vlan_stripped(pkt)) {
1008 rxcd->ts = 1;
1009 rxcd->tci = net_rx_pkt_get_vlan_tag(pkt);
1010 }
1011
1012 if (!net_rx_pkt_has_virt_hdr(pkt)) {
1013 goto nocsum;
1014 }
1015
1016 vhdr = net_rx_pkt_get_vhdr(pkt);
1017 /*
1018 * Checksum is valid when lower level tell so or when lower level
1019 * requires checksum offload telling that packet produced/bridged
1020 * locally and did travel over network after last checksum calculation
1021 * or production
1022 */
1023 csum_ok = VMXNET_FLAG_IS_SET(vhdr->flags, VIRTIO_NET_HDR_F_DATA_VALID) ||
1024 VMXNET_FLAG_IS_SET(vhdr->flags, VIRTIO_NET_HDR_F_NEEDS_CSUM);
1025
1026 offload_type = vhdr->gso_type & ~VIRTIO_NET_HDR_GSO_ECN;
1027 is_gso = (offload_type != VIRTIO_NET_HDR_GSO_NONE) ? 1 : 0;
1028
1029 if (!csum_ok && !is_gso) {
1030 goto nocsum;
1031 }
1032
1033 net_rx_pkt_get_protocols(pkt, &isip4, &isip6, &isudp, &istcp);
1034 if ((!istcp && !isudp) || (!isip4 && !isip6)) {
1035 goto nocsum;
1036 }
1037
1038 rxcd->cnc = 0;
1039 rxcd->v4 = isip4 ? 1 : 0;
1040 rxcd->v6 = isip6 ? 1 : 0;
1041 rxcd->tcp = istcp ? 1 : 0;
1042 rxcd->udp = isudp ? 1 : 0;
1043 rxcd->fcs = rxcd->tuc = rxcd->ipc = 1;
1044 return;
1045
1046 nocsum:
1047 rxcd->cnc = 1;
1048 return;
1049 }
1050
1051 static void
1052 vmxnet3_pci_dma_writev(PCIDevice *pci_dev,
1053 const struct iovec *iov,
1054 size_t start_iov_off,
1055 hwaddr target_addr,
1056 size_t bytes_to_copy)
1057 {
1058 size_t curr_off = 0;
1059 size_t copied = 0;
1060
1061 while (bytes_to_copy) {
1062 if (start_iov_off < (curr_off + iov->iov_len)) {
1063 size_t chunk_len =
1064 MIN((curr_off + iov->iov_len) - start_iov_off, bytes_to_copy);
1065
1066 pci_dma_write(pci_dev, target_addr + copied,
1067 iov->iov_base + start_iov_off - curr_off,
1068 chunk_len);
1069
1070 copied += chunk_len;
1071 start_iov_off += chunk_len;
1072 curr_off = start_iov_off;
1073 bytes_to_copy -= chunk_len;
1074 } else {
1075 curr_off += iov->iov_len;
1076 }
1077 iov++;
1078 }
1079 }
1080
1081 static void
1082 vmxnet3_pci_dma_write_rxcd(PCIDevice *pcidev, dma_addr_t pa,
1083 struct Vmxnet3_RxCompDesc *rxcd)
1084 {
1085 rxcd->val1 = cpu_to_le32(rxcd->val1);
1086 rxcd->val2 = cpu_to_le32(rxcd->val2);
1087 rxcd->val3 = cpu_to_le32(rxcd->val3);
1088 pci_dma_write(pcidev, pa, rxcd, sizeof(*rxcd));
1089 }
1090
1091 static bool
1092 vmxnet3_indicate_packet(VMXNET3State *s)
1093 {
1094 struct Vmxnet3_RxDesc rxd;
1095 PCIDevice *d = PCI_DEVICE(s);
1096 bool is_head = true;
1097 uint32_t rxd_idx;
1098 uint32_t rx_ridx = 0;
1099
1100 struct Vmxnet3_RxCompDesc rxcd;
1101 uint32_t new_rxcd_gen = VMXNET3_INIT_GEN;
1102 hwaddr new_rxcd_pa = 0;
1103 hwaddr ready_rxcd_pa = 0;
1104 struct iovec *data = net_rx_pkt_get_iovec(s->rx_pkt);
1105 size_t bytes_copied = 0;
1106 size_t bytes_left = net_rx_pkt_get_total_len(s->rx_pkt);
1107 uint16_t num_frags = 0;
1108 size_t chunk_size;
1109
1110 net_rx_pkt_dump(s->rx_pkt);
1111
1112 while (bytes_left > 0) {
1113
1114 /* cannot add more frags to packet */
1115 if (num_frags == s->max_rx_frags) {
1116 break;
1117 }
1118
1119 new_rxcd_pa = vmxnet3_pop_rxc_descr(s, RXQ_IDX, &new_rxcd_gen);
1120 if (!new_rxcd_pa) {
1121 break;
1122 }
1123
1124 if (!vmxnet3_get_next_rx_descr(s, is_head, &rxd, &rxd_idx, &rx_ridx)) {
1125 break;
1126 }
1127
1128 chunk_size = MIN(bytes_left, rxd.len);
1129 vmxnet3_pci_dma_writev(d, data, bytes_copied, rxd.addr, chunk_size);
1130 bytes_copied += chunk_size;
1131 bytes_left -= chunk_size;
1132
1133 vmxnet3_dump_rx_descr(&rxd);
1134
1135 if (ready_rxcd_pa != 0) {
1136 vmxnet3_pci_dma_write_rxcd(d, ready_rxcd_pa, &rxcd);
1137 }
1138
1139 memset(&rxcd, 0, sizeof(struct Vmxnet3_RxCompDesc));
1140 rxcd.rxdIdx = rxd_idx;
1141 rxcd.len = chunk_size;
1142 rxcd.sop = is_head;
1143 rxcd.gen = new_rxcd_gen;
1144 rxcd.rqID = RXQ_IDX + rx_ridx * s->rxq_num;
1145
1146 if (bytes_left == 0) {
1147 vmxnet3_rx_update_descr(s->rx_pkt, &rxcd);
1148 }
1149
1150 VMW_RIPRN("RX Completion descriptor: rxRing: %lu rxIdx %lu len %lu "
1151 "sop %d csum_correct %lu",
1152 (unsigned long) rx_ridx,
1153 (unsigned long) rxcd.rxdIdx,
1154 (unsigned long) rxcd.len,
1155 (int) rxcd.sop,
1156 (unsigned long) rxcd.tuc);
1157
1158 is_head = false;
1159 ready_rxcd_pa = new_rxcd_pa;
1160 new_rxcd_pa = 0;
1161 num_frags++;
1162 }
1163
1164 if (ready_rxcd_pa != 0) {
1165 rxcd.eop = 1;
1166 rxcd.err = (bytes_left != 0);
1167
1168 vmxnet3_pci_dma_write_rxcd(d, ready_rxcd_pa, &rxcd);
1169
1170 /* Flush RX descriptor changes */
1171 smp_wmb();
1172 }
1173
1174 if (new_rxcd_pa != 0) {
1175 vmxnet3_revert_rxc_descr(s, RXQ_IDX);
1176 }
1177
1178 vmxnet3_trigger_interrupt(s, s->rxq_descr[RXQ_IDX].intr_idx);
1179
1180 if (bytes_left == 0) {
1181 vmxnet3_on_rx_done_update_stats(s, RXQ_IDX, VMXNET3_PKT_STATUS_OK);
1182 return true;
1183 } else if (num_frags == s->max_rx_frags) {
1184 vmxnet3_on_rx_done_update_stats(s, RXQ_IDX, VMXNET3_PKT_STATUS_ERROR);
1185 return false;
1186 } else {
1187 vmxnet3_on_rx_done_update_stats(s, RXQ_IDX,
1188 VMXNET3_PKT_STATUS_OUT_OF_BUF);
1189 return false;
1190 }
1191 }
1192
1193 static void
1194 vmxnet3_io_bar0_write(void *opaque, hwaddr addr,
1195 uint64_t val, unsigned size)
1196 {
1197 VMXNET3State *s = opaque;
1198
1199 if (!s->device_active) {
1200 return;
1201 }
1202
1203 if (VMW_IS_MULTIREG_ADDR(addr, VMXNET3_REG_TXPROD,
1204 VMXNET3_DEVICE_MAX_TX_QUEUES, VMXNET3_REG_ALIGN)) {
1205 int tx_queue_idx =
1206 VMW_MULTIREG_IDX_BY_ADDR(addr, VMXNET3_REG_TXPROD,
1207 VMXNET3_REG_ALIGN);
1208 assert(tx_queue_idx <= s->txq_num);
1209 vmxnet3_process_tx_queue(s, tx_queue_idx);
1210 return;
1211 }
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
1218 VMW_CBPRN("Interrupt mask for line %d written: 0x%" PRIx64, l, val);
1219
1220 vmxnet3_on_interrupt_mask_changed(s, l, val);
1221 return;
1222 }
1223
1224 if (VMW_IS_MULTIREG_ADDR(addr, VMXNET3_REG_RXPROD,
1225 VMXNET3_DEVICE_MAX_RX_QUEUES, VMXNET3_REG_ALIGN) ||
1226 VMW_IS_MULTIREG_ADDR(addr, VMXNET3_REG_RXPROD2,
1227 VMXNET3_DEVICE_MAX_RX_QUEUES, VMXNET3_REG_ALIGN)) {
1228 return;
1229 }
1230
1231 VMW_WRPRN("BAR0 unknown write [%" PRIx64 "] = %" PRIx64 ", size %d",
1232 (uint64_t) addr, val, size);
1233 }
1234
1235 static uint64_t
1236 vmxnet3_io_bar0_read(void *opaque, hwaddr addr, unsigned size)
1237 {
1238 VMXNET3State *s = opaque;
1239
1240 if (VMW_IS_MULTIREG_ADDR(addr, VMXNET3_REG_IMR,
1241 VMXNET3_MAX_INTRS, VMXNET3_REG_ALIGN)) {
1242 int l = VMW_MULTIREG_IDX_BY_ADDR(addr, VMXNET3_REG_IMR,
1243 VMXNET3_REG_ALIGN);
1244 return s->interrupt_states[l].is_masked;
1245 }
1246
1247 VMW_CBPRN("BAR0 unknown read [%" PRIx64 "], size %d", addr, size);
1248 return 0;
1249 }
1250
1251 static void vmxnet3_reset_interrupt_states(VMXNET3State *s)
1252 {
1253 int i;
1254 for (i = 0; i < ARRAY_SIZE(s->interrupt_states); i++) {
1255 s->interrupt_states[i].is_asserted = false;
1256 s->interrupt_states[i].is_pending = false;
1257 s->interrupt_states[i].is_masked = true;
1258 }
1259 }
1260
1261 static void vmxnet3_reset_mac(VMXNET3State *s)
1262 {
1263 memcpy(&s->conf.macaddr.a, &s->perm_mac.a, sizeof(s->perm_mac.a));
1264 VMW_CFPRN("MAC address set to: " MAC_FMT, MAC_ARG(s->conf.macaddr.a));
1265 }
1266
1267 static void vmxnet3_deactivate_device(VMXNET3State *s)
1268 {
1269 if (s->device_active) {
1270 VMW_CBPRN("Deactivating vmxnet3...");
1271 net_tx_pkt_reset(s->tx_pkt);
1272 net_tx_pkt_uninit(s->tx_pkt);
1273 net_rx_pkt_uninit(s->rx_pkt);
1274 s->device_active = false;
1275 }
1276 }
1277
1278 static void vmxnet3_reset(VMXNET3State *s)
1279 {
1280 VMW_CBPRN("Resetting vmxnet3...");
1281
1282 vmxnet3_deactivate_device(s);
1283 vmxnet3_reset_interrupt_states(s);
1284 s->drv_shmem = 0;
1285 s->tx_sop = true;
1286 s->skip_current_tx_pkt = false;
1287 }
1288
1289 static void vmxnet3_update_rx_mode(VMXNET3State *s)
1290 {
1291 PCIDevice *d = PCI_DEVICE(s);
1292
1293 s->rx_mode = VMXNET3_READ_DRV_SHARED32(d, s->drv_shmem,
1294 devRead.rxFilterConf.rxMode);
1295 VMW_CFPRN("RX mode: 0x%08X", s->rx_mode);
1296 }
1297
1298 static void vmxnet3_update_vlan_filters(VMXNET3State *s)
1299 {
1300 int i;
1301 PCIDevice *d = PCI_DEVICE(s);
1302
1303 /* Copy configuration from shared memory */
1304 VMXNET3_READ_DRV_SHARED(d, s->drv_shmem,
1305 devRead.rxFilterConf.vfTable,
1306 s->vlan_table,
1307 sizeof(s->vlan_table));
1308
1309 /* Invert byte order when needed */
1310 for (i = 0; i < ARRAY_SIZE(s->vlan_table); i++) {
1311 s->vlan_table[i] = le32_to_cpu(s->vlan_table[i]);
1312 }
1313
1314 /* Dump configuration for debugging purposes */
1315 VMW_CFPRN("Configured VLANs:");
1316 for (i = 0; i < sizeof(s->vlan_table) * 8; i++) {
1317 if (VMXNET3_VFTABLE_ENTRY_IS_SET(s->vlan_table, i)) {
1318 VMW_CFPRN("\tVLAN %d is present", i);
1319 }
1320 }
1321 }
1322
1323 static void vmxnet3_update_mcast_filters(VMXNET3State *s)
1324 {
1325 PCIDevice *d = PCI_DEVICE(s);
1326
1327 uint16_t list_bytes =
1328 VMXNET3_READ_DRV_SHARED16(d, s->drv_shmem,
1329 devRead.rxFilterConf.mfTableLen);
1330
1331 s->mcast_list_len = list_bytes / sizeof(s->mcast_list[0]);
1332
1333 s->mcast_list = g_realloc(s->mcast_list, list_bytes);
1334 if (!s->mcast_list) {
1335 if (s->mcast_list_len == 0) {
1336 VMW_CFPRN("Current multicast list is empty");
1337 } else {
1338 VMW_ERPRN("Failed to allocate multicast list of %d elements",
1339 s->mcast_list_len);
1340 }
1341 s->mcast_list_len = 0;
1342 } else {
1343 int i;
1344 hwaddr mcast_list_pa =
1345 VMXNET3_READ_DRV_SHARED64(d, s->drv_shmem,
1346 devRead.rxFilterConf.mfTablePA);
1347
1348 pci_dma_read(d, mcast_list_pa, s->mcast_list, list_bytes);
1349
1350 VMW_CFPRN("Current multicast list len is %d:", s->mcast_list_len);
1351 for (i = 0; i < s->mcast_list_len; i++) {
1352 VMW_CFPRN("\t" MAC_FMT, MAC_ARG(s->mcast_list[i].a));
1353 }
1354 }
1355 }
1356
1357 static void vmxnet3_setup_rx_filtering(VMXNET3State *s)
1358 {
1359 vmxnet3_update_rx_mode(s);
1360 vmxnet3_update_vlan_filters(s);
1361 vmxnet3_update_mcast_filters(s);
1362 }
1363
1364 static uint32_t vmxnet3_get_interrupt_config(VMXNET3State *s)
1365 {
1366 uint32_t interrupt_mode = VMXNET3_IT_AUTO | (VMXNET3_IMM_AUTO << 2);
1367 VMW_CFPRN("Interrupt config is 0x%X", interrupt_mode);
1368 return interrupt_mode;
1369 }
1370
1371 static void vmxnet3_fill_stats(VMXNET3State *s)
1372 {
1373 int i;
1374 PCIDevice *d = PCI_DEVICE(s);
1375
1376 if (!s->device_active)
1377 return;
1378
1379 for (i = 0; i < s->txq_num; i++) {
1380 pci_dma_write(d,
1381 s->txq_descr[i].tx_stats_pa,
1382 &s->txq_descr[i].txq_stats,
1383 sizeof(s->txq_descr[i].txq_stats));
1384 }
1385
1386 for (i = 0; i < s->rxq_num; i++) {
1387 pci_dma_write(d,
1388 s->rxq_descr[i].rx_stats_pa,
1389 &s->rxq_descr[i].rxq_stats,
1390 sizeof(s->rxq_descr[i].rxq_stats));
1391 }
1392 }
1393
1394 static void vmxnet3_adjust_by_guest_type(VMXNET3State *s)
1395 {
1396 struct Vmxnet3_GOSInfo gos;
1397 PCIDevice *d = PCI_DEVICE(s);
1398
1399 VMXNET3_READ_DRV_SHARED(d, s->drv_shmem, devRead.misc.driverInfo.gos,
1400 &gos, sizeof(gos));
1401 s->rx_packets_compound =
1402 (gos.gosType == VMXNET3_GOS_TYPE_WIN) ? false : true;
1403
1404 VMW_CFPRN("Guest type specifics: RXCOMPOUND: %d", s->rx_packets_compound);
1405 }
1406
1407 static void
1408 vmxnet3_dump_conf_descr(const char *name,
1409 struct Vmxnet3_VariableLenConfDesc *pm_descr)
1410 {
1411 VMW_CFPRN("%s descriptor dump: Version %u, Length %u",
1412 name, pm_descr->confVer, pm_descr->confLen);
1413
1414 };
1415
1416 static void vmxnet3_update_pm_state(VMXNET3State *s)
1417 {
1418 struct Vmxnet3_VariableLenConfDesc pm_descr;
1419 PCIDevice *d = PCI_DEVICE(s);
1420
1421 pm_descr.confLen =
1422 VMXNET3_READ_DRV_SHARED32(d, s->drv_shmem, devRead.pmConfDesc.confLen);
1423 pm_descr.confVer =
1424 VMXNET3_READ_DRV_SHARED32(d, s->drv_shmem, devRead.pmConfDesc.confVer);
1425 pm_descr.confPA =
1426 VMXNET3_READ_DRV_SHARED64(d, s->drv_shmem, devRead.pmConfDesc.confPA);
1427
1428 vmxnet3_dump_conf_descr("PM State", &pm_descr);
1429 }
1430
1431 static void vmxnet3_update_features(VMXNET3State *s)
1432 {
1433 uint32_t guest_features;
1434 int rxcso_supported;
1435 PCIDevice *d = PCI_DEVICE(s);
1436
1437 guest_features = VMXNET3_READ_DRV_SHARED32(d, s->drv_shmem,
1438 devRead.misc.uptFeatures);
1439
1440 rxcso_supported = VMXNET_FLAG_IS_SET(guest_features, UPT1_F_RXCSUM);
1441 s->rx_vlan_stripping = VMXNET_FLAG_IS_SET(guest_features, UPT1_F_RXVLAN);
1442 s->lro_supported = VMXNET_FLAG_IS_SET(guest_features, UPT1_F_LRO);
1443
1444 VMW_CFPRN("Features configuration: LRO: %d, RXCSUM: %d, VLANSTRIP: %d",
1445 s->lro_supported, rxcso_supported,
1446 s->rx_vlan_stripping);
1447 if (s->peer_has_vhdr) {
1448 qemu_set_offload(qemu_get_queue(s->nic)->peer,
1449 rxcso_supported,
1450 s->lro_supported,
1451 s->lro_supported,
1452 0,
1453 0);
1454 }
1455 }
1456
1457 static bool vmxnet3_verify_intx(VMXNET3State *s, int intx)
1458 {
1459 return s->msix_used || msi_enabled(PCI_DEVICE(s))
1460 || intx == pci_get_byte(s->parent_obj.config + PCI_INTERRUPT_PIN) - 1;
1461 }
1462
1463 static void vmxnet3_validate_interrupt_idx(bool is_msix, int idx)
1464 {
1465 int max_ints = is_msix ? VMXNET3_MAX_INTRS : VMXNET3_MAX_NMSIX_INTRS;
1466 if (idx >= max_ints) {
1467 hw_error("Bad interrupt index: %d\n", idx);
1468 }
1469 }
1470
1471 static void vmxnet3_validate_interrupts(VMXNET3State *s)
1472 {
1473 int i;
1474
1475 VMW_CFPRN("Verifying event interrupt index (%d)", s->event_int_idx);
1476 vmxnet3_validate_interrupt_idx(s->msix_used, s->event_int_idx);
1477
1478 for (i = 0; i < s->txq_num; i++) {
1479 int idx = s->txq_descr[i].intr_idx;
1480 VMW_CFPRN("Verifying TX queue %d interrupt index (%d)", i, idx);
1481 vmxnet3_validate_interrupt_idx(s->msix_used, idx);
1482 }
1483
1484 for (i = 0; i < s->rxq_num; i++) {
1485 int idx = s->rxq_descr[i].intr_idx;
1486 VMW_CFPRN("Verifying RX queue %d interrupt index (%d)", i, idx);
1487 vmxnet3_validate_interrupt_idx(s->msix_used, idx);
1488 }
1489 }
1490
1491 static void vmxnet3_validate_queues(VMXNET3State *s)
1492 {
1493 /*
1494 * txq_num and rxq_num are total number of queues
1495 * configured by guest. These numbers must not
1496 * exceed corresponding maximal values.
1497 */
1498
1499 if (s->txq_num > VMXNET3_DEVICE_MAX_TX_QUEUES) {
1500 hw_error("Bad TX queues number: %d\n", s->txq_num);
1501 }
1502
1503 if (s->rxq_num > VMXNET3_DEVICE_MAX_RX_QUEUES) {
1504 hw_error("Bad RX queues number: %d\n", s->rxq_num);
1505 }
1506 }
1507
1508 static void vmxnet3_activate_device(VMXNET3State *s)
1509 {
1510 int i;
1511 static const uint32_t VMXNET3_DEF_TX_THRESHOLD = 1;
1512 PCIDevice *d = PCI_DEVICE(s);
1513 hwaddr qdescr_table_pa;
1514 uint64_t pa;
1515 uint32_t size;
1516
1517 /* Verify configuration consistency */
1518 if (!vmxnet3_verify_driver_magic(d, s->drv_shmem)) {
1519 VMW_ERPRN("Device configuration received from driver is invalid");
1520 return;
1521 }
1522
1523 /* Verify if device is active */
1524 if (s->device_active) {
1525 VMW_CFPRN("Vmxnet3 device is active");
1526 return;
1527 }
1528
1529 vmxnet3_adjust_by_guest_type(s);
1530 vmxnet3_update_features(s);
1531 vmxnet3_update_pm_state(s);
1532 vmxnet3_setup_rx_filtering(s);
1533 /* Cache fields from shared memory */
1534 s->mtu = VMXNET3_READ_DRV_SHARED32(d, s->drv_shmem, devRead.misc.mtu);
1535 VMW_CFPRN("MTU is %u", s->mtu);
1536
1537 s->max_rx_frags =
1538 VMXNET3_READ_DRV_SHARED16(d, s->drv_shmem, devRead.misc.maxNumRxSG);
1539
1540 if (s->max_rx_frags == 0) {
1541 s->max_rx_frags = 1;
1542 }
1543
1544 VMW_CFPRN("Max RX fragments is %u", s->max_rx_frags);
1545
1546 s->event_int_idx =
1547 VMXNET3_READ_DRV_SHARED8(d, s->drv_shmem, devRead.intrConf.eventIntrIdx);
1548 assert(vmxnet3_verify_intx(s, s->event_int_idx));
1549 VMW_CFPRN("Events interrupt line is %u", s->event_int_idx);
1550
1551 s->auto_int_masking =
1552 VMXNET3_READ_DRV_SHARED8(d, s->drv_shmem, devRead.intrConf.autoMask);
1553 VMW_CFPRN("Automatic interrupt masking is %d", (int)s->auto_int_masking);
1554
1555 s->txq_num =
1556 VMXNET3_READ_DRV_SHARED8(d, s->drv_shmem, devRead.misc.numTxQueues);
1557 s->rxq_num =
1558 VMXNET3_READ_DRV_SHARED8(d, s->drv_shmem, devRead.misc.numRxQueues);
1559
1560 VMW_CFPRN("Number of TX/RX queues %u/%u", s->txq_num, s->rxq_num);
1561 vmxnet3_validate_queues(s);
1562
1563 qdescr_table_pa =
1564 VMXNET3_READ_DRV_SHARED64(d, s->drv_shmem, devRead.misc.queueDescPA);
1565 VMW_CFPRN("TX queues descriptors table is at 0x%" PRIx64, qdescr_table_pa);
1566
1567 /*
1568 * Worst-case scenario is a packet that holds all TX rings space so
1569 * we calculate total size of all TX rings for max TX fragments number
1570 */
1571 s->max_tx_frags = 0;
1572
1573 /* TX queues */
1574 for (i = 0; i < s->txq_num; i++) {
1575 hwaddr qdescr_pa =
1576 qdescr_table_pa + i * sizeof(struct Vmxnet3_TxQueueDesc);
1577
1578 /* Read interrupt number for this TX queue */
1579 s->txq_descr[i].intr_idx =
1580 VMXNET3_READ_TX_QUEUE_DESCR8(d, qdescr_pa, conf.intrIdx);
1581 assert(vmxnet3_verify_intx(s, s->txq_descr[i].intr_idx));
1582
1583 VMW_CFPRN("TX Queue %d interrupt: %d", i, s->txq_descr[i].intr_idx);
1584
1585 /* Read rings memory locations for TX queues */
1586 pa = VMXNET3_READ_TX_QUEUE_DESCR64(d, qdescr_pa, conf.txRingBasePA);
1587 size = VMXNET3_READ_TX_QUEUE_DESCR32(d, qdescr_pa, conf.txRingSize);
1588
1589 vmxnet3_ring_init(d, &s->txq_descr[i].tx_ring, pa, size,
1590 sizeof(struct Vmxnet3_TxDesc), false);
1591 VMXNET3_RING_DUMP(VMW_CFPRN, "TX", i, &s->txq_descr[i].tx_ring);
1592
1593 s->max_tx_frags += size;
1594
1595 /* TXC ring */
1596 pa = VMXNET3_READ_TX_QUEUE_DESCR64(d, qdescr_pa, conf.compRingBasePA);
1597 size = VMXNET3_READ_TX_QUEUE_DESCR32(d, qdescr_pa, conf.compRingSize);
1598 vmxnet3_ring_init(d, &s->txq_descr[i].comp_ring, pa, size,
1599 sizeof(struct Vmxnet3_TxCompDesc), true);
1600 VMXNET3_RING_DUMP(VMW_CFPRN, "TXC", i, &s->txq_descr[i].comp_ring);
1601
1602 s->txq_descr[i].tx_stats_pa =
1603 qdescr_pa + offsetof(struct Vmxnet3_TxQueueDesc, stats);
1604
1605 memset(&s->txq_descr[i].txq_stats, 0,
1606 sizeof(s->txq_descr[i].txq_stats));
1607
1608 /* Fill device-managed parameters for queues */
1609 VMXNET3_WRITE_TX_QUEUE_DESCR32(d, qdescr_pa,
1610 ctrl.txThreshold,
1611 VMXNET3_DEF_TX_THRESHOLD);
1612 }
1613
1614 /* Preallocate TX packet wrapper */
1615 VMW_CFPRN("Max TX fragments is %u", s->max_tx_frags);
1616 net_tx_pkt_init(&s->tx_pkt, PCI_DEVICE(s),
1617 s->max_tx_frags, s->peer_has_vhdr);
1618 net_rx_pkt_init(&s->rx_pkt, s->peer_has_vhdr);
1619
1620 /* Read rings memory locations for RX queues */
1621 for (i = 0; i < s->rxq_num; i++) {
1622 int j;
1623 hwaddr qd_pa =
1624 qdescr_table_pa + s->txq_num * sizeof(struct Vmxnet3_TxQueueDesc) +
1625 i * sizeof(struct Vmxnet3_RxQueueDesc);
1626
1627 /* Read interrupt number for this RX queue */
1628 s->rxq_descr[i].intr_idx =
1629 VMXNET3_READ_TX_QUEUE_DESCR8(d, qd_pa, conf.intrIdx);
1630 assert(vmxnet3_verify_intx(s, s->rxq_descr[i].intr_idx));
1631
1632 VMW_CFPRN("RX Queue %d interrupt: %d", i, s->rxq_descr[i].intr_idx);
1633
1634 /* Read rings memory locations */
1635 for (j = 0; j < VMXNET3_RX_RINGS_PER_QUEUE; j++) {
1636 /* RX rings */
1637 pa = VMXNET3_READ_RX_QUEUE_DESCR64(d, qd_pa, conf.rxRingBasePA[j]);
1638 size = VMXNET3_READ_RX_QUEUE_DESCR32(d, qd_pa, conf.rxRingSize[j]);
1639 vmxnet3_ring_init(d, &s->rxq_descr[i].rx_ring[j], pa, size,
1640 sizeof(struct Vmxnet3_RxDesc), false);
1641 VMW_CFPRN("RX queue %d:%d: Base: %" PRIx64 ", Size: %d",
1642 i, j, pa, size);
1643 }
1644
1645 /* RXC ring */
1646 pa = VMXNET3_READ_RX_QUEUE_DESCR64(d, qd_pa, conf.compRingBasePA);
1647 size = VMXNET3_READ_RX_QUEUE_DESCR32(d, qd_pa, conf.compRingSize);
1648 vmxnet3_ring_init(d, &s->rxq_descr[i].comp_ring, pa, size,
1649 sizeof(struct Vmxnet3_RxCompDesc), true);
1650 VMW_CFPRN("RXC queue %d: Base: %" PRIx64 ", Size: %d", i, pa, size);
1651
1652 s->rxq_descr[i].rx_stats_pa =
1653 qd_pa + offsetof(struct Vmxnet3_RxQueueDesc, stats);
1654 memset(&s->rxq_descr[i].rxq_stats, 0,
1655 sizeof(s->rxq_descr[i].rxq_stats));
1656 }
1657
1658 vmxnet3_validate_interrupts(s);
1659
1660 /* Make sure everything is in place before device activation */
1661 smp_wmb();
1662
1663 vmxnet3_reset_mac(s);
1664
1665 s->device_active = true;
1666 }
1667
1668 static void vmxnet3_handle_command(VMXNET3State *s, uint64_t cmd)
1669 {
1670 s->last_command = cmd;
1671
1672 switch (cmd) {
1673 case VMXNET3_CMD_GET_PERM_MAC_HI:
1674 VMW_CBPRN("Set: Get upper part of permanent MAC");
1675 break;
1676
1677 case VMXNET3_CMD_GET_PERM_MAC_LO:
1678 VMW_CBPRN("Set: Get lower part of permanent MAC");
1679 break;
1680
1681 case VMXNET3_CMD_GET_STATS:
1682 VMW_CBPRN("Set: Get device statistics");
1683 vmxnet3_fill_stats(s);
1684 break;
1685
1686 case VMXNET3_CMD_ACTIVATE_DEV:
1687 VMW_CBPRN("Set: Activating vmxnet3 device");
1688 vmxnet3_activate_device(s);
1689 break;
1690
1691 case VMXNET3_CMD_UPDATE_RX_MODE:
1692 VMW_CBPRN("Set: Update rx mode");
1693 vmxnet3_update_rx_mode(s);
1694 break;
1695
1696 case VMXNET3_CMD_UPDATE_VLAN_FILTERS:
1697 VMW_CBPRN("Set: Update VLAN filters");
1698 vmxnet3_update_vlan_filters(s);
1699 break;
1700
1701 case VMXNET3_CMD_UPDATE_MAC_FILTERS:
1702 VMW_CBPRN("Set: Update MAC filters");
1703 vmxnet3_update_mcast_filters(s);
1704 break;
1705
1706 case VMXNET3_CMD_UPDATE_FEATURE:
1707 VMW_CBPRN("Set: Update features");
1708 vmxnet3_update_features(s);
1709 break;
1710
1711 case VMXNET3_CMD_UPDATE_PMCFG:
1712 VMW_CBPRN("Set: Update power management config");
1713 vmxnet3_update_pm_state(s);
1714 break;
1715
1716 case VMXNET3_CMD_GET_LINK:
1717 VMW_CBPRN("Set: Get link");
1718 break;
1719
1720 case VMXNET3_CMD_RESET_DEV:
1721 VMW_CBPRN("Set: Reset device");
1722 vmxnet3_reset(s);
1723 break;
1724
1725 case VMXNET3_CMD_QUIESCE_DEV:
1726 VMW_CBPRN("Set: VMXNET3_CMD_QUIESCE_DEV - deactivate the device");
1727 vmxnet3_deactivate_device(s);
1728 break;
1729
1730 case VMXNET3_CMD_GET_CONF_INTR:
1731 VMW_CBPRN("Set: VMXNET3_CMD_GET_CONF_INTR - interrupt configuration");
1732 break;
1733
1734 case VMXNET3_CMD_GET_ADAPTIVE_RING_INFO:
1735 VMW_CBPRN("Set: VMXNET3_CMD_GET_ADAPTIVE_RING_INFO - "
1736 "adaptive ring info flags");
1737 break;
1738
1739 case VMXNET3_CMD_GET_DID_LO:
1740 VMW_CBPRN("Set: Get lower part of device ID");
1741 break;
1742
1743 case VMXNET3_CMD_GET_DID_HI:
1744 VMW_CBPRN("Set: Get upper part of device ID");
1745 break;
1746
1747 case VMXNET3_CMD_GET_DEV_EXTRA_INFO:
1748 VMW_CBPRN("Set: Get device extra info");
1749 break;
1750
1751 default:
1752 VMW_CBPRN("Received unknown command: %" PRIx64, cmd);
1753 break;
1754 }
1755 }
1756
1757 static uint64_t vmxnet3_get_command_status(VMXNET3State *s)
1758 {
1759 uint64_t ret;
1760
1761 switch (s->last_command) {
1762 case VMXNET3_CMD_ACTIVATE_DEV:
1763 ret = (s->device_active) ? 0 : 1;
1764 VMW_CFPRN("Device active: %" PRIx64, ret);
1765 break;
1766
1767 case VMXNET3_CMD_RESET_DEV:
1768 case VMXNET3_CMD_QUIESCE_DEV:
1769 case VMXNET3_CMD_GET_QUEUE_STATUS:
1770 case VMXNET3_CMD_GET_DEV_EXTRA_INFO:
1771 ret = 0;
1772 break;
1773
1774 case VMXNET3_CMD_GET_LINK:
1775 ret = s->link_status_and_speed;
1776 VMW_CFPRN("Link and speed: %" PRIx64, ret);
1777 break;
1778
1779 case VMXNET3_CMD_GET_PERM_MAC_LO:
1780 ret = vmxnet3_get_mac_low(&s->perm_mac);
1781 break;
1782
1783 case VMXNET3_CMD_GET_PERM_MAC_HI:
1784 ret = vmxnet3_get_mac_high(&s->perm_mac);
1785 break;
1786
1787 case VMXNET3_CMD_GET_CONF_INTR:
1788 ret = vmxnet3_get_interrupt_config(s);
1789 break;
1790
1791 case VMXNET3_CMD_GET_ADAPTIVE_RING_INFO:
1792 ret = VMXNET3_DISABLE_ADAPTIVE_RING;
1793 break;
1794
1795 case VMXNET3_CMD_GET_DID_LO:
1796 ret = PCI_DEVICE_ID_VMWARE_VMXNET3;
1797 break;
1798
1799 case VMXNET3_CMD_GET_DID_HI:
1800 ret = VMXNET3_DEVICE_REVISION;
1801 break;
1802
1803 default:
1804 VMW_WRPRN("Received request for unknown command: %x", s->last_command);
1805 ret = 0;
1806 break;
1807 }
1808
1809 return ret;
1810 }
1811
1812 static void vmxnet3_set_events(VMXNET3State *s, uint32_t val)
1813 {
1814 uint32_t events;
1815 PCIDevice *d = PCI_DEVICE(s);
1816
1817 VMW_CBPRN("Setting events: 0x%x", val);
1818 events = VMXNET3_READ_DRV_SHARED32(d, s->drv_shmem, ecr) | val;
1819 VMXNET3_WRITE_DRV_SHARED32(d, s->drv_shmem, ecr, events);
1820 }
1821
1822 static void vmxnet3_ack_events(VMXNET3State *s, uint32_t val)
1823 {
1824 PCIDevice *d = PCI_DEVICE(s);
1825 uint32_t events;
1826
1827 VMW_CBPRN("Clearing events: 0x%x", val);
1828 events = VMXNET3_READ_DRV_SHARED32(d, s->drv_shmem, ecr) & ~val;
1829 VMXNET3_WRITE_DRV_SHARED32(d, s->drv_shmem, ecr, events);
1830 }
1831
1832 static void
1833 vmxnet3_io_bar1_write(void *opaque,
1834 hwaddr addr,
1835 uint64_t val,
1836 unsigned size)
1837 {
1838 VMXNET3State *s = opaque;
1839
1840 switch (addr) {
1841 /* Vmxnet3 Revision Report Selection */
1842 case VMXNET3_REG_VRRS:
1843 VMW_CBPRN("Write BAR1 [VMXNET3_REG_VRRS] = %" PRIx64 ", size %d",
1844 val, size);
1845 break;
1846
1847 /* UPT Version Report Selection */
1848 case VMXNET3_REG_UVRS:
1849 VMW_CBPRN("Write BAR1 [VMXNET3_REG_UVRS] = %" PRIx64 ", size %d",
1850 val, size);
1851 break;
1852
1853 /* Driver Shared Address Low */
1854 case VMXNET3_REG_DSAL:
1855 VMW_CBPRN("Write BAR1 [VMXNET3_REG_DSAL] = %" PRIx64 ", size %d",
1856 val, size);
1857 /*
1858 * Guest driver will first write the low part of the shared
1859 * memory address. We save it to temp variable and set the
1860 * shared address only after we get the high part
1861 */
1862 if (val == 0) {
1863 vmxnet3_deactivate_device(s);
1864 }
1865 s->temp_shared_guest_driver_memory = val;
1866 s->drv_shmem = 0;
1867 break;
1868
1869 /* Driver Shared Address High */
1870 case VMXNET3_REG_DSAH:
1871 VMW_CBPRN("Write BAR1 [VMXNET3_REG_DSAH] = %" PRIx64 ", size %d",
1872 val, size);
1873 /*
1874 * Set the shared memory between guest driver and device.
1875 * We already should have low address part.
1876 */
1877 s->drv_shmem = s->temp_shared_guest_driver_memory | (val << 32);
1878 break;
1879
1880 /* Command */
1881 case VMXNET3_REG_CMD:
1882 VMW_CBPRN("Write BAR1 [VMXNET3_REG_CMD] = %" PRIx64 ", size %d",
1883 val, size);
1884 vmxnet3_handle_command(s, val);
1885 break;
1886
1887 /* MAC Address Low */
1888 case VMXNET3_REG_MACL:
1889 VMW_CBPRN("Write BAR1 [VMXNET3_REG_MACL] = %" PRIx64 ", size %d",
1890 val, size);
1891 s->temp_mac = val;
1892 break;
1893
1894 /* MAC Address High */
1895 case VMXNET3_REG_MACH:
1896 VMW_CBPRN("Write BAR1 [VMXNET3_REG_MACH] = %" PRIx64 ", size %d",
1897 val, size);
1898 vmxnet3_set_variable_mac(s, val, s->temp_mac);
1899 break;
1900
1901 /* Interrupt Cause Register */
1902 case VMXNET3_REG_ICR:
1903 VMW_CBPRN("Write BAR1 [VMXNET3_REG_ICR] = %" PRIx64 ", size %d",
1904 val, size);
1905 g_assert_not_reached();
1906 break;
1907
1908 /* Event Cause Register */
1909 case VMXNET3_REG_ECR:
1910 VMW_CBPRN("Write BAR1 [VMXNET3_REG_ECR] = %" PRIx64 ", size %d",
1911 val, size);
1912 vmxnet3_ack_events(s, val);
1913 break;
1914
1915 default:
1916 VMW_CBPRN("Unknown Write to BAR1 [%" PRIx64 "] = %" PRIx64 ", size %d",
1917 addr, val, size);
1918 break;
1919 }
1920 }
1921
1922 static uint64_t
1923 vmxnet3_io_bar1_read(void *opaque, hwaddr addr, unsigned size)
1924 {
1925 VMXNET3State *s = opaque;
1926 uint64_t ret = 0;
1927
1928 switch (addr) {
1929 /* Vmxnet3 Revision Report Selection */
1930 case VMXNET3_REG_VRRS:
1931 VMW_CBPRN("Read BAR1 [VMXNET3_REG_VRRS], size %d", size);
1932 ret = VMXNET3_DEVICE_REVISION;
1933 break;
1934
1935 /* UPT Version Report Selection */
1936 case VMXNET3_REG_UVRS:
1937 VMW_CBPRN("Read BAR1 [VMXNET3_REG_UVRS], size %d", size);
1938 ret = VMXNET3_UPT_REVISION;
1939 break;
1940
1941 /* Command */
1942 case VMXNET3_REG_CMD:
1943 VMW_CBPRN("Read BAR1 [VMXNET3_REG_CMD], size %d", size);
1944 ret = vmxnet3_get_command_status(s);
1945 break;
1946
1947 /* MAC Address Low */
1948 case VMXNET3_REG_MACL:
1949 VMW_CBPRN("Read BAR1 [VMXNET3_REG_MACL], size %d", size);
1950 ret = vmxnet3_get_mac_low(&s->conf.macaddr);
1951 break;
1952
1953 /* MAC Address High */
1954 case VMXNET3_REG_MACH:
1955 VMW_CBPRN("Read BAR1 [VMXNET3_REG_MACH], size %d", size);
1956 ret = vmxnet3_get_mac_high(&s->conf.macaddr);
1957 break;
1958
1959 /*
1960 * Interrupt Cause Register
1961 * Used for legacy interrupts only so interrupt index always 0
1962 */
1963 case VMXNET3_REG_ICR:
1964 VMW_CBPRN("Read BAR1 [VMXNET3_REG_ICR], size %d", size);
1965 if (vmxnet3_interrupt_asserted(s, 0)) {
1966 vmxnet3_clear_interrupt(s, 0);
1967 ret = true;
1968 } else {
1969 ret = false;
1970 }
1971 break;
1972
1973 default:
1974 VMW_CBPRN("Unknow read BAR1[%" PRIx64 "], %d bytes", addr, size);
1975 break;
1976 }
1977
1978 return ret;
1979 }
1980
1981 static int
1982 vmxnet3_can_receive(NetClientState *nc)
1983 {
1984 VMXNET3State *s = qemu_get_nic_opaque(nc);
1985 return s->device_active &&
1986 VMXNET_FLAG_IS_SET(s->link_status_and_speed, VMXNET3_LINK_STATUS_UP);
1987 }
1988
1989 static inline bool
1990 vmxnet3_is_registered_vlan(VMXNET3State *s, const void *data)
1991 {
1992 uint16_t vlan_tag = eth_get_pkt_tci(data) & VLAN_VID_MASK;
1993 if (IS_SPECIAL_VLAN_ID(vlan_tag)) {
1994 return true;
1995 }
1996
1997 return VMXNET3_VFTABLE_ENTRY_IS_SET(s->vlan_table, vlan_tag);
1998 }
1999
2000 static bool
2001 vmxnet3_is_allowed_mcast_group(VMXNET3State *s, const uint8_t *group_mac)
2002 {
2003 int i;
2004 for (i = 0; i < s->mcast_list_len; i++) {
2005 if (!memcmp(group_mac, s->mcast_list[i].a, sizeof(s->mcast_list[i]))) {
2006 return true;
2007 }
2008 }
2009 return false;
2010 }
2011
2012 static bool
2013 vmxnet3_rx_filter_may_indicate(VMXNET3State *s, const void *data,
2014 size_t size)
2015 {
2016 struct eth_header *ehdr = PKT_GET_ETH_HDR(data);
2017
2018 if (VMXNET_FLAG_IS_SET(s->rx_mode, VMXNET3_RXM_PROMISC)) {
2019 return true;
2020 }
2021
2022 if (!vmxnet3_is_registered_vlan(s, data)) {
2023 return false;
2024 }
2025
2026 switch (net_rx_pkt_get_packet_type(s->rx_pkt)) {
2027 case ETH_PKT_UCAST:
2028 if (!VMXNET_FLAG_IS_SET(s->rx_mode, VMXNET3_RXM_UCAST)) {
2029 return false;
2030 }
2031 if (memcmp(s->conf.macaddr.a, ehdr->h_dest, ETH_ALEN)) {
2032 return false;
2033 }
2034 break;
2035
2036 case ETH_PKT_BCAST:
2037 if (!VMXNET_FLAG_IS_SET(s->rx_mode, VMXNET3_RXM_BCAST)) {
2038 return false;
2039 }
2040 break;
2041
2042 case ETH_PKT_MCAST:
2043 if (VMXNET_FLAG_IS_SET(s->rx_mode, VMXNET3_RXM_ALL_MULTI)) {
2044 return true;
2045 }
2046 if (!VMXNET_FLAG_IS_SET(s->rx_mode, VMXNET3_RXM_MCAST)) {
2047 return false;
2048 }
2049 if (!vmxnet3_is_allowed_mcast_group(s, ehdr->h_dest)) {
2050 return false;
2051 }
2052 break;
2053
2054 default:
2055 g_assert_not_reached();
2056 }
2057
2058 return true;
2059 }
2060
2061 static ssize_t
2062 vmxnet3_receive(NetClientState *nc, const uint8_t *buf, size_t size)
2063 {
2064 VMXNET3State *s = qemu_get_nic_opaque(nc);
2065 size_t bytes_indicated;
2066 uint8_t min_buf[MIN_BUF_SIZE];
2067
2068 if (!vmxnet3_can_receive(nc)) {
2069 VMW_PKPRN("Cannot receive now");
2070 return -1;
2071 }
2072
2073 if (s->peer_has_vhdr) {
2074 net_rx_pkt_set_vhdr(s->rx_pkt, (struct virtio_net_hdr *)buf);
2075 buf += sizeof(struct virtio_net_hdr);
2076 size -= sizeof(struct virtio_net_hdr);
2077 }
2078
2079 /* Pad to minimum Ethernet frame length */
2080 if (size < sizeof(min_buf)) {
2081 memcpy(min_buf, buf, size);
2082 memset(&min_buf[size], 0, sizeof(min_buf) - size);
2083 buf = min_buf;
2084 size = sizeof(min_buf);
2085 }
2086
2087 net_rx_pkt_set_packet_type(s->rx_pkt,
2088 get_eth_packet_type(PKT_GET_ETH_HDR(buf)));
2089
2090 if (vmxnet3_rx_filter_may_indicate(s, buf, size)) {
2091 net_rx_pkt_set_protocols(s->rx_pkt, buf, size);
2092 vmxnet3_rx_need_csum_calculate(s->rx_pkt, buf, size);
2093 net_rx_pkt_attach_data(s->rx_pkt, buf, size, s->rx_vlan_stripping);
2094 bytes_indicated = vmxnet3_indicate_packet(s) ? size : -1;
2095 if (bytes_indicated < size) {
2096 VMW_PKPRN("RX: %zu of %zu bytes indicated", bytes_indicated, size);
2097 }
2098 } else {
2099 VMW_PKPRN("Packet dropped by RX filter");
2100 bytes_indicated = size;
2101 }
2102
2103 assert(size > 0);
2104 assert(bytes_indicated != 0);
2105 return bytes_indicated;
2106 }
2107
2108 static void vmxnet3_set_link_status(NetClientState *nc)
2109 {
2110 VMXNET3State *s = qemu_get_nic_opaque(nc);
2111
2112 if (nc->link_down) {
2113 s->link_status_and_speed &= ~VMXNET3_LINK_STATUS_UP;
2114 } else {
2115 s->link_status_and_speed |= VMXNET3_LINK_STATUS_UP;
2116 }
2117
2118 vmxnet3_set_events(s, VMXNET3_ECR_LINK);
2119 vmxnet3_trigger_interrupt(s, s->event_int_idx);
2120 }
2121
2122 static NetClientInfo net_vmxnet3_info = {
2123 .type = NET_CLIENT_DRIVER_NIC,
2124 .size = sizeof(NICState),
2125 .receive = vmxnet3_receive,
2126 .link_status_changed = vmxnet3_set_link_status,
2127 };
2128
2129 static bool vmxnet3_peer_has_vnet_hdr(VMXNET3State *s)
2130 {
2131 NetClientState *nc = qemu_get_queue(s->nic);
2132
2133 if (qemu_has_vnet_hdr(nc->peer)) {
2134 return true;
2135 }
2136
2137 return false;
2138 }
2139
2140 static void vmxnet3_net_uninit(VMXNET3State *s)
2141 {
2142 g_free(s->mcast_list);
2143 vmxnet3_deactivate_device(s);
2144 qemu_del_nic(s->nic);
2145 }
2146
2147 static void vmxnet3_net_init(VMXNET3State *s)
2148 {
2149 DeviceState *d = DEVICE(s);
2150
2151 VMW_CBPRN("vmxnet3_net_init called...");
2152
2153 qemu_macaddr_default_if_unset(&s->conf.macaddr);
2154
2155 /* Windows guest will query the address that was set on init */
2156 memcpy(&s->perm_mac.a, &s->conf.macaddr.a, sizeof(s->perm_mac.a));
2157
2158 s->mcast_list = NULL;
2159 s->mcast_list_len = 0;
2160
2161 s->link_status_and_speed = VMXNET3_LINK_SPEED | VMXNET3_LINK_STATUS_UP;
2162
2163 VMW_CFPRN("Permanent MAC: " MAC_FMT, MAC_ARG(s->perm_mac.a));
2164
2165 s->nic = qemu_new_nic(&net_vmxnet3_info, &s->conf,
2166 object_get_typename(OBJECT(s)),
2167 d->id, s);
2168
2169 s->peer_has_vhdr = vmxnet3_peer_has_vnet_hdr(s);
2170 s->tx_sop = true;
2171 s->skip_current_tx_pkt = false;
2172 s->tx_pkt = NULL;
2173 s->rx_pkt = NULL;
2174 s->rx_vlan_stripping = false;
2175 s->lro_supported = false;
2176
2177 if (s->peer_has_vhdr) {
2178 qemu_set_vnet_hdr_len(qemu_get_queue(s->nic)->peer,
2179 sizeof(struct virtio_net_hdr));
2180
2181 qemu_using_vnet_hdr(qemu_get_queue(s->nic)->peer, 1);
2182 }
2183
2184 qemu_format_nic_info_str(qemu_get_queue(s->nic), s->conf.macaddr.a);
2185 }
2186
2187 static void
2188 vmxnet3_unuse_msix_vectors(VMXNET3State *s, int num_vectors)
2189 {
2190 PCIDevice *d = PCI_DEVICE(s);
2191 int i;
2192 for (i = 0; i < num_vectors; i++) {
2193 msix_vector_unuse(d, i);
2194 }
2195 }
2196
2197 static bool
2198 vmxnet3_use_msix_vectors(VMXNET3State *s, int num_vectors)
2199 {
2200 PCIDevice *d = PCI_DEVICE(s);
2201 int i;
2202 for (i = 0; i < num_vectors; i++) {
2203 int res = msix_vector_use(d, i);
2204 if (0 > res) {
2205 VMW_WRPRN("Failed to use MSI-X vector %d, error %d", i, res);
2206 vmxnet3_unuse_msix_vectors(s, i);
2207 return false;
2208 }
2209 }
2210 return true;
2211 }
2212
2213 static bool
2214 vmxnet3_init_msix(VMXNET3State *s)
2215 {
2216 PCIDevice *d = PCI_DEVICE(s);
2217 int res = msix_init(d, VMXNET3_MAX_INTRS,
2218 &s->msix_bar,
2219 VMXNET3_MSIX_BAR_IDX, VMXNET3_OFF_MSIX_TABLE,
2220 &s->msix_bar,
2221 VMXNET3_MSIX_BAR_IDX, VMXNET3_OFF_MSIX_PBA(s),
2222 VMXNET3_MSIX_OFFSET(s), NULL);
2223
2224 if (0 > res) {
2225 VMW_WRPRN("Failed to initialize MSI-X, error %d", res);
2226 s->msix_used = false;
2227 } else {
2228 if (!vmxnet3_use_msix_vectors(s, VMXNET3_MAX_INTRS)) {
2229 VMW_WRPRN("Failed to use MSI-X vectors, error %d", res);
2230 msix_uninit(d, &s->msix_bar, &s->msix_bar);
2231 s->msix_used = false;
2232 } else {
2233 s->msix_used = true;
2234 }
2235 }
2236 return s->msix_used;
2237 }
2238
2239 static void
2240 vmxnet3_cleanup_msix(VMXNET3State *s)
2241 {
2242 PCIDevice *d = PCI_DEVICE(s);
2243
2244 if (s->msix_used) {
2245 vmxnet3_unuse_msix_vectors(s, VMXNET3_MAX_INTRS);
2246 msix_uninit(d, &s->msix_bar, &s->msix_bar);
2247 }
2248 }
2249
2250 static void
2251 vmxnet3_cleanup_msi(VMXNET3State *s)
2252 {
2253 PCIDevice *d = PCI_DEVICE(s);
2254
2255 msi_uninit(d);
2256 }
2257
2258 static void
2259 vmxnet3_msix_save(QEMUFile *f, void *opaque)
2260 {
2261 PCIDevice *d = PCI_DEVICE(opaque);
2262 msix_save(d, f);
2263 }
2264
2265 static int
2266 vmxnet3_msix_load(QEMUFile *f, void *opaque, int version_id)
2267 {
2268 PCIDevice *d = PCI_DEVICE(opaque);
2269 msix_load(d, f);
2270 return 0;
2271 }
2272
2273 static const MemoryRegionOps b0_ops = {
2274 .read = vmxnet3_io_bar0_read,
2275 .write = vmxnet3_io_bar0_write,
2276 .endianness = DEVICE_LITTLE_ENDIAN,
2277 .impl = {
2278 .min_access_size = 4,
2279 .max_access_size = 4,
2280 },
2281 };
2282
2283 static const MemoryRegionOps b1_ops = {
2284 .read = vmxnet3_io_bar1_read,
2285 .write = vmxnet3_io_bar1_write,
2286 .endianness = DEVICE_LITTLE_ENDIAN,
2287 .impl = {
2288 .min_access_size = 4,
2289 .max_access_size = 4,
2290 },
2291 };
2292
2293 static SaveVMHandlers savevm_vmxnet3_msix = {
2294 .save_state = vmxnet3_msix_save,
2295 .load_state = vmxnet3_msix_load,
2296 };
2297
2298 static uint64_t vmxnet3_device_serial_num(VMXNET3State *s)
2299 {
2300 uint64_t dsn_payload;
2301 uint8_t *dsnp = (uint8_t *)&dsn_payload;
2302
2303 dsnp[0] = 0xfe;
2304 dsnp[1] = s->conf.macaddr.a[3];
2305 dsnp[2] = s->conf.macaddr.a[4];
2306 dsnp[3] = s->conf.macaddr.a[5];
2307 dsnp[4] = s->conf.macaddr.a[0];
2308 dsnp[5] = s->conf.macaddr.a[1];
2309 dsnp[6] = s->conf.macaddr.a[2];
2310 dsnp[7] = 0xff;
2311 return dsn_payload;
2312 }
2313
2314
2315 #define VMXNET3_USE_64BIT (true)
2316 #define VMXNET3_PER_VECTOR_MASK (false)
2317
2318 static void vmxnet3_pci_realize(PCIDevice *pci_dev, Error **errp)
2319 {
2320 DeviceState *dev = DEVICE(pci_dev);
2321 VMXNET3State *s = VMXNET3(pci_dev);
2322 int ret;
2323
2324 VMW_CBPRN("Starting init...");
2325
2326 memory_region_init_io(&s->bar0, OBJECT(s), &b0_ops, s,
2327 "vmxnet3-b0", VMXNET3_PT_REG_SIZE);
2328 pci_register_bar(pci_dev, VMXNET3_BAR0_IDX,
2329 PCI_BASE_ADDRESS_SPACE_MEMORY, &s->bar0);
2330
2331 memory_region_init_io(&s->bar1, OBJECT(s), &b1_ops, s,
2332 "vmxnet3-b1", VMXNET3_VD_REG_SIZE);
2333 pci_register_bar(pci_dev, VMXNET3_BAR1_IDX,
2334 PCI_BASE_ADDRESS_SPACE_MEMORY, &s->bar1);
2335
2336 memory_region_init(&s->msix_bar, OBJECT(s), "vmxnet3-msix-bar",
2337 VMXNET3_MSIX_BAR_SIZE);
2338 pci_register_bar(pci_dev, VMXNET3_MSIX_BAR_IDX,
2339 PCI_BASE_ADDRESS_SPACE_MEMORY, &s->msix_bar);
2340
2341 vmxnet3_reset_interrupt_states(s);
2342
2343 /* Interrupt pin A */
2344 pci_dev->config[PCI_INTERRUPT_PIN] = 0x01;
2345
2346 ret = msi_init(pci_dev, VMXNET3_MSI_OFFSET(s), VMXNET3_MAX_NMSIX_INTRS,
2347 VMXNET3_USE_64BIT, VMXNET3_PER_VECTOR_MASK, NULL);
2348 /* Any error other than -ENOTSUP(board's MSI support is broken)
2349 * is a programming error. Fall back to INTx silently on -ENOTSUP */
2350 assert(!ret || ret == -ENOTSUP);
2351
2352 if (!vmxnet3_init_msix(s)) {
2353 VMW_WRPRN("Failed to initialize MSI-X, configuration is inconsistent.");
2354 }
2355
2356 vmxnet3_net_init(s);
2357
2358 if (pci_is_express(pci_dev)) {
2359 if (pci_bus_is_express(pci_get_bus(pci_dev))) {
2360 pcie_endpoint_cap_init(pci_dev, VMXNET3_EXP_EP_OFFSET);
2361 }
2362
2363 pcie_dev_ser_num_init(pci_dev, VMXNET3_DSN_OFFSET,
2364 vmxnet3_device_serial_num(s));
2365 }
2366
2367 register_savevm_live(dev, "vmxnet3-msix", -1, 1, &savevm_vmxnet3_msix, s);
2368 }
2369
2370 static void vmxnet3_instance_init(Object *obj)
2371 {
2372 VMXNET3State *s = VMXNET3(obj);
2373 device_add_bootindex_property(obj, &s->conf.bootindex,
2374 "bootindex", "/ethernet-phy@0",
2375 DEVICE(obj), NULL);
2376 }
2377
2378 static void vmxnet3_pci_uninit(PCIDevice *pci_dev)
2379 {
2380 DeviceState *dev = DEVICE(pci_dev);
2381 VMXNET3State *s = VMXNET3(pci_dev);
2382
2383 VMW_CBPRN("Starting uninit...");
2384
2385 unregister_savevm(dev, "vmxnet3-msix", s);
2386
2387 vmxnet3_net_uninit(s);
2388
2389 vmxnet3_cleanup_msix(s);
2390
2391 vmxnet3_cleanup_msi(s);
2392 }
2393
2394 static void vmxnet3_qdev_reset(DeviceState *dev)
2395 {
2396 PCIDevice *d = PCI_DEVICE(dev);
2397 VMXNET3State *s = VMXNET3(d);
2398
2399 VMW_CBPRN("Starting QDEV reset...");
2400 vmxnet3_reset(s);
2401 }
2402
2403 static bool vmxnet3_mc_list_needed(void *opaque)
2404 {
2405 return true;
2406 }
2407
2408 static int vmxnet3_mcast_list_pre_load(void *opaque)
2409 {
2410 VMXNET3State *s = opaque;
2411
2412 s->mcast_list = g_malloc(s->mcast_list_buff_size);
2413
2414 return 0;
2415 }
2416
2417
2418 static int vmxnet3_pre_save(void *opaque)
2419 {
2420 VMXNET3State *s = opaque;
2421
2422 s->mcast_list_buff_size = s->mcast_list_len * sizeof(MACAddr);
2423
2424 return 0;
2425 }
2426
2427 static const VMStateDescription vmxstate_vmxnet3_mcast_list = {
2428 .name = "vmxnet3/mcast_list",
2429 .version_id = 1,
2430 .minimum_version_id = 1,
2431 .pre_load = vmxnet3_mcast_list_pre_load,
2432 .needed = vmxnet3_mc_list_needed,
2433 .fields = (VMStateField[]) {
2434 VMSTATE_VBUFFER_UINT32(mcast_list, VMXNET3State, 0, NULL,
2435 mcast_list_buff_size),
2436 VMSTATE_END_OF_LIST()
2437 }
2438 };
2439
2440 static const VMStateDescription vmstate_vmxnet3_ring = {
2441 .name = "vmxnet3-ring",
2442 .version_id = 0,
2443 .fields = (VMStateField[]) {
2444 VMSTATE_UINT64(pa, Vmxnet3Ring),
2445 VMSTATE_UINT32(size, Vmxnet3Ring),
2446 VMSTATE_UINT32(cell_size, Vmxnet3Ring),
2447 VMSTATE_UINT32(next, Vmxnet3Ring),
2448 VMSTATE_UINT8(gen, Vmxnet3Ring),
2449 VMSTATE_END_OF_LIST()
2450 }
2451 };
2452
2453 static const VMStateDescription vmstate_vmxnet3_tx_stats = {
2454 .name = "vmxnet3-tx-stats",
2455 .version_id = 0,
2456 .fields = (VMStateField[]) {
2457 VMSTATE_UINT64(TSOPktsTxOK, struct UPT1_TxStats),
2458 VMSTATE_UINT64(TSOBytesTxOK, struct UPT1_TxStats),
2459 VMSTATE_UINT64(ucastPktsTxOK, struct UPT1_TxStats),
2460 VMSTATE_UINT64(ucastBytesTxOK, struct UPT1_TxStats),
2461 VMSTATE_UINT64(mcastPktsTxOK, struct UPT1_TxStats),
2462 VMSTATE_UINT64(mcastBytesTxOK, struct UPT1_TxStats),
2463 VMSTATE_UINT64(bcastPktsTxOK, struct UPT1_TxStats),
2464 VMSTATE_UINT64(bcastBytesTxOK, struct UPT1_TxStats),
2465 VMSTATE_UINT64(pktsTxError, struct UPT1_TxStats),
2466 VMSTATE_UINT64(pktsTxDiscard, struct UPT1_TxStats),
2467 VMSTATE_END_OF_LIST()
2468 }
2469 };
2470
2471 static const VMStateDescription vmstate_vmxnet3_txq_descr = {
2472 .name = "vmxnet3-txq-descr",
2473 .version_id = 0,
2474 .fields = (VMStateField[]) {
2475 VMSTATE_STRUCT(tx_ring, Vmxnet3TxqDescr, 0, vmstate_vmxnet3_ring,
2476 Vmxnet3Ring),
2477 VMSTATE_STRUCT(comp_ring, Vmxnet3TxqDescr, 0, vmstate_vmxnet3_ring,
2478 Vmxnet3Ring),
2479 VMSTATE_UINT8(intr_idx, Vmxnet3TxqDescr),
2480 VMSTATE_UINT64(tx_stats_pa, Vmxnet3TxqDescr),
2481 VMSTATE_STRUCT(txq_stats, Vmxnet3TxqDescr, 0, vmstate_vmxnet3_tx_stats,
2482 struct UPT1_TxStats),
2483 VMSTATE_END_OF_LIST()
2484 }
2485 };
2486
2487 static const VMStateDescription vmstate_vmxnet3_rx_stats = {
2488 .name = "vmxnet3-rx-stats",
2489 .version_id = 0,
2490 .fields = (VMStateField[]) {
2491 VMSTATE_UINT64(LROPktsRxOK, struct UPT1_RxStats),
2492 VMSTATE_UINT64(LROBytesRxOK, struct UPT1_RxStats),
2493 VMSTATE_UINT64(ucastPktsRxOK, struct UPT1_RxStats),
2494 VMSTATE_UINT64(ucastBytesRxOK, struct UPT1_RxStats),
2495 VMSTATE_UINT64(mcastPktsRxOK, struct UPT1_RxStats),
2496 VMSTATE_UINT64(mcastBytesRxOK, struct UPT1_RxStats),
2497 VMSTATE_UINT64(bcastPktsRxOK, struct UPT1_RxStats),
2498 VMSTATE_UINT64(bcastBytesRxOK, struct UPT1_RxStats),
2499 VMSTATE_UINT64(pktsRxOutOfBuf, struct UPT1_RxStats),
2500 VMSTATE_UINT64(pktsRxError, struct UPT1_RxStats),
2501 VMSTATE_END_OF_LIST()
2502 }
2503 };
2504
2505 static const VMStateDescription vmstate_vmxnet3_rxq_descr = {
2506 .name = "vmxnet3-rxq-descr",
2507 .version_id = 0,
2508 .fields = (VMStateField[]) {
2509 VMSTATE_STRUCT_ARRAY(rx_ring, Vmxnet3RxqDescr,
2510 VMXNET3_RX_RINGS_PER_QUEUE, 0,
2511 vmstate_vmxnet3_ring, Vmxnet3Ring),
2512 VMSTATE_STRUCT(comp_ring, Vmxnet3RxqDescr, 0, vmstate_vmxnet3_ring,
2513 Vmxnet3Ring),
2514 VMSTATE_UINT8(intr_idx, Vmxnet3RxqDescr),
2515 VMSTATE_UINT64(rx_stats_pa, Vmxnet3RxqDescr),
2516 VMSTATE_STRUCT(rxq_stats, Vmxnet3RxqDescr, 0, vmstate_vmxnet3_rx_stats,
2517 struct UPT1_RxStats),
2518 VMSTATE_END_OF_LIST()
2519 }
2520 };
2521
2522 static int vmxnet3_post_load(void *opaque, int version_id)
2523 {
2524 VMXNET3State *s = opaque;
2525 PCIDevice *d = PCI_DEVICE(s);
2526
2527 net_tx_pkt_init(&s->tx_pkt, PCI_DEVICE(s),
2528 s->max_tx_frags, s->peer_has_vhdr);
2529 net_rx_pkt_init(&s->rx_pkt, s->peer_has_vhdr);
2530
2531 if (s->msix_used) {
2532 if (!vmxnet3_use_msix_vectors(s, VMXNET3_MAX_INTRS)) {
2533 VMW_WRPRN("Failed to re-use MSI-X vectors");
2534 msix_uninit(d, &s->msix_bar, &s->msix_bar);
2535 s->msix_used = false;
2536 return -1;
2537 }
2538 }
2539
2540 vmxnet3_validate_queues(s);
2541 vmxnet3_validate_interrupts(s);
2542
2543 return 0;
2544 }
2545
2546 static const VMStateDescription vmstate_vmxnet3_int_state = {
2547 .name = "vmxnet3-int-state",
2548 .version_id = 0,
2549 .fields = (VMStateField[]) {
2550 VMSTATE_BOOL(is_masked, Vmxnet3IntState),
2551 VMSTATE_BOOL(is_pending, Vmxnet3IntState),
2552 VMSTATE_BOOL(is_asserted, Vmxnet3IntState),
2553 VMSTATE_END_OF_LIST()
2554 }
2555 };
2556
2557 static bool vmxnet3_vmstate_need_pcie_device(void *opaque)
2558 {
2559 VMXNET3State *s = VMXNET3(opaque);
2560
2561 return !(s->compat_flags & VMXNET3_COMPAT_FLAG_DISABLE_PCIE);
2562 }
2563
2564 static bool vmxnet3_vmstate_test_pci_device(void *opaque, int version_id)
2565 {
2566 return !vmxnet3_vmstate_need_pcie_device(opaque);
2567 }
2568
2569 static const VMStateDescription vmstate_vmxnet3_pcie_device = {
2570 .name = "vmxnet3/pcie",
2571 .version_id = 1,
2572 .minimum_version_id = 1,
2573 .needed = vmxnet3_vmstate_need_pcie_device,
2574 .fields = (VMStateField[]) {
2575 VMSTATE_PCI_DEVICE(parent_obj, VMXNET3State),
2576 VMSTATE_END_OF_LIST()
2577 }
2578 };
2579
2580 static const VMStateDescription vmstate_vmxnet3 = {
2581 .name = "vmxnet3",
2582 .version_id = 1,
2583 .minimum_version_id = 1,
2584 .pre_save = vmxnet3_pre_save,
2585 .post_load = vmxnet3_post_load,
2586 .fields = (VMStateField[]) {
2587 VMSTATE_STRUCT_TEST(parent_obj, VMXNET3State,
2588 vmxnet3_vmstate_test_pci_device, 0,
2589 vmstate_pci_device, PCIDevice),
2590 VMSTATE_BOOL(rx_packets_compound, VMXNET3State),
2591 VMSTATE_BOOL(rx_vlan_stripping, VMXNET3State),
2592 VMSTATE_BOOL(lro_supported, VMXNET3State),
2593 VMSTATE_UINT32(rx_mode, VMXNET3State),
2594 VMSTATE_UINT32(mcast_list_len, VMXNET3State),
2595 VMSTATE_UINT32(mcast_list_buff_size, VMXNET3State),
2596 VMSTATE_UINT32_ARRAY(vlan_table, VMXNET3State, VMXNET3_VFT_SIZE),
2597 VMSTATE_UINT32(mtu, VMXNET3State),
2598 VMSTATE_UINT16(max_rx_frags, VMXNET3State),
2599 VMSTATE_UINT32(max_tx_frags, VMXNET3State),
2600 VMSTATE_UINT8(event_int_idx, VMXNET3State),
2601 VMSTATE_BOOL(auto_int_masking, VMXNET3State),
2602 VMSTATE_UINT8(txq_num, VMXNET3State),
2603 VMSTATE_UINT8(rxq_num, VMXNET3State),
2604 VMSTATE_UINT32(device_active, VMXNET3State),
2605 VMSTATE_UINT32(last_command, VMXNET3State),
2606 VMSTATE_UINT32(link_status_and_speed, VMXNET3State),
2607 VMSTATE_UINT32(temp_mac, VMXNET3State),
2608 VMSTATE_UINT64(drv_shmem, VMXNET3State),
2609 VMSTATE_UINT64(temp_shared_guest_driver_memory, VMXNET3State),
2610
2611 VMSTATE_STRUCT_ARRAY(txq_descr, VMXNET3State,
2612 VMXNET3_DEVICE_MAX_TX_QUEUES, 0, vmstate_vmxnet3_txq_descr,
2613 Vmxnet3TxqDescr),
2614 VMSTATE_STRUCT_ARRAY(rxq_descr, VMXNET3State,
2615 VMXNET3_DEVICE_MAX_RX_QUEUES, 0, vmstate_vmxnet3_rxq_descr,
2616 Vmxnet3RxqDescr),
2617 VMSTATE_STRUCT_ARRAY(interrupt_states, VMXNET3State,
2618 VMXNET3_MAX_INTRS, 0, vmstate_vmxnet3_int_state,
2619 Vmxnet3IntState),
2620
2621 VMSTATE_END_OF_LIST()
2622 },
2623 .subsections = (const VMStateDescription*[]) {
2624 &vmxstate_vmxnet3_mcast_list,
2625 &vmstate_vmxnet3_pcie_device,
2626 NULL
2627 }
2628 };
2629
2630 static Property vmxnet3_properties[] = {
2631 DEFINE_NIC_PROPERTIES(VMXNET3State, conf),
2632 DEFINE_PROP_BIT("x-old-msi-offsets", VMXNET3State, compat_flags,
2633 VMXNET3_COMPAT_FLAG_OLD_MSI_OFFSETS_BIT, false),
2634 DEFINE_PROP_BIT("x-disable-pcie", VMXNET3State, compat_flags,
2635 VMXNET3_COMPAT_FLAG_DISABLE_PCIE_BIT, false),
2636 DEFINE_PROP_END_OF_LIST(),
2637 };
2638
2639 static void vmxnet3_realize(DeviceState *qdev, Error **errp)
2640 {
2641 VMXNET3Class *vc = VMXNET3_DEVICE_GET_CLASS(qdev);
2642 PCIDevice *pci_dev = PCI_DEVICE(qdev);
2643 VMXNET3State *s = VMXNET3(qdev);
2644
2645 if (!(s->compat_flags & VMXNET3_COMPAT_FLAG_DISABLE_PCIE)) {
2646 pci_dev->cap_present |= QEMU_PCI_CAP_EXPRESS;
2647 }
2648
2649 vc->parent_dc_realize(qdev, errp);
2650 }
2651
2652 static void vmxnet3_class_init(ObjectClass *class, void *data)
2653 {
2654 DeviceClass *dc = DEVICE_CLASS(class);
2655 PCIDeviceClass *c = PCI_DEVICE_CLASS(class);
2656 VMXNET3Class *vc = VMXNET3_DEVICE_CLASS(class);
2657
2658 c->realize = vmxnet3_pci_realize;
2659 c->exit = vmxnet3_pci_uninit;
2660 c->vendor_id = PCI_VENDOR_ID_VMWARE;
2661 c->device_id = PCI_DEVICE_ID_VMWARE_VMXNET3;
2662 c->revision = PCI_DEVICE_ID_VMWARE_VMXNET3_REVISION;
2663 c->romfile = "efi-vmxnet3.rom";
2664 c->class_id = PCI_CLASS_NETWORK_ETHERNET;
2665 c->subsystem_vendor_id = PCI_VENDOR_ID_VMWARE;
2666 c->subsystem_id = PCI_DEVICE_ID_VMWARE_VMXNET3;
2667 device_class_set_parent_realize(dc, vmxnet3_realize,
2668 &vc->parent_dc_realize);
2669 dc->desc = "VMWare Paravirtualized Ethernet v3";
2670 dc->reset = vmxnet3_qdev_reset;
2671 dc->vmsd = &vmstate_vmxnet3;
2672 dc->props = vmxnet3_properties;
2673 set_bit(DEVICE_CATEGORY_NETWORK, dc->categories);
2674 }
2675
2676 static const TypeInfo vmxnet3_info = {
2677 .name = TYPE_VMXNET3,
2678 .parent = TYPE_PCI_DEVICE,
2679 .class_size = sizeof(VMXNET3Class),
2680 .instance_size = sizeof(VMXNET3State),
2681 .class_init = vmxnet3_class_init,
2682 .instance_init = vmxnet3_instance_init,
2683 .interfaces = (InterfaceInfo[]) {
2684 { INTERFACE_PCIE_DEVICE },
2685 { INTERFACE_CONVENTIONAL_PCI_DEVICE },
2686 { }
2687 },
2688 };
2689
2690 static void vmxnet3_register_types(void)
2691 {
2692 VMW_CBPRN("vmxnet3_register_types called...");
2693 type_register_static(&vmxnet3_info);
2694 }
2695
2696 type_init(vmxnet3_register_types)
QEmu