-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathif_sume.c
1628 lines (1379 loc) · 47 KB
/
if_sume.c
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
934
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
956
957
958
959
960
961
962
963
964
965
966
967
968
969
970
971
972
973
974
975
976
977
978
979
980
981
982
983
984
985
986
987
988
989
990
991
992
993
994
995
996
997
998
999
1000
/*-
* SPDX-License-Identifier: BSD-2-Clause-FreeBSD
*
* Copyright (c) 2015 Bjoern A. Zeeb
* Copyright (c) 2020 Denis Salopek
*
* This software was developed by SRI International and the University of
* Cambridge Computer Laboratory under DARPA/AFRL contract FA8750-11-C-0249
* ("MRC2"), as part of the DARPA MRC research programme.
*
* Redistribution and use in source and binary forms, with or without
* modification, are permitted provided that the following conditions
* are met:
* 1. Redistributions of source code must retain the above copyright
* notice, this list of conditions and the following disclaimer.
* 2. Redistributions in binary form must reproduce the above copyright
* notice, this list of conditions and the following disclaimer in the
* documentation and/or other materials provided with the distribution.
*
* THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
* AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
* IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
* ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE
* LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
* CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
* SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
* INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
* CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
* ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
* POSSIBILITY OF SUCH DAMAGE.
*/
#include <sys/cdefs.h>
__FBSDID("$FreeBSD: $");
#include <sys/param.h>
#include <sys/bus.h>
#include <sys/endian.h>
#include <sys/kernel.h>
#include <sys/limits.h>
#include <sys/module.h>
#include <sys/rman.h>
#include <sys/socket.h>
#include <sys/sockio.h>
#include <sys/sysctl.h>
#include <sys/taskqueue.h>
#include <net/if.h>
#include <net/if_media.h>
#include <net/if_types.h>
#include <net/if_var.h>
#include <netinet/in.h>
#include <netinet/if_ether.h>
#include <dev/pci/pcivar.h>
#include <dev/pci/pcireg.h>
#include <machine/bus.h>
#include "adapter.h"
#define PCI_VENDOR_ID_XILINX 0x10ee
#define PCI_DEVICE_ID_SUME 0x7028
/* SUME bus driver interface */
static int sume_probe(device_t);
static int sume_attach(device_t);
static int sume_detach(device_t);
static device_method_t sume_methods[] = {
DEVMETHOD(device_probe, sume_probe),
DEVMETHOD(device_attach, sume_attach),
DEVMETHOD(device_detach, sume_detach),
DEVMETHOD_END
};
static driver_t sume_driver = {
"sume",
sume_methods,
sizeof(struct sume_adapter)
};
/*
* The DMA engine for SUME generates interrupts for each RX/TX transaction.
* Depending on the channel (0 if packet transaction, 1 if register transaction)
* the used bits of the interrupt vector will be the lowest or the second lowest
* 5 bits.
*
* When receiving packets from SUME (RX):
* (1) SUME received a packet on one of the interfaces.
* (2) SUME generates an interrupt vector, bit 00001 is set (channel 0 - new RX
* transaction).
* (3) We read the length of the incoming packet and the offset along with the
* 'last' flag from the SUME registers.
* (4) We prepare for the DMA transaction by setting the bouncebuffer on the
* address buf_addr. For now, this is how it's done:
* - First 3*sizeof(uint32_t) bytes are: lower and upper 32 bits of physical
* address where we want the data to arrive (buf_addr[0] and buf_addr[1]),
* and length of incoming data (buf_addr[2]).
* - Data will start right after, at buf_addr+3*sizeof(uint32_t). The
* physical address buf_hw_addr is a block of contiguous memory mapped to
* buf_addr, so we can set the incoming data's physical address (buf_addr[0]
* and buf_addr[1]) to buf_hw_addr+3*sizeof(uint32_t).
* (5) We notify SUME that the bouncebuffer is ready for the transaction by
* writing the lower/upper physical address buf_hw_addr to the SUME
* registers RIFFA_TX_SG_ADDR_LO_REG_OFF and RIFFA_TX_SG_ADDR_HI_REG_OFF as
* well as the number of segments to the register RIFFA_TX_SG_LEN_REG_OFF.
* (6) SUME generates an interrupt vector, bit 00010 is set (channel 0 -
* bouncebuffer received).
* (7) SUME generates an interrupt vector, bit 00100 is set (channel 0 -
* transaction is done).
* (8) SUME can do both steps (6) and (7) using the same interrupt.
* (8) We read the first 16 bytes (metadata) of the received data and note the
* incoming interface so we can later forward it to the right one in the OS
* (sume0, sume1, sume2 or sume3).
* (10) We create an mbuf and copy the data from the bouncebuffer to the mbuf
* and set the mbuf rcvif to the incoming interface.
* (11) We forward the mbuf to the appropriate interface via ifp->if_input.
*
* When sending packets to SUME (TX):
* (1) The OS calls sume_if_start() function on TX.
* (2) We get the mbuf packet data and copy it to the
* buf_addr+3*sizeof(uint32_t) + metadata 16 bytes.
* (3) We create the metadata based on the output interface and copy it to the
* buf_addr+3*sizeof(uint32_t).
* (4) We write the offset/last and length of the packet to the SUME registers
* RIFFA_RX_OFFLAST_REG_OFF and RIFFA_RX_LEN_REG_OFF.
* (5) We fill the bouncebuffer by filling the first 3*sizeof(uint32_t) bytes
* with the physical address and length just as in RX step (4).
* (6) We notify SUME that the bouncebuffer is ready by writing to SUME
* registers RIFFA_RX_SG_ADDR_LO_REG_OFF, RIFFA_RX_SG_ADDR_HI_REG_OFF and
* RIFFA_RX_SG_LEN_REG_OFF just as in RX step (5).
* (7) SUME generates an interrupt vector, bit 01000 is set (channel 0 -
* bouncebuffer is read).
* (8) SUME generates an interrupt vector, bit 10000 is set (channel 0 -
* transaction is done).
* (9) SUME can do both steps (7) and (8) using the same interrupt.
*
* Internal registers
* Every module in the SUME hardware has its own set of internal registers
* (IDs, for debugging and statistic purposes, etc.). Their base addresses are
* defined in 'projects/reference_nic/hw/tcl/reference_nic_defines.tcl' and the
* offsets to different memory locations of every module are defined in their
* corresponding folder inside the library. These registers can be RO/RW and
* there is a special method to fetch/change this data over 1 or 2 DMA
* transactions. For writing, by calling the sume_module_reg_write(). For
* reading, by calling the sume_module_reg_write() and then
* sume_module_reg_read(). Check those functions for more information.
*/
MALLOC_DECLARE(M_SUME);
MALLOC_DEFINE(M_SUME, "sume", "NetFPGA SUME device driver");
static void check_tx_queues(struct sume_adapter *);
static void sume_fill_bb_desc(struct sume_adapter *, struct riffa_chnl_dir *,
uint64_t);
static struct unrhdr *unr;
static struct {
uint16_t device;
char *desc;
} sume_pciids[] = {
{PCI_DEVICE_ID_SUME, "NetFPGA SUME reference NIC"},
};
static inline uint32_t
read_reg(struct sume_adapter *adapter, int offset)
{
return (bus_space_read_4(adapter->bt, adapter->bh, offset << 2));
}
static inline void
write_reg(struct sume_adapter *adapter, int offset, uint32_t val)
{
bus_space_write_4(adapter->bt, adapter->bh, offset << 2, val);
}
static int
sume_probe(device_t dev)
{
int i;
uint16_t v = pci_get_vendor(dev);
uint16_t d = pci_get_device(dev);
if (v != PCI_VENDOR_ID_XILINX)
return (ENXIO);
for (i = 0; i < nitems(sume_pciids); i++) {
if (d == sume_pciids[i].device) {
device_set_desc(dev, sume_pciids[i].desc);
return (BUS_PROBE_DEFAULT);
}
}
return (ENXIO);
}
/*
* Building mbuf for packet received from SUME. We expect to receive 'len'
* bytes of data (including metadata) written from the bouncebuffer address
* buf_addr+3*sizeof(uint32_t). Metadata will tell us which SUME interface
* received the packet (sport will be 1, 2, 4 or 8), the packet length (plen),
* and the magic word needs to be 0xcafe. When we have the packet data, we
* create an mbuf and copy the data to it using m_copyback() function, set the
* correct interface to rcvif and return the mbuf to be later sent to the OS
* with if_input.
*/
static struct mbuf *
sume_rx_build_mbuf(struct sume_adapter *adapter, uint32_t len)
{
struct nf_priv *nf_priv;
struct mbuf *m;
struct ifnet *ifp = NULL;
int np;
uint16_t dport, plen, magic;
device_t dev = adapter->dev;
uint8_t *indata = (uint8_t *)
adapter->recv[SUME_RIFFA_CHANNEL_DATA]->buf_addr +
sizeof(struct nf_bb_desc);
struct nf_metadata *mdata = (struct nf_metadata *) indata;
/* The metadata header is 16 bytes. */
if (len < sizeof(struct nf_metadata)) {
device_printf(dev, "short frame (%d)\n", len);
adapter->packets_err++;
adapter->bytes_err += len;
return (NULL);
}
dport = le16toh(mdata->dport);
plen = le16toh(mdata->plen);
magic = le16toh(mdata->magic);
if (sizeof(struct nf_metadata) + plen > len ||
magic != SUME_RIFFA_MAGIC) {
device_printf(dev, "corrupted packet (%zd + %d > %d || magic "
"0x%04x != 0x%04x)\n", sizeof(struct nf_metadata), plen,
len, magic, SUME_RIFFA_MAGIC);
return (NULL);
}
/* We got the packet from one of the even bits */
np = (ffs(dport & SUME_DPORT_MASK) >> 1) - 1;
if (np < 0 || np > SUME_NPORTS) {
device_printf(dev, "invalid destination port 0x%04x (%d)\n",
dport, np);
adapter->packets_err++;
adapter->bytes_err += plen;
return (NULL);
}
ifp = adapter->ifp[np];
nf_priv = ifp->if_softc;
nf_priv->stats.rx_packets++;
nf_priv->stats.rx_bytes += plen;
/* If the interface is down, well, we are done. */
if (!(ifp->if_flags & IFF_UP)) {
nf_priv->update_ctr = 1;
nf_priv->stats.ifc_down_packets++;
nf_priv->stats.ifc_down_bytes += plen;
return (NULL);
}
if (adapter->sume_debug)
device_printf(dev, "building mbuf with length: %d\n", plen);
m = m_getm(NULL, plen, M_NOWAIT, MT_DATA);
if (m == NULL) {
adapter->packets_err++;
adapter->bytes_err += plen;
return (NULL);
}
/* Copy the data in at the right offset. */
m_copyback(m, 0, plen, (void *) (indata + sizeof(struct nf_metadata)));
m->m_pkthdr.rcvif = ifp;
return (m);
}
/*
* SUME interrupt handler for when we get a valid interrupt from the board.
* Theoretically, we can receive interrupt for any of the available channels,
* but RIFFA DMA uses only 2: 0 and 1, so we use only vect0. The vector is a 32
* bit number, using 5 bits for every channel, the least significant bits
* correspond to channel 0 and the next 5 bits correspond to channel 1. Vector
* bits for RX/TX are:
* RX
* bit 0 - new transaction from SUME
* bit 1 - SUME received our bouncebuffer address
* bit 2 - SUME copied the received data to our bouncebuffer, transaction done
* TX
* bit 3 - SUME received our bouncebuffer address
* bit 4 - SUME copied the data from our bouncebuffer, transaction done
*
* There are two finite state machines (one for TX, one for RX). We loop
* through channels 0 and 1 to check and our current state and which interrupt
* bit is set.
* TX
* SUME_RIFFA_CHAN_STATE_IDLE: waiting for the first TX transaction.
* SUME_RIFFA_CHAN_STATE_READY: we prepared (filled with data) the bouncebuffer
* and triggered the SUME for the TX transaction. Waiting for interrupt bit 3
* to go to the next state.
* SUME_RIFFA_CHAN_STATE_READ: waiting for interrupt bit 4 (for SUME to send
* our packet). Then we get the length of the sent data and go back to the
* IDLE state.
* RX
* SUME_RIFFA_CHAN_STATE_IDLE: waiting for the interrupt bit 0 (new RX
* transaction). When we get it, we prepare our bouncebuffer for reading and
* trigger the SUME to start the transaction. Go to the next state.
* SUME_RIFFA_CHAN_STATE_READY: waiting for the interrupt bit 1 (SUME got our
* bouncebuffer). Go to the next state.
* SUME_RIFFA_CHAN_STATE_READ: SUME copied data and our bouncebuffer is ready,
* we can build the mbuf and go back to the IDLE state.
*/
static void
sume_intr_handler(void *arg)
{
struct sume_adapter *adapter = arg;
uint32_t vect, vect0, len;
int ch, loops;
device_t dev = adapter->dev;
struct mbuf *m = NULL;
struct ifnet *ifp = NULL;
struct riffa_chnl_dir *send, *recv;
SUME_LOCK(adapter);
vect0 = read_reg(adapter, RIFFA_IRQ_REG0_OFF);
if ((vect0 & SUME_INVALID_VECT) != 0) {
SUME_UNLOCK(adapter);
return;
}
/*
* We only have one interrupt for all channels and no way
* to quickly lookup for which channel(s) we got an interrupt?
*/
for (ch = 0; ch < SUME_RIFFA_CHANNELS; ch++) {
vect = vect0 >> (5 * ch);
send = adapter->send[ch];
recv = adapter->recv[ch];
loops = 0;
while ((vect & (SUME_MSI_TXBUF | SUME_MSI_TXDONE)) &&
loops <= 5) {
if (adapter->sume_debug)
device_printf(dev, "TX ch %d state %u vect = "
"0x%08x\n", ch, send->state, vect);
switch (send->state) {
case SUME_RIFFA_CHAN_STATE_IDLE:
break;
case SUME_RIFFA_CHAN_STATE_READY:
if (!(vect & SUME_MSI_TXBUF)) {
device_printf(dev, "ch %d unexpected "
"interrupt in send+3 state %u: "
"vect = 0x%08x\n", ch, send->state,
vect);
send->recovery = 1;
break;
}
send->state = SUME_RIFFA_CHAN_STATE_READ;
vect &= ~SUME_MSI_TXBUF;
break;
case SUME_RIFFA_CHAN_STATE_READ:
if (!(vect & SUME_MSI_TXDONE)) {
device_printf(dev, "ch %d unexpected "
"interrupt in send+4 state %u: "
"vect = 0x%08x\n", ch, send->state,
vect);
send->recovery = 1;
break;
}
send->state = SUME_RIFFA_CHAN_STATE_LEN;
len = read_reg(adapter, RIFFA_CHNL_REG(ch,
RIFFA_RX_TNFR_LEN_REG_OFF));
if (ch == SUME_RIFFA_CHANNEL_DATA) {
send->state =
SUME_RIFFA_CHAN_STATE_IDLE;
check_tx_queues(adapter);
} else if (ch == SUME_RIFFA_CHANNEL_REG)
wakeup(&send->event);
else {
device_printf(dev, "ch %d unexpected "
"interrupt in send+4 state %u: "
"vect = 0x%08x\n", ch, send->state,
vect);
send->recovery = 1;
}
vect &= ~SUME_MSI_TXDONE;
break;
case SUME_RIFFA_CHAN_STATE_LEN:
break;
default:
device_printf(dev, "unknown TX state!\n");
}
loops++;
}
if ((vect & (SUME_MSI_TXBUF | SUME_MSI_TXDONE)) &&
send->recovery) {
send->recovery_ctr++;
device_printf(dev, "ch %d ignoring vect = 0x%08x "
"during TX; not in recovery; state = %d loops = "
"%d\n", ch, vect, send->state, loops);
}
loops = 0;
while ((vect & (SUME_MSI_RXQUE | SUME_MSI_RXBUF |
SUME_MSI_RXDONE)) && loops < 5) {
if (adapter->sume_debug)
device_printf(dev, "RX ch %d state %u vect = "
"0x%08x\n", ch, recv->state, vect);
switch (recv->state) {
case SUME_RIFFA_CHAN_STATE_IDLE:
if (!(vect & SUME_MSI_RXQUE)) {
device_printf(dev, "ch %d unexpected "
"interrupt in recv+0 state %u: "
"vect = 0x%08x\n", ch, recv->state,
vect);
recv->recovery = 1;
break;
}
uint32_t max_ptr;
/* Clear recovery state. */
recv->recovery = 0;
/* Get offset and length. */
recv->offlast = read_reg(adapter,
RIFFA_CHNL_REG(ch,
RIFFA_TX_OFFLAST_REG_OFF));
recv->len = read_reg(adapter, RIFFA_CHNL_REG(ch,
RIFFA_TX_LEN_REG_OFF));
/* Boundary checks. */
max_ptr = (uint32_t)((uintptr_t)recv->buf_addr
+ SUME_RIFFA_OFFSET(recv->offlast)
+ SUME_RIFFA_LEN(recv->len) - 1);
if (max_ptr <
(uint32_t)((uintptr_t)recv->buf_addr))
device_printf(dev, "receive buffer "
"wrap-around overflow.\n");
if (SUME_RIFFA_OFFSET(recv->offlast) +
SUME_RIFFA_LEN(recv->len) >
adapter->sg_buf_size)
device_printf(dev, "receive buffer too"
" small.\n");
/* Fill the bouncebuf "descriptor". */
sume_fill_bb_desc(adapter, recv,
SUME_RIFFA_LEN(recv->len));
bus_dmamap_sync(recv->ch_tag, recv->ch_map,
BUS_DMASYNC_PREREAD |
BUS_DMASYNC_PREWRITE);
write_reg(adapter, RIFFA_CHNL_REG(ch,
RIFFA_TX_SG_ADDR_LO_REG_OFF),
SUME_RIFFA_LO_ADDR(recv->buf_hw_addr));
write_reg(adapter, RIFFA_CHNL_REG(ch,
RIFFA_TX_SG_ADDR_HI_REG_OFF),
SUME_RIFFA_HI_ADDR(recv->buf_hw_addr));
write_reg(adapter, RIFFA_CHNL_REG(ch,
RIFFA_TX_SG_LEN_REG_OFF),
4 * recv->num_sg);
bus_dmamap_sync(recv->ch_tag, recv->ch_map,
BUS_DMASYNC_POSTREAD |
BUS_DMASYNC_POSTWRITE);
recv->state = SUME_RIFFA_CHAN_STATE_READY;
vect &= ~SUME_MSI_RXQUE;
break;
case SUME_RIFFA_CHAN_STATE_READY:
if (!(vect & SUME_MSI_RXBUF)) {
device_printf(dev, "ch %d unexpected "
"interrupt in recv+1 state %u: "
"vect = 0x%08x\n", ch, recv->state,
vect);
recv->recovery = 1;
break;
}
recv->state = SUME_RIFFA_CHAN_STATE_READ;
vect &= ~SUME_MSI_RXBUF;
break;
case SUME_RIFFA_CHAN_STATE_READ:
if (!(vect & SUME_MSI_RXDONE)) {
device_printf(dev, "ch %d unexpected "
"interrupt in recv+2 state %u: "
"vect = 0x%08x\n", ch, recv->state,
vect);
recv->recovery = 1;
break;
}
len = read_reg(adapter, RIFFA_CHNL_REG(ch,
RIFFA_TX_TNFR_LEN_REG_OFF));
/* Remember, len and recv->len are words. */
if (ch == SUME_RIFFA_CHANNEL_DATA) {
m = sume_rx_build_mbuf(adapter,
len << 2);
recv->state =
SUME_RIFFA_CHAN_STATE_IDLE;
} else if (ch == SUME_RIFFA_CHANNEL_REG)
wakeup(&recv->event);
else {
device_printf(dev, "ch %d unexpected "
"interrupt in recv+2 state %u: "
"vect = 0x%08x\n", ch, recv->state,
vect);
recv->recovery = 1;
}
vect &= ~SUME_MSI_RXDONE;
break;
case SUME_RIFFA_CHAN_STATE_LEN:
break;
default:
device_printf(dev, "unknown RX state!\n");
}
loops++;
}
if ((vect & (SUME_MSI_RXQUE | SUME_MSI_RXBUF |
SUME_MSI_RXDONE)) && recv->recovery) {
recv->recovery_ctr++;
device_printf(dev, "ch %d ignoring vect = 0x%08x "
"during RX; not in recovery; state = %d, loops = "
"%d\n", ch, vect, recv->state, loops);
/* Clean the unfinished transaction. */
if (ch == SUME_RIFFA_CHANNEL_REG &&
vect & SUME_MSI_RXDONE) {
read_reg(adapter, RIFFA_CHNL_REG(ch,
RIFFA_TX_TNFR_LEN_REG_OFF));
recv->recovery = 0;
} else if (ch == SUME_RIFFA_CHANNEL_DATA &&
vect & SUME_MSI_RXQUE) {
recv->recovery = 0;
recv->state = SUME_RIFFA_CHAN_STATE_IDLE;
}
}
}
SUME_UNLOCK(adapter);
if (m != NULL) {
ifp = m->m_pkthdr.rcvif;
(*ifp->if_input)(ifp, m);
}
}
/*
* As we cannot disable interrupt generation, ignore early interrupts by waiting
* for the adapter to go into the 'running' state.
*/
static int
sume_intr_filter(void *arg)
{
struct sume_adapter *adapter = arg;
if (adapter->running == 0)
return (FILTER_STRAY);
return (FILTER_SCHEDULE_THREAD);
}
static int
sume_probe_riffa_pci(struct sume_adapter *adapter)
{
device_t dev = adapter->dev;
int error, count, capmem;
uint32_t reg, devctl, linkctl;
pci_enable_busmaster(dev);
adapter->rid = PCIR_BAR(0);
adapter->bar0_addr = bus_alloc_resource_any(dev, SYS_RES_MEMORY,
&adapter->rid, RF_ACTIVE);
if (adapter->bar0_addr == NULL) {
device_printf(dev, "unable to allocate bus resource: "
"BAR0 address\n");
return (ENXIO);
}
adapter->bt = rman_get_bustag(adapter->bar0_addr);
adapter->bh = rman_get_bushandle(adapter->bar0_addr);
adapter->bar0_len = rman_get_size(adapter->bar0_addr);
if (adapter->bar0_len != 1024) {
device_printf(dev, "BAR0 resource length %lu != 1024\n",
adapter->bar0_len);
return (ENXIO);
}
count = pci_msi_count(dev);
error = pci_alloc_msi(dev, &count);
if (error) {
device_printf(dev, "unable to allocate bus resource: PCI "
"MSI\n");
return (error);
}
adapter->irq.rid = 1; /* Should be 1, thus says pci_alloc_msi() */
adapter->irq.res = bus_alloc_resource_any(dev, SYS_RES_IRQ,
&adapter->irq.rid, RF_SHAREABLE | RF_ACTIVE);
if (adapter->irq.res == NULL) {
device_printf(dev, "unable to allocate bus resource: IRQ "
"memory\n");
return (ENXIO);
}
error = bus_setup_intr(dev, adapter->irq.res, INTR_MPSAFE |
INTR_TYPE_NET, sume_intr_filter, sume_intr_handler, adapter,
&adapter->irq.tag);
if (error) {
device_printf(dev, "failed to setup interrupt for rid %d, name"
" %s: %d\n", adapter->irq.rid, "SUME_INTR", error);
return (ENXIO);
}
if (pci_find_cap(dev, PCIY_EXPRESS, &capmem) != 0) {
device_printf(dev, "PCI not PCIe capable\n");
return (ENXIO);
}
devctl = pci_read_config(dev, capmem + PCIER_DEVICE_CTL, 2);
pci_write_config(dev, capmem + PCIER_DEVICE_CTL, (devctl |
PCIEM_CTL_EXT_TAG_FIELD), 2);
devctl = pci_read_config(dev, capmem + PCIER_DEVICE_CTL2, 2);
pci_write_config(dev, capmem + PCIER_DEVICE_CTL2, (devctl |
PCIEM_CTL2_ID_ORDERED_REQ_EN), 2);
linkctl = pci_read_config(dev, capmem + PCIER_LINK_CTL, 2);
pci_write_config(dev, capmem + PCIER_LINK_CTL, (linkctl |
PCIEM_LINK_CTL_RCB), 2);
reg = read_reg(adapter, RIFFA_INFO_REG_OFF);
adapter->num_sg = RIFFA_SG_ELEMS * ((reg >> 19) & 0xf);
adapter->sg_buf_size = RIFFA_SG_BUF_SIZE * ((reg >> 19) & 0xf);
error = ENODEV;
/* Check bus master is enabled. */
if (((reg >> 4) & 0x1) != 1) {
device_printf(dev, "bus master not enabled: %d\n",
(reg >> 4) & 0x1);
return (error);
}
/* Check link parameters are valid. */
if (((reg >> 5) & 0x3f) == 0 || ((reg >> 11) & 0x3) == 0) {
device_printf(dev, "link parameters not valid: %d %d\n",
(reg >> 5) & 0x3f, (reg >> 11) & 0x3);
return (error);
}
/* Check # of channels are within valid range. */
if ((reg & 0xf) == 0 || (reg & 0xf) > RIFFA_MAX_CHNLS) {
device_printf(dev, "number of channels out of range: %d\n",
reg & 0xf);
return (error);
}
/* Check bus width. */
if (((reg >> 19) & 0xf) == 0 ||
((reg >> 19) & 0xf) > RIFFA_MAX_BUS_WIDTH_PARAM) {
device_printf(dev, "bus width out of range: %d\n",
(reg >> 19) & 0xf);
return (error);
}
device_printf(dev, "[riffa] # of channels: %d\n",
reg & 0xf);
device_printf(dev, "[riffa] bus interface width: %d\n",
((reg >> 19) & 0xf) << 5);
device_printf(dev, "[riffa] bus master enabled: %d\n",
(reg >> 4) & 0x1);
device_printf(dev, "[riffa] negotiated link width: %d\n",
(reg >> 5) & 0x3f);
device_printf(dev, "[riffa] negotiated rate width: %d MTs\n",
((reg >> 11) & 0x3) * 2500);
device_printf(dev, "[riffa] max downstream payload: %d B\n",
128 << ((reg >> 13) & 0x7));
device_printf(dev, "[riffa] max upstream payload: %d B\n",
128 << ((reg >> 16) & 0x7));
return (0);
}
/* If there is no sume_if_init, the ether_ioctl panics. */
static void
sume_if_init(void *sc)
{
}
/* Write the address and length for our incoming / outgoing transaction. */
static void
sume_fill_bb_desc(struct sume_adapter *adapter, struct riffa_chnl_dir *p,
uint64_t len)
{
struct nf_bb_desc *bouncebuf = (struct nf_bb_desc *) p->buf_addr;
bouncebuf->lower = (p->buf_hw_addr + sizeof(struct nf_bb_desc));
bouncebuf->upper = (p->buf_hw_addr + sizeof(struct nf_bb_desc)) >> 32;
bouncebuf->len = len >> 2;
}
/* Module register locked write. */
static int
sume_modreg_write_locked(struct sume_adapter *adapter)
{
struct riffa_chnl_dir *send = adapter->send[SUME_RIFFA_CHANNEL_REG];
/* Let the FPGA know about the transfer. */
write_reg(adapter, RIFFA_CHNL_REG(SUME_RIFFA_CHANNEL_REG,
RIFFA_RX_OFFLAST_REG_OFF), SUME_OFFLAST);
write_reg(adapter, RIFFA_CHNL_REG(SUME_RIFFA_CHANNEL_REG,
RIFFA_RX_LEN_REG_OFF), send->len); /* words */
/* Fill the bouncebuf "descriptor". */
sume_fill_bb_desc(adapter, send, SUME_RIFFA_LEN(send->len));
/* Update the state before intiating the DMA to avoid races. */
send->state = SUME_RIFFA_CHAN_STATE_READY;
bus_dmamap_sync(send->ch_tag, send->ch_map,
BUS_DMASYNC_PREREAD | BUS_DMASYNC_PREWRITE);
/* DMA. */
write_reg(adapter, RIFFA_CHNL_REG(SUME_RIFFA_CHANNEL_REG,
RIFFA_RX_SG_ADDR_LO_REG_OFF),
SUME_RIFFA_LO_ADDR(send->buf_hw_addr));
write_reg(adapter, RIFFA_CHNL_REG(SUME_RIFFA_CHANNEL_REG,
RIFFA_RX_SG_ADDR_HI_REG_OFF),
SUME_RIFFA_HI_ADDR(send->buf_hw_addr));
write_reg(adapter, RIFFA_CHNL_REG(SUME_RIFFA_CHANNEL_REG,
RIFFA_RX_SG_LEN_REG_OFF), 4 * send->num_sg);
bus_dmamap_sync(send->ch_tag, send->ch_map,
BUS_DMASYNC_POSTREAD | BUS_DMASYNC_POSTWRITE);
return (0);
}
/*
* Request a register read or write (depending on optype).
* If optype is set (0x1f) this will result in a register write,
* otherwise this will result in a register read request at the given
* address and the result will need to be DMAed back.
*/
static int
sume_module_reg_write(struct nf_priv *nf_priv, struct sume_ifreq *sifr,
uint32_t optype)
{
struct sume_adapter *adapter = nf_priv->adapter;
struct riffa_chnl_dir *send = adapter->send[SUME_RIFFA_CHANNEL_REG];
struct nf_regop_data *data;
int error;
/*
* 1. Make sure the channel is free; otherwise return EBUSY.
* 2. Prepare the memory in the bounce buffer (which we always
* use for regs).
* 3. Start the DMA process.
* 4. Sleep and wait for result and return success or error.
*/
SUME_LOCK(adapter);
if (send->state != SUME_RIFFA_CHAN_STATE_IDLE) {
SUME_UNLOCK(adapter);
return (EBUSY);
}
data = (struct nf_regop_data *) (send->buf_addr +
sizeof(struct nf_bb_desc));
data->addr = htole32(sifr->addr);
data->val = htole32(sifr->val);
/* Tag to indentify request. */
data->rtag = htole32(++send->rtag);
data->optype = htole32(optype);
send->len = sizeof(struct nf_regop_data) / 4; /* words */
error = sume_modreg_write_locked(adapter);
if (error) {
SUME_UNLOCK(adapter);
return (EFAULT);
}
/* Timeout after 1s. */
if (send->state != SUME_RIFFA_CHAN_STATE_LEN)
error = msleep(&send->event, &adapter->lock, 0,
"Waiting recv finish", 1 * hz);
/* This was a write so we are done; were interrupted, or timed out. */
if (optype != SUME_MR_READ || error != 0 || error == EWOULDBLOCK) {
send->state = SUME_RIFFA_CHAN_STATE_IDLE;
if (optype == SUME_MR_READ)
error = EWOULDBLOCK;
else
error = 0;
} else
error = 0;
/*
* For read requests we will update state once we are done
* having read the result to avoid any two outstanding
* transactions, or we need a queue and validate tags,
* which is a lot of work for a low priority, infrequent
* event.
*/
SUME_UNLOCK(adapter);
return (error);
}
/* Module register read. */
static int
sume_module_reg_read(struct nf_priv *nf_priv, struct sume_ifreq *sifr)
{
struct sume_adapter *adapter = nf_priv->adapter;
struct riffa_chnl_dir *recv = adapter->recv[SUME_RIFFA_CHANNEL_REG];
struct riffa_chnl_dir *send = adapter->send[SUME_RIFFA_CHANNEL_REG];
struct nf_regop_data *data;
int error = 0;
/*
* 0. Sleep waiting for result if needed (unless condition is
* true already).
* 1. Read DMA results.
* 2. Update state on *TX* to IDLE to allow next read to start.
*/
SUME_LOCK(adapter);
bus_dmamap_sync(recv->ch_tag, recv->ch_map,
BUS_DMASYNC_PREREAD | BUS_DMASYNC_PREWRITE);
/*
* We only need to be woken up at the end of the transaction.
* Timeout after 1s.
*/
if (recv->state != SUME_RIFFA_CHAN_STATE_READ)
error = msleep(&recv->event, &adapter->lock, 0,
"Waiting transaction finish", 1 * hz);
if (recv->state != SUME_RIFFA_CHAN_STATE_READ || error == EWOULDBLOCK) {
SUME_UNLOCK(adapter);
device_printf(adapter->dev, "wait error: %d\n", error);
return (EWOULDBLOCK);
}
bus_dmamap_sync(recv->ch_tag, recv->ch_map,
BUS_DMASYNC_POSTREAD | BUS_DMASYNC_POSTWRITE);
/*
* Read reply data and validate address and tag.
* Note: we do access the send side without lock but the state
* machine does prevent the data from changing.
*/
data = (struct nf_regop_data *) (recv->buf_addr +
sizeof(struct nf_bb_desc));
if (le32toh(data->rtag) != send->rtag)
device_printf(adapter->dev, "rtag error: 0x%08x 0x%08x\n",
le32toh(data->rtag), send->rtag);
sifr->val = le32toh(data->val);
recv->state = SUME_RIFFA_CHAN_STATE_IDLE;
/* We are done. */
send->state = SUME_RIFFA_CHAN_STATE_IDLE;
SUME_UNLOCK(adapter);
return (0);
}
/* Read value from a module register and return it to a sume_ifreq. */
static int
get_modreg_value(struct nf_priv *nf_priv, struct sume_ifreq *sifr)
{
int error;
error = sume_module_reg_write(nf_priv, sifr, SUME_MR_READ);
if (!error)
error = sume_module_reg_read(nf_priv, sifr);
return (error);
}
static int
sume_if_ioctl(struct ifnet *ifp, unsigned long cmd, caddr_t data)
{
struct ifreq *ifr = (struct ifreq *) data;
struct nf_priv *nf_priv = ifp->if_softc;
struct sume_ifreq sifr;
int error = 0;
switch (cmd) {
case SIOCGIFMEDIA:
case SIOCGIFXMEDIA:
error = ifmedia_ioctl(ifp, ifr, &nf_priv->media, cmd);
break;
case SUME_IOCTL_CMD_WRITE_REG:
error = copyin(ifr_data_get_ptr(ifr), &sifr, sizeof(sifr));
if (error) {
error = EINVAL;
break;
}
error = sume_module_reg_write(nf_priv, &sifr, SUME_MR_WRITE);
break;
case SUME_IOCTL_CMD_READ_REG:
error = copyin(ifr_data_get_ptr(ifr), &sifr, sizeof(sifr));
if (error) {
error = EINVAL;
break;
}
error = get_modreg_value(nf_priv, &sifr);
if (error)
break;
error = copyout(&sifr, ifr_data_get_ptr(ifr), sizeof(sifr));
if (error)
error = EINVAL;
break;
case SIOCSIFFLAGS:
/* Silence tcpdump 'promisc mode not supported' warning. */
if (ifp->if_flags & IFF_PROMISC)
break;
default:
error = ether_ioctl(ifp, cmd, data);
break;
}
return (error);
}
static int
sume_media_change(struct ifnet *ifp)
{
struct nf_priv *nf_priv = ifp->if_softc;
struct ifmedia *ifm = &nf_priv->media;
if (IFM_TYPE(ifm->ifm_media) != IFM_ETHER)
return (EINVAL);
if (IFM_SUBTYPE(ifm->ifm_media) == IFM_10G_SR)
ifp->if_baudrate = ifmedia_baudrate(IFM_ETHER | IFM_10G_SR);
else
ifp->if_baudrate = ifmedia_baudrate(ifm->ifm_media);
return (0);
}
static void
sume_update_link_status(struct ifnet *ifp)
{
struct nf_priv *nf_priv = ifp->if_softc;
struct sume_adapter *adapter = nf_priv->adapter;
struct sume_ifreq sifr;
int link_status;
sifr.addr = SUME_STATUS_ADDR(nf_priv->port);
sifr.val = 0;
if (get_modreg_value(nf_priv, &sifr))
return;
link_status = SUME_LINK_STATUS(sifr.val);
if (!link_status && nf_priv->link_up) {
if_link_state_change(ifp, LINK_STATE_DOWN);
nf_priv->link_up = 0;
if (adapter->sume_debug)
device_printf(adapter->dev, "port %d link state "
"changed to DOWN\n", nf_priv->unit);
} else if (link_status && !nf_priv->link_up) {
nf_priv->link_up = 1;
if_link_state_change(ifp, LINK_STATE_UP);
if (adapter->sume_debug)
device_printf(adapter->dev, "port %d link state "
"changed to UP\n", nf_priv->unit);
}
}
static void
sume_media_status(struct ifnet *ifp, struct ifmediareq *ifmr)
{
struct nf_priv *nf_priv = ifp->if_softc;
struct ifmedia *ifm = &nf_priv->media;
if (ifm->ifm_cur->ifm_media == (IFM_ETHER | IFM_10G_SR) &&
(ifp->if_flags & IFF_UP))
ifmr->ifm_active = IFM_ETHER | IFM_10G_SR;
else
ifmr->ifm_active = ifm->ifm_cur->ifm_media;
ifmr->ifm_status |= IFM_AVALID;