-
Notifications
You must be signed in to change notification settings - Fork 134
/
npu2.c
2318 lines (1965 loc) · 65.8 KB
/
npu2.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: Apache-2.0
/*
* NPU - NVlink and OpenCAPI
*
* Copyright 2013-2019 IBM Corp.
*/
#include <skiboot.h>
#include <io.h>
#include <timebase.h>
#include <pci-cfg.h>
#include <pci.h>
#include <pci-slot.h>
#include <pci-virt.h>
#include <opal.h>
#include <opal-api.h>
#include <cpu.h>
#include <device.h>
#include <ccan/str/str.h>
#include <ccan/array_size/array_size.h>
#include <affinity.h>
#include <npu2.h>
#include <lock.h>
#include <xscom.h>
#include <bitutils.h>
#include <chip.h>
#include <phys-map.h>
#include <nvram.h>
#include <xscom-p9-regs.h>
#include <phb4.h>
#include <cache-p9.h>
#define VENDOR_CAP_START 0x80
#define VENDOR_CAP_END 0x90
#define VENDOR_CAP_LEN 0x10
#define VENDOR_CAP_VERSION 0x01
#define VENDOR_CAP_PCI_DEV_OFFSET 0x0d
/*
* NPU2 BAR layout definition. We have 3 stacks and each of them
* contains 2 bricks. So every NPU2 has 6 bricks in total. There are 2
* PHY BARs and each of them is shared by 3 bricks. Every brick has
* one NTL BAR and two bricks share one GENID BAR. There is also a
* global MMIO BAR. We only expose DL and GENID BARs to the OS and all
* other BARs will be hidden in skiboot.
*
* Before the global MMIO BAR is configured, scom is the only way to
* access the BAR registers. At NPU2 PHB probing time, we rely on scom
* to assign all BARs until the global MMIO BAR is established.
*
* We need to access 4 SM registers in the same stack in order to
* configure one particular BAR.
*/
/* Set a specific flag in the vendor config space */
void npu2_set_link_flag(struct npu2_dev *ndev, uint8_t flag)
{
ndev->nvlink.link_flags |= flag;
PCI_VIRT_CFG_INIT_RO(ndev->nvlink.pvd, VENDOR_CAP_START +
VENDOR_CAP_PCI_DEV_OFFSET, 1, ndev->nvlink.link_flags);
}
void npu2_clear_link_flag(struct npu2_dev *ndev, uint8_t flag)
{
ndev->nvlink.link_flags &= ~flag;
PCI_VIRT_CFG_INIT_RO(ndev->nvlink.pvd, VENDOR_CAP_START +
VENDOR_CAP_PCI_DEV_OFFSET, 1, ndev->nvlink.link_flags);
}
static inline void npu2_ioda_sel(struct npu2 *p, uint32_t table,
uint32_t index, bool autoinc)
{
out_be64(p->regs + NPU2_ATS_IODA_TBL,
(autoinc ? NPU2_ATS_IODA_TBL_AUTOINC : 0ul) |
SETFIELD(NPU2_ATS_IODA_TBL_SELECT, 0ul, table) |
SETFIELD(NPU2_ATS_IODA_TBL_INDEX, 0ul, index));
}
static struct npu2_dev *npu2_bdf_to_dev(struct npu2 *p,
uint32_t bdfn)
{
struct pci_virt_device *pvd;
/* All emulated devices are attached to root bus */
if (bdfn & ~0xff)
return NULL;
pvd = pci_virt_find_device(&p->phb_nvlink, bdfn);
if (pvd)
return pvd->data;
return NULL;
}
static inline void npu2_get_bar(uint32_t gcid, struct npu2_bar *bar)
{
phys_map_get(gcid, bar->type, bar->index, &bar->base, &bar->size);
}
static void npu2_read_bar(struct npu2 *p, struct npu2_bar *bar)
{
uint64_t reg, val;
int enabled;
reg = NPU2_REG_OFFSET(0, NPU2_BLOCK_SM_0, bar->reg);
val = npu2_read(p, reg);
switch (NPU2_REG(bar->reg)) {
case NPU2_PHY_BAR:
bar->base = GETFIELD(NPU2_PHY_BAR_ADDR, val) << 21;
enabled = GETFIELD(NPU2_PHY_BAR_ENABLE, val);
if (NPU2_REG_STACK(reg) == NPU2_STACK_STCK_2)
/* This is the global MMIO BAR */
bar->size = 0x1000000;
else
bar->size = 0x200000;
break;
case NPU2_NTL0_BAR:
case NPU2_NTL1_BAR:
bar->base = GETFIELD(NPU2_NTL_BAR_ADDR, val) << 16;
enabled = GETFIELD(NPU2_NTL_BAR_ENABLE, val);
bar->size = 0x10000 << GETFIELD(NPU2_NTL_BAR_SIZE, val);
break;
case NPU2_GENID_BAR:
bar->base = GETFIELD(NPU2_GENID_BAR_ADDR, val) << 16;
enabled = GETFIELD(NPU2_GENID_BAR_ENABLE, val);
bar->size = 0x20000;
break;
default:
bar->base = 0ul;
enabled = 0;
bar->size = 0;
break;
}
bar->flags = SETFIELD(NPU2_BAR_FLAG_ENABLED, bar->flags, enabled);
}
static void npu2_write_bar(struct npu2 *p,
struct npu2_bar *bar,
uint32_t gcid,
uint32_t scom)
{
uint64_t reg, val, enable = !!(bar->flags & NPU2_BAR_FLAG_ENABLED);
int block;
switch (NPU2_REG(bar->reg)) {
case NPU2_PHY_BAR:
val = SETFIELD(NPU2_PHY_BAR_ADDR, 0ul, bar->base >> 21);
val = SETFIELD(NPU2_PHY_BAR_ENABLE, val, enable);
break;
case NPU2_NTL0_BAR:
case NPU2_NTL1_BAR:
val = SETFIELD(NPU2_NTL_BAR_ADDR, 0ul, bar->base >> 16);
val = SETFIELD(NPU2_NTL_BAR_ENABLE, val, enable);
val = SETFIELD(NPU2_NTL_BAR_SIZE, val, 1);
break;
case NPU2_GENID_BAR:
val = SETFIELD(NPU2_GENID_BAR_ADDR, 0ul, bar->base >> 16);
val = SETFIELD(NPU2_GENID_BAR_ENABLE, val, enable);
break;
default:
val = 0ul;
}
for (block = NPU2_BLOCK_SM_0; block <= NPU2_BLOCK_SM_3; block++) {
reg = NPU2_REG_OFFSET(0, block, bar->reg);
if (p)
npu2_write(p, reg, val);
else
npu2_scom_write(gcid, scom, reg, NPU2_MISC_DA_LEN_8B, val);
}
}
/* Trap for PCI command (0x4) to enable or disable device's BARs */
static int64_t npu2_cfg_write_cmd(void *dev,
struct pci_cfg_reg_filter *pcrf __unused,
uint32_t offset, uint32_t size,
uint32_t *data, bool write)
{
struct pci_virt_device *pvd = dev;
struct npu2_dev *ndev = pvd->data;
struct npu2_bar *ntl_npu_bar, *genid_npu_bar;
bool enabled;
if (!write)
return OPAL_PARTIAL;
if (offset != PCI_CFG_CMD)
return OPAL_PARAMETER;
if (size != 1 && size != 2 && size != 4)
return OPAL_PARAMETER;
/*
* Enable or disable NTL and GENID BAR. Two bricks share
* one GENID BAR, which is exposed via the first brick.
*/
enabled = !!(*data & PCI_CFG_CMD_MEM_EN);
ntl_npu_bar = &ndev->bars[0].npu2_bar;
genid_npu_bar = &ndev->bars[1].npu2_bar;
ntl_npu_bar->flags = SETFIELD(NPU2_BAR_FLAG_ENABLED, ntl_npu_bar->flags, enabled);
npu2_write_bar(ndev->npu, ntl_npu_bar, 0, 0);
/*
* Enable/disable the GENID BAR. Two bricks share one GENID
* BAR which is exposed via the first brick so we need to
* track the enables separately.
*/
if (NPU2DEV_BRICK(ndev))
genid_npu_bar->flags = SETFIELD(NPU2_BAR_FLAG_ENABLED1, genid_npu_bar->flags,
enabled);
else
genid_npu_bar->flags = SETFIELD(NPU2_BAR_FLAG_ENABLED0, genid_npu_bar->flags,
enabled);
/* Enable the BAR if either device requests it enabled, otherwise disable it */
genid_npu_bar->flags = SETFIELD(NPU2_BAR_FLAG_ENABLED, genid_npu_bar->flags,
!!(genid_npu_bar->flags & (NPU2_BAR_FLAG_ENABLED0 |
NPU2_BAR_FLAG_ENABLED1)));
npu2_write_bar(ndev->npu, genid_npu_bar, 0, 0);
return OPAL_PARTIAL;
}
static int64_t npu2_cfg_read_bar(struct npu2_dev *dev __unused,
struct pci_cfg_reg_filter *pcrf,
uint32_t offset, uint32_t size,
uint32_t *data)
{
struct npu2_pcie_bar *bar = (struct npu2_pcie_bar *) pcrf->data;
if (!(bar->flags & NPU2_PCIE_BAR_FLAG_TRAPPED))
return OPAL_PARTIAL;
if ((size != 4) ||
(offset != pcrf->start && offset != pcrf->start + 4))
return OPAL_PARAMETER;
if (bar->flags & NPU2_PCIE_BAR_FLAG_SIZE_HI)
*data = bar->npu2_bar.size >> 32;
else
*data = bar->npu2_bar.size;
bar->flags &= ~(NPU2_PCIE_BAR_FLAG_TRAPPED | NPU2_PCIE_BAR_FLAG_SIZE_HI);
return OPAL_SUCCESS;
}
static int64_t npu2_cfg_write_bar(struct npu2_dev *dev,
struct pci_cfg_reg_filter *pcrf,
uint32_t offset, uint32_t size,
uint32_t data)
{
struct npu2_pcie_bar *bar = (struct npu2_pcie_bar *) pcrf->data;
struct npu2_bar old_bar, *npu2_bar = &bar->npu2_bar;
if ((size != 4) ||
(offset != pcrf->start && offset != pcrf->start + 4))
return OPAL_PARAMETER;
/* Return BAR size on next read */
if (data == 0xffffffff) {
bar->flags |= NPU2_PCIE_BAR_FLAG_TRAPPED;
if (offset == pcrf->start + 4)
bar->flags |= NPU2_PCIE_BAR_FLAG_SIZE_HI;
return OPAL_SUCCESS;
}
if (offset == pcrf->start) {
npu2_bar->base &= 0xffffffff00000000UL;
npu2_bar->base |= (data & 0xfffffff0);
} else {
npu2_bar->base &= 0x00000000ffffffffUL;
npu2_bar->base |= ((uint64_t)data << 32);
if (NPU2_REG(npu2_bar->reg) == NPU2_GENID_BAR && NPU2DEV_BRICK(dev))
npu2_bar->base -= 0x10000;
old_bar.reg = npu2_bar->reg;
npu2_read_bar(dev->npu, &old_bar);
/* Only allow changing the base address if the BAR is not enabled */
if ((npu2_bar->flags & NPU2_BAR_FLAG_ENABLED) &&
(npu2_bar->base != old_bar.base)) {
npu2_bar->base = old_bar.base;
return OPAL_HARDWARE;
}
npu2_write_bar(dev->npu, &bar->npu2_bar, 0, 0);
}
/* To update the config cache */
return OPAL_PARTIAL;
}
static int64_t npu2_dev_cfg_bar(void *dev, struct pci_cfg_reg_filter *pcrf,
uint32_t offset, uint32_t len, uint32_t *data,
bool write)
{
struct pci_virt_device *pvd = dev;
struct npu2_dev *ndev = (struct npu2_dev *) pvd->data;
if (write)
return npu2_cfg_write_bar(ndev, pcrf, offset, len, *data);
return npu2_cfg_read_bar(ndev, pcrf, offset, len, data);
}
static int64_t npu2_dev_cfg_exp_devcap(void *dev,
struct pci_cfg_reg_filter *pcrf __unused,
uint32_t offset, uint32_t size,
uint32_t *data, bool write)
{
struct pci_virt_device *pvd = dev;
struct npu2_dev *ndev = pvd->data;
int rc;
assert(write);
if ((size != 2) || (offset & 1)) {
/* Short config writes are not supported */
prlog(PR_ERR, "NPU%d: Unsupported write to pcie control register\n",
ndev->nvlink.phb->opal_id);
return OPAL_PARAMETER;
}
if (*data & PCICAP_EXP_DEVCTL_FUNC_RESET)
npu2_dev_procedure_reset(ndev);
rc = purge_l2_l3_caches();
if (rc)
return rc;
return OPAL_PARTIAL;
}
#define NPU2_CFG_READ(size, type) \
static int64_t npu2_cfg_read##size(struct phb *phb, uint32_t bdfn, \
uint32_t offset, type *data) \
{ \
uint32_t val; \
int64_t ret; \
\
ret = pci_virt_cfg_read(phb, bdfn, offset, \
sizeof(*data), &val); \
*data = (type)val; \
return ret; \
}
#define NPU2_CFG_WRITE(size, type) \
static int64_t npu2_cfg_write##size(struct phb *phb, uint32_t bdfn, \
uint32_t offset, type data) \
{ \
uint32_t val = data; \
int64_t ret; \
\
ret = pci_virt_cfg_write(phb, bdfn, offset, \
sizeof(data), val); \
return ret; \
}
NPU2_CFG_READ(8, u8);
NPU2_CFG_READ(16, u16);
NPU2_CFG_READ(32, u32);
NPU2_CFG_WRITE(8, u8);
NPU2_CFG_WRITE(16, u16);
NPU2_CFG_WRITE(32, u32);
static int __npu2_dev_bind_pci_dev(struct phb *phb __unused,
struct pci_device *pd,
void *data)
{
struct npu2_dev *dev = data;
struct dt_node *pci_dt_node;
char *pcislot;
/* Ignore non-nvidia PCI devices */
if ((pd->vdid & 0xffff) != 0x10de)
return 0;
/* Find the PCI device's slot location */
for (pci_dt_node = pd->dn;
pci_dt_node && !dt_find_property(pci_dt_node, "ibm,loc-code");
pci_dt_node = pci_dt_node->parent);
if (!pci_dt_node)
return 0;
pcislot = (char *)dt_prop_get(pci_dt_node, "ibm,loc-code");
NPU2DEVDBG(dev, "Comparing GPU '%s' and NPU2 '%s'\n",
pcislot, dev->nvlink.slot_label);
if (streq(pcislot, dev->nvlink.slot_label))
return 1;
return 0;
}
static int64_t npu2_gpu_bridge_sec_bus_reset(void *dev,
struct pci_cfg_reg_filter *pcrf __unused,
uint32_t offset, uint32_t len,
uint32_t *data, bool write)
{
struct pci_device *pd = dev;
struct pci_device *gpu;
struct phb *npphb;
struct npu2 *npu;
struct dt_node *np;
struct npu2_dev *ndev;
int i;
assert(write);
if ((len != 2) || (offset & 1)) {
/* Short config writes are not supported */
PCIERR(pd->phb, pd->bdfn,
"Unsupported write to bridge control register\n");
return OPAL_PARAMETER;
}
gpu = list_top(&pd->children, struct pci_device, link);
if (gpu && (*data & PCI_CFG_BRCTL_SECONDARY_RESET)) {
int64_t rc;
dt_for_each_compatible(dt_root, np, "ibm,power9-npu-pciex") {
npphb = pci_get_phb(dt_prop_get_cell(np,
"ibm,opal-phbid", 1));
if (!npphb || npphb->phb_type != phb_type_npu_v2)
continue;
npu = phb_to_npu2_nvlink(npphb);
for (i = 0; i < npu->total_devices; ++i) {
ndev = &npu->devices[i];
if (ndev->nvlink.pd == gpu)
npu2_dev_procedure_reset(ndev);
}
}
rc = purge_l2_l3_caches();
if (rc)
return rc;
}
return OPAL_PARTIAL;
}
static void npu2_dev_bind_pci_dev(struct npu2_dev *dev)
{
struct phb *phb;
uint32_t i;
if (dev->nvlink.pd)
return;
for (i = 0; i < 64; i++) {
if (dev->npu->phb_nvlink.opal_id == i)
continue;
phb = pci_get_phb(i);
if (!phb)
continue;
dev->nvlink.pd = pci_walk_dev(phb, NULL, __npu2_dev_bind_pci_dev, dev);
if (dev->nvlink.pd) {
dev->nvlink.phb = phb;
/* Found the device, set the bit in config space */
npu2_set_link_flag(dev, NPU2_DEV_PCI_LINKED);
/*
* We define a custom sec bus reset handler for a slot
* with an NVLink-connected GPU to prevent HMIs which
* will otherwise happen if we reset GPU before
* resetting NVLinks.
*/
if (dev->nvlink.pd->parent &&
dev->nvlink.pd->parent->slot)
pci_add_cfg_reg_filter(dev->nvlink.pd->parent,
PCI_CFG_BRCTL, 2,
PCI_REG_FLAG_WRITE,
npu2_gpu_bridge_sec_bus_reset);
return;
}
}
NPU2DEVINF(dev, "No PCI device found for slot '%s'\n",
dev->nvlink.slot_label);
}
static struct lock pci_npu_phandle_lock = LOCK_UNLOCKED;
static void npu2_append_phandle(struct dt_node *dn,
u32 phandle)
{
struct dt_property *prop;
uint32_t *npu_phandles;
size_t len;
/*
* Use a lock to make sure no one else has a reference to an
* ibm,npu property (this assumes this is the only function
* that holds a reference to it)
*/
lock(&pci_npu_phandle_lock);
/* This function shouldn't be called unless ibm,npu exists */
prop = (struct dt_property *)dt_require_property(dn, "ibm,npu", -1);
/* Need to append to the properties */
len = prop->len + sizeof(*npu_phandles);
dt_resize_property(&prop, len);
prop->len = len;
npu_phandles = (uint32_t *)prop->prop;
npu_phandles[len / sizeof(*npu_phandles) - 1] = phandle;
unlock(&pci_npu_phandle_lock);
}
static struct dt_node *npu2_create_memory_dn(uint64_t addr, uint64_t size)
{
struct dt_node *mem;
static u32 chip_id = 255;
mem = dt_find_by_name_addr(dt_root, "memory", addr);
if (mem)
return mem;
mem = dt_new_addr(dt_root, "memory", addr);
if (!mem)
return NULL;
dt_add_property_string(mem, "device_type", "memory");
dt_add_property_string(mem, "compatible", "ibm,coherent-device-memory");
dt_add_property_u64s(mem, "reg", addr, size);
dt_add_property_cells(mem, "ibm,chip-id", chip_id);
dt_add_property_u64s(mem, "linux,usable-memory", addr, 0);
dt_add_property_cells(mem, "ibm,associativity", 4, chip_id, chip_id, chip_id, chip_id);
chip_id--;
assert(chip_id);
return mem;
}
/* There are potentially multiple links per GPU, so lookup the GPU memory based
* on bdfn. */
static void npu2_get_gpu_base(struct npu2_dev *ndev, uint64_t *addr, uint64_t *size)
{
struct npu2 *p = ndev->npu;
int group;
group = (ndev->bdfn >> 3) & 0x1f;
phys_map_get(ndev->npu->chip_id, p->gpu_map_type, group, addr, size);
}
static void npu2_dn_fixup_gmb(struct dt_node *pd_dn, struct npu2_dev *ndev)
{
uint64_t gpu_base, gpu_size, gta;
struct dt_node *mem_dn;
npu2_get_gpu_base(ndev, &gpu_base, &gpu_size);
mem_dn = npu2_create_memory_dn(gpu_base, gpu_size);
assert(mem_dn);
dt_add_property_cells(pd_dn, "memory-region", mem_dn->phandle);
/* Coral mode address compression. This is documented in Figure 3.5
* "P9->GPU RA Compression (Coral) of the NPU2 workbook". */
gta = ((gpu_base >> 42) & 0x1) << 42;
gta |= ((gpu_base >> 45) & 0x3) << 43;
gta |= ((gpu_base >> 49) & 0x3) << 45;
gta |= gpu_base & ((1UL << 43) - 1);
dt_add_property_u64s(pd_dn, "ibm,device-tgt-addr", gta);
}
static int npu2_assign_gmb(struct npu2_dev *ndev)
{
struct npu2 *p = ndev->npu;
int peers, mode;
uint32_t bdfn;
uint64_t base, size, reg, val, gmb;
/* Need to work out number of link peers. This amount to
* working out the maximum function number. So work start at
* the highest bdfn (fn = 6) and count back until we find a
* npu2_dev. */
for (bdfn = (ndev->bdfn & ~0x7) | NPU2_LINKS_PER_CHIP;
(bdfn & 0x7) != 0x7; bdfn = (bdfn & ~0x7) | ((bdfn & 0x7) - 1))
if (npu2_bdf_to_dev(p, bdfn))
break;
peers = bdfn & 0x7;
npu2_get_gpu_base(ndev, &base, &size);
NPU2DBG(p, "Setting BAR region dt:%llx\n", base);
val = SETFIELD(NPU2_MEM_BAR_EN, 0ULL, 1);
val = SETFIELD(NPU2_MEM_BAR_SEL_MEM, val, base >> (63-14));
val = SETFIELD(NPU2_MEM_BAR_GROUP, val, base >> (63-18));
val = SETFIELD(NPU2_MEM_BAR_CHIP, val, base >> (63-21));
val = SETFIELD(NPU2_MEM_BAR_NODE_ADDR, val, base >> (63-33));
val = SETFIELD(NPU2_MEM_BAR_POISON, val, 1);
val = SETFIELD(NPU2_MEM_BAR_GRANULE, val, 0);
/* We don't know how much memory the GPU has, so we may as well just
* pass the whole aperture through at this point. */
val = SETFIELD(NPU2_MEM_BAR_BAR_SIZE, val, ilog2(size >> 30));
switch (peers) {
case 0:
mode = 0;
break;
case 1:
mode = 1;
break;
case 2:
mode = 3;
break;
case 3:
mode = 6;
break;
case 5:
mode = 10;
break;
default:
/* Hardware does not support this configuration */
assert(0);
}
mode += ndev->bdfn & 0x7;
val = SETFIELD(NPU2_MEM_BAR_MODE, val, mode);
gmb = NPU2_GPU0_MEM_BAR;
if (NPU2DEV_BRICK(ndev))
gmb = NPU2_GPU1_MEM_BAR;
reg = NPU2_REG_OFFSET(NPU2_STACK_STCK_0 + NPU2DEV_STACK(ndev),
NPU2_BLOCK_SM_0, gmb);
npu2_write(p, reg, val);
reg = NPU2_REG_OFFSET(NPU2_STACK_STCK_0 + NPU2DEV_STACK(ndev),
NPU2_BLOCK_SM_1, gmb);
npu2_write(p, reg, val);
reg = NPU2_REG_OFFSET(NPU2_STACK_STCK_0 + NPU2DEV_STACK(ndev),
NPU2_BLOCK_SM_2, gmb);
npu2_write(p, reg, val);
reg = NPU2_REG_OFFSET(NPU2_STACK_STCK_0 + NPU2DEV_STACK(ndev),
NPU2_BLOCK_SM_3, gmb);
npu2_write(p, reg, val);
return 0;
}
static int npu2_dn_fixup(struct phb *phb,
struct pci_device *pd,
void *data __unused)
{
struct npu2 *p = phb_to_npu2_nvlink(phb);
struct npu2_dev *dev;
uint32_t speed;
const char *label;
dev = npu2_bdf_to_dev(p, pd->bdfn);
assert(dev);
if (dev->nvlink.phb || dev->nvlink.pd)
return 0;
npu2_assign_gmb(dev);
npu2_dn_fixup_gmb(pd->dn, dev);
dt_add_property_cells(pd->dn, "ibm,nvlink", dev->dt_node->phandle);
/*
* NVLink supports multiple speeds and device drivers need to know what
* speed has been set by firmware. Hostboot does the inits that set the
* link speed and tell us via HDAT and we need to copy that from the
* link node.
*/
speed = dt_prop_get_u32_def(dev->dt_node, "nvidia,link-speed", 0xff);
if (speed != 0xff)
dt_add_property_cells(pd->dn, "ibm,nvlink-speed", speed);
/*
* NPU2 devices have a slot label that indicates which GPU slot
* this NPU is connected to. Add a location code to the NVlink
* device node based on the slot label.
*/
label = dt_prop_get_def(dev->dt_node, "ibm,slot-label", NULL);
if (!label) {
/**
* @fwts-label NPUNoPHBSlotLabel
* @fwts-advice No GPU/NPU2 slot information was found.
* NVLink2 functionality will not work.
*/
prlog(PR_ERR, "NPU: Cannot find GPU slot information\n");
return 0;
}
dt_add_property_string(pd->dn, "ibm,loc-code", label);
dev->nvlink.slot_label = label;
/*
* Bind the emulated PCI device with the real one, which can't
* be done until the PCI devices are populated. Once the real
* PCI device is identified, we also need fix the device-tree
* for it
*/
npu2_dev_bind_pci_dev(dev);
if (dev->nvlink.phb && dev->nvlink.pd && dev->nvlink.pd->dn) {
if (dt_find_property(dev->nvlink.pd->dn, "ibm,npu"))
npu2_append_phandle(dev->nvlink.pd->dn, pd->dn->phandle);
else
dt_add_property_cells(dev->nvlink.pd->dn, "ibm,npu", pd->dn->phandle);
dt_add_property_cells(pd->dn, "ibm,gpu", dev->nvlink.pd->dn->phandle);
dev->nvlink.gpu_bdfn = dev->nvlink.pd->bdfn;
}
return 0;
}
static int npu2_links_per_gpu(struct phb *phb,
struct pci_device *pd,
void *data)
{
struct npu2 *p = phb_to_npu2_nvlink(phb);
struct npu2_dev *dev;
int *nlinks = (int *)data;
dev = npu2_bdf_to_dev(p, pd->bdfn);
assert(dev);
if (dev->nvlink.phb && dev->nvlink.pd && dev->nvlink.pd->dn) {
const struct dt_property *prop;
int n;
/* The link count is the number of phandles in "ibm,npu" */
prop = dt_find_property(dev->nvlink.pd->dn, "ibm,npu");
if (!prop)
return 0;
/* Count could vary by gpu, so find the max */
n = prop->len / sizeof(uint32_t);
if (n > *nlinks)
*nlinks = n;
}
return 0;
}
static void npu2_phb_fixup_scominit(struct dt_node *dn, int links_per_gpu)
{
uint32_t gcid = dt_get_chip_id(dn);
uint64_t val, mask;
/*
* MRBSP settings for 2- and 3-link GPU systems. These can improve
* GPU peer-to-peer fully ordered write performance.
*/
if (links_per_gpu == 3) {
val = PPC_BIT(30) | PPC_BIT(34) | PPC_BIT(36) | PPC_BIT(37) |
PPC_BIT(44) | PPC_BIT(45);
mask = PPC_BITMASK(28,39) | PPC_BITMASK(44,47);
} else if (links_per_gpu == 2) {
val = PPC_BIT(46) | PPC_BIT(47);
mask = PPC_BITMASK(44,47);
} else
return;
xscom_write_mask(gcid, 0x50110c0, val, mask);
xscom_write_mask(gcid, 0x50112c0, val, mask);
xscom_write_mask(gcid, 0x50114c0, val, mask);
}
static void npu2_phb_final_fixup(struct phb *phb)
{
int links_per_gpu = 0;
struct dt_node *np;
pci_walk_dev(phb, NULL, npu2_dn_fixup, NULL);
/*
* Now that the emulated devices are bound to the real ones, we can
* determine links_per_gpu and do some final init.
*/
pci_walk_dev(phb, NULL, npu2_links_per_gpu, &links_per_gpu);
dt_for_each_compatible(dt_root, np, "ibm,power9-npu")
npu2_phb_fixup_scominit(np, links_per_gpu);
}
static void npu2_init_ioda_cache(struct npu2 *p)
{
/* TVT */
memset(p->tve_cache, 0, sizeof(p->tve_cache));
}
static int64_t npu2_ioda_reset(struct phb *phb, bool purge)
{
struct npu2 *p = phb_to_npu2_nvlink(phb);
uint32_t i;
if (purge) {
NPU2DBG(p, "Purging all IODA tables...\n");
npu2_init_ioda_cache(p);
}
/* TVT */
npu2_ioda_sel(p, NPU2_ATS_IODA_TBL_TVT, 0, true);
for (i = 0; i < ARRAY_SIZE(p->tve_cache); i++)
out_be64(p->regs + NPU2_ATS_IODA_DATA, p->tve_cache[i]);
return OPAL_SUCCESS;
}
static void npu2_write_mcd(struct npu2 *p, uint64_t pcb_addr, uint64_t addr,
uint64_t size)
{
uint64_t val;
NPU2DBG(p, "Setting MCD addr:%llx\n", pcb_addr);
assert(is_pow2(size));
val = MCD_BANK_CN_VALID;
val = SETFIELD(MCD_BANK_CN_SIZE, val, (size >> 25) - 1);
val = SETFIELD(MCD_BANK_CN_ADDR, val, addr >> 25);
xscom_write(p->chip_id, pcb_addr, val);
}
static void npu2_mcd_init(struct npu2 *p)
{
int i;
uint64_t size, addr, gpu_min_addr, gpu_max_addr, total_size;
/* Init memory cache directory (MCD) registers. */
phys_map_get(p->chip_id, p->gpu_map_type, NPU2_LINKS_PER_CHIP - 1,
&gpu_min_addr, NULL);
phys_map_get(p->chip_id, p->gpu_map_type, 0, &gpu_max_addr, &size);
gpu_max_addr += size;
/* We assume GPU memory is contiguous from the first possible GPU to the
* last and that the size is the same so best to check that. */
for (i = 0; i < NPU2_LINKS_PER_CHIP; i++) {
uint64_t tmp;
phys_map_get(p->chip_id, p->gpu_map_type, i, &addr, &tmp);
assert((addr >= gpu_min_addr) && (addr + tmp <= gpu_max_addr));
assert(tmp == size);
}
/* We have two MCDs, so if neccessary we can split the region covered
* across both if total_size is not a power of two. */
total_size = gpu_max_addr - gpu_min_addr;
size = 1ull << ilog2(total_size);
/* Allocate the biggest chunk first as we assume gpu_max_addr has the
* highest alignment. */
addr = gpu_max_addr - size;
npu2_write_mcd(p, MCD0_BANK0_CN3, addr, size);
total_size -= size;
if (total_size) {
/* total_size was not a power of two, but the remainder should
* be if all GPUs were assigned the same size. */
assert(is_pow2(total_size));
size = 1ull << ilog2(total_size);
addr -= size;
assert(addr <= gpu_min_addr);
npu2_write_mcd(p, MCD1_BANK0_CN3, addr, size);
}
}
static void npu2_hw_init(struct npu2 *p)
{
uint64_t reg, val;
int s, b;
npu2_ioda_reset(&p->phb_nvlink, false);
/* Enable XTS retry mode */
val = npu2_read(p, NPU2_XTS_CFG);
npu2_write(p, NPU2_XTS_CFG, val | NPU2_XTS_CFG_MMIOSD | NPU2_XTS_CFG_TRY_ATR_RO);
val = npu2_read(p, NPU2_XTS_CFG2);
npu2_write(p, NPU2_XTS_CFG2, val | NPU2_XTS_CFG2_NO_FLUSH_ENA);
/*
* There are three different ways we configure the MCD and memory map.
* 1) Old way
* Skiboot configures the MCD and puts GPUs at 4TB and below
* 2) New way with MCD
* Hostboot configures the MCD and skiboot puts GPU at 4TB and above
* 3) New way without MCD
* No one configures the MCD and skiboot puts GPU at 4TB and below
*
* 1) Will go away evenutally as it's a configuration that can
* cause an xstop or data integrity problems. We are keeping
* it around to support existing hostboot. Print error
* message if used.
* 2) Is for smaller memory configurations and will be used
* initially for GPUs on Witherspoon. Supports only to
* 512GB of memory and 4 GPUs per socket.
* 3) Is for fully populated configurations of 4TB of memory
* and 6GPUs per socket. May have performance impacts.
*
* The different configurations can be detected via the following scoms:
* 1) 0x5011c0c bit 2 = 1, 0x5011c0a bits 42:48 = 0
* 2) 0x5011c0c bit 2 = 1, 0x5011c0a bits 42:48 = 7
* 3) 0x5011c0c bit 2 = 0, 0x5011c0a bits 42:48 = 0
*/
/* Get 0x05011c0c bit 2 = 1 */
xscom_read(p->chip_id, PB_CENT_HP_MODE_CURR, &val);
if ((val & PB_CFG_CHG_RATE_GP_MASTER) != 0) {
/* Get 0x05011c0a bits 42:48 */
xscom_read(p->chip_id, PB_CENT_MODE, &val);
if (GETFIELD(PB_CFG_CHIP_ADDR_EXTENSION_MASK_CENT, val) == 0) {
/* 1) */
NPU2DBG(p, "Using old memory map + MCD enabled in skiboot\n");
NPU2ERR(p, "!!! Old firmware detected. Update hostboot for new MCD mapping !!!\n");
p->gpu_map_type = GPU_MEM_4T_DOWN;
npu2_mcd_init(p);
} else if (GETFIELD(PB_CFG_CHIP_ADDR_EXTENSION_MASK_CENT, val) == 7) {
/* 2) */
NPU2DBG(p, "Using small memory map + MCD enabled\n");
p->gpu_map_type = GPU_MEM_4T_UP;
} else
NPU2ERR(p, "!!! Unsupported NPU2 configuration. "
"0x%llx!!!\n", val);
} else {
/* 3) */
NPU2DBG(p, "Using large memory map + MCD disabled\n");
p->gpu_map_type = GPU_MEM_4T_DOWN;
}
/* Static initialization of every relaxed-ordering cfg[2] register */
val = NPU2_RELAXED_ORDERING_CMD_CL_DMA_W |
NPU2_RELAXED_ORDERING_CMD_CL_DMA_W_HP |
NPU2_RELAXED_ORDERING_CMD_CL_DMA_INJ |
NPU2_RELAXED_ORDERING_CMD_PR_DMA_INJ |
NPU2_RELAXED_ORDERING_CMD_DMA_PR_W |
NPU2_RELAXED_ORDERING_CMD_CL_RD_NC_F0 |
NPU2_RELAXED_ORDERING_SOURCE4_RDENA;
for (s = NPU2_STACK_STCK_0; s <= NPU2_STACK_STCK_2; s++) {
for (b = NPU2_BLOCK_SM_0; b <= NPU2_BLOCK_SM_3; b++) {
reg = NPU2_REG_OFFSET(s, b, NPU2_RELAXED_ORDERING_CFG(2));
npu2_write(p, reg, val);
}
}
}
static int64_t npu2_map_pe_dma_window_real(struct phb *phb,
uint64_t pe_num,
uint16_t window_id,
uint64_t pci_start_addr __unused,
uint64_t pci_mem_size __unused)
{
struct npu2 *p = phb_to_npu2_nvlink(phb);
uint64_t tve;
/* Sanity check. Each PE has one corresponding TVE */
if (pe_num >= NPU2_MAX_PE_NUM ||
window_id != pe_num)
return OPAL_PARAMETER;
if (pci_mem_size) {
/* GPUs need to be able to access the MMIO memory space as well.
* On POWER9 this is above the top of ram so disable the TVT
* range check allowing access to all memory addresses. */
tve = 0;
} else {
/* Disable */
tve = PPC_BIT(51);
}
npu2_ioda_sel(p, NPU2_ATS_IODA_TBL_TVT, window_id, false);
out_be64(p->regs + NPU2_ATS_IODA_DATA, tve);
p->tve_cache[window_id] = tve;
return OPAL_SUCCESS;
}
static int64_t npu2_map_pe_dma_window(struct phb *phb,
uint64_t pe_num,
uint16_t window_id,
uint16_t tce_levels,
uint64_t tce_table_addr,
uint64_t tce_table_size,
uint64_t tce_page_size)
{
struct npu2 *p = phb_to_npu2_nvlink(phb);
uint64_t tts_encoded;
uint64_t data64 = 0;
/* Sanity check. Each PE has one corresponding TVE */
if (pe_num >= NPU2_MAX_PE_NUM ||
window_id != pe_num)
return OPAL_PARAMETER;
/*
* Special condition, zero TCE table size used to disable
* the TVE.
*/
if (!tce_table_size) {
npu2_ioda_sel(p, NPU2_ATS_IODA_TBL_TVT, window_id, false);