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
|
/*
* QEMU paravirtual RDMA - Command channel
*
* Copyright (C) 2018 Oracle
* Copyright (C) 2018 Red Hat Inc
*
* Authors:
* Yuval Shaia <yuval.shaia@oracle.com>
* Marcel Apfelbaum <marcel@redhat.com>
*
* This work is licensed under the terms of the GNU GPL, version 2 or later.
* See the COPYING file in the top-level directory.
*
*/
#include "qemu/osdep.h"
#include "cpu.h"
#include "hw/hw.h"
#include "hw/pci/pci.h"
#include "hw/pci/pci_ids.h"
#include "../rdma_backend.h"
#include "../rdma_rm.h"
#include "../rdma_utils.h"
#include "trace.h"
#include "pvrdma.h"
#include "standard-headers/rdma/vmw_pvrdma-abi.h"
static void *pvrdma_map_to_pdir(PCIDevice *pdev, uint64_t pdir_dma,
uint32_t nchunks, size_t length)
{
uint64_t *dir, *tbl;
int tbl_idx, dir_idx, addr_idx;
void *host_virt = NULL, *curr_page;
if (!nchunks) {
rdma_error_report("Got nchunks=0");
return NULL;
}
dir = rdma_pci_dma_map(pdev, pdir_dma, TARGET_PAGE_SIZE);
if (!dir) {
rdma_error_report("Failed to map to page directory");
return NULL;
}
tbl = rdma_pci_dma_map(pdev, dir[0], TARGET_PAGE_SIZE);
if (!tbl) {
rdma_error_report("Failed to map to page table 0");
goto out_unmap_dir;
}
curr_page = rdma_pci_dma_map(pdev, (dma_addr_t)tbl[0], TARGET_PAGE_SIZE);
if (!curr_page) {
rdma_error_report("Failed to map the page 0");
goto out_unmap_tbl;
}
host_virt = mremap(curr_page, 0, length, MREMAP_MAYMOVE);
if (host_virt == MAP_FAILED) {
host_virt = NULL;
rdma_error_report("Failed to remap memory for host_virt");
goto out_unmap_tbl;
}
trace_pvrdma_map_to_pdir_host_virt(curr_page, host_virt);
rdma_pci_dma_unmap(pdev, curr_page, TARGET_PAGE_SIZE);
dir_idx = 0;
tbl_idx = 1;
addr_idx = 1;
while (addr_idx < nchunks) {
if (tbl_idx == TARGET_PAGE_SIZE / sizeof(uint64_t)) {
tbl_idx = 0;
dir_idx++;
rdma_pci_dma_unmap(pdev, tbl, TARGET_PAGE_SIZE);
tbl = rdma_pci_dma_map(pdev, dir[dir_idx], TARGET_PAGE_SIZE);
if (!tbl) {
rdma_error_report("Failed to map to page table %d", dir_idx);
goto out_unmap_host_virt;
}
}
curr_page = rdma_pci_dma_map(pdev, (dma_addr_t)tbl[tbl_idx],
TARGET_PAGE_SIZE);
if (!curr_page) {
rdma_error_report("Failed to map to page %d, dir %d", tbl_idx,
dir_idx);
goto out_unmap_host_virt;
}
mremap(curr_page, 0, TARGET_PAGE_SIZE, MREMAP_MAYMOVE | MREMAP_FIXED,
host_virt + TARGET_PAGE_SIZE * addr_idx);
trace_pvrdma_map_to_pdir_next_page(addr_idx, curr_page, host_virt +
TARGET_PAGE_SIZE * addr_idx);
rdma_pci_dma_unmap(pdev, curr_page, TARGET_PAGE_SIZE);
addr_idx++;
tbl_idx++;
}
goto out_unmap_tbl;
out_unmap_host_virt:
munmap(host_virt, length);
host_virt = NULL;
out_unmap_tbl:
rdma_pci_dma_unmap(pdev, tbl, TARGET_PAGE_SIZE);
out_unmap_dir:
rdma_pci_dma_unmap(pdev, dir, TARGET_PAGE_SIZE);
return host_virt;
}
static int query_port(PVRDMADev *dev, union pvrdma_cmd_req *req,
union pvrdma_cmd_resp *rsp)
{
struct pvrdma_cmd_query_port *cmd = &req->query_port;
struct pvrdma_cmd_query_port_resp *resp = &rsp->query_port_resp;
struct pvrdma_port_attr attrs = {};
if (cmd->port_num > MAX_PORTS) {
return -EINVAL;
}
if (rdma_backend_query_port(&dev->backend_dev,
(struct ibv_port_attr *)&attrs)) {
return -ENOMEM;
}
memset(resp, 0, sizeof(*resp));
resp->attrs.state = dev->func0->device_active ? attrs.state :
PVRDMA_PORT_DOWN;
resp->attrs.max_mtu = attrs.max_mtu;
resp->attrs.active_mtu = attrs.active_mtu;
resp->attrs.phys_state = attrs.phys_state;
resp->attrs.gid_tbl_len = MIN(MAX_PORT_GIDS, attrs.gid_tbl_len);
resp->attrs.max_msg_sz = 1024;
resp->attrs.pkey_tbl_len = MIN(MAX_PORT_PKEYS, attrs.pkey_tbl_len);
resp->attrs.active_width = 1;
resp->attrs.active_speed = 1;
return 0;
}
static int query_pkey(PVRDMADev *dev, union pvrdma_cmd_req *req,
union pvrdma_cmd_resp *rsp)
{
struct pvrdma_cmd_query_pkey *cmd = &req->query_pkey;
struct pvrdma_cmd_query_pkey_resp *resp = &rsp->query_pkey_resp;
if (cmd->port_num > MAX_PORTS) {
return -EINVAL;
}
if (cmd->index > MAX_PKEYS) {
return -EINVAL;
}
memset(resp, 0, sizeof(*resp));
resp->pkey = PVRDMA_PKEY;
return 0;
}
static int create_pd(PVRDMADev *dev, union pvrdma_cmd_req *req,
union pvrdma_cmd_resp *rsp)
{
struct pvrdma_cmd_create_pd *cmd = &req->create_pd;
struct pvrdma_cmd_create_pd_resp *resp = &rsp->create_pd_resp;
int rc;
memset(resp, 0, sizeof(*resp));
rc = rdma_rm_alloc_pd(&dev->rdma_dev_res, &dev->backend_dev,
&resp->pd_handle, cmd->ctx_handle);
return rc;
}
static int destroy_pd(PVRDMADev *dev, union pvrdma_cmd_req *req,
union pvrdma_cmd_resp *rsp)
{
struct pvrdma_cmd_destroy_pd *cmd = &req->destroy_pd;
rdma_rm_dealloc_pd(&dev->rdma_dev_res, cmd->pd_handle);
return 0;
}
static int create_mr(PVRDMADev *dev, union pvrdma_cmd_req *req,
union pvrdma_cmd_resp *rsp)
{
struct pvrdma_cmd_create_mr *cmd = &req->create_mr;
struct pvrdma_cmd_create_mr_resp *resp = &rsp->create_mr_resp;
PCIDevice *pci_dev = PCI_DEVICE(dev);
void *host_virt = NULL;
int rc = 0;
memset(resp, 0, sizeof(*resp));
if (!(cmd->flags & PVRDMA_MR_FLAG_DMA)) {
host_virt = pvrdma_map_to_pdir(pci_dev, cmd->pdir_dma, cmd->nchunks,
cmd->length);
if (!host_virt) {
rdma_error_report("Failed to map to pdir");
return -EINVAL;
}
}
rc = rdma_rm_alloc_mr(&dev->rdma_dev_res, cmd->pd_handle, cmd->start,
cmd->length, host_virt, cmd->access_flags,
&resp->mr_handle, &resp->lkey, &resp->rkey);
if (rc && host_virt) {
munmap(host_virt, cmd->length);
}
return rc;
}
static int destroy_mr(PVRDMADev *dev, union pvrdma_cmd_req *req,
union pvrdma_cmd_resp *rsp)
{
struct pvrdma_cmd_destroy_mr *cmd = &req->destroy_mr;
rdma_rm_dealloc_mr(&dev->rdma_dev_res, cmd->mr_handle);
return 0;
}
static int create_cq_ring(PCIDevice *pci_dev , PvrdmaRing **ring,
uint64_t pdir_dma, uint32_t nchunks, uint32_t cqe)
{
uint64_t *dir = NULL, *tbl = NULL;
PvrdmaRing *r;
int rc = -EINVAL;
char ring_name[MAX_RING_NAME_SZ];
if (!nchunks || nchunks > PVRDMA_MAX_FAST_REG_PAGES) {
rdma_error_report("Got invalid nchunks: %d", nchunks);
return rc;
}
dir = rdma_pci_dma_map(pci_dev, pdir_dma, TARGET_PAGE_SIZE);
if (!dir) {
rdma_error_report("Failed to map to CQ page directory");
goto out;
}
tbl = rdma_pci_dma_map(pci_dev, dir[0], TARGET_PAGE_SIZE);
if (!tbl) {
rdma_error_report("Failed to map to CQ page table");
goto out;
}
r = g_malloc(sizeof(*r));
*ring = r;
r->ring_state = (struct pvrdma_ring *)
rdma_pci_dma_map(pci_dev, tbl[0], TARGET_PAGE_SIZE);
if (!r->ring_state) {
rdma_error_report("Failed to map to CQ ring state");
goto out_free_ring;
}
sprintf(ring_name, "cq_ring_%" PRIx64, pdir_dma);
rc = pvrdma_ring_init(r, ring_name, pci_dev, &r->ring_state[1],
cqe, sizeof(struct pvrdma_cqe),
/* first page is ring state */
(dma_addr_t *)&tbl[1], nchunks - 1);
if (rc) {
goto out_unmap_ring_state;
}
goto out;
out_unmap_ring_state:
/* ring_state was in slot 1, not 0 so need to jump back */
rdma_pci_dma_unmap(pci_dev, --r->ring_state, TARGET_PAGE_SIZE);
out_free_ring:
g_free(r);
out:
rdma_pci_dma_unmap(pci_dev, tbl, TARGET_PAGE_SIZE);
rdma_pci_dma_unmap(pci_dev, dir, TARGET_PAGE_SIZE);
return rc;
}
static void destroy_cq_ring(PvrdmaRing *ring)
{
pvrdma_ring_free(ring);
/* ring_state was in slot 1, not 0 so need to jump back */
rdma_pci_dma_unmap(ring->dev, --ring->ring_state, TARGET_PAGE_SIZE);
g_free(ring);
}
static int create_cq(PVRDMADev *dev, union pvrdma_cmd_req *req,
union pvrdma_cmd_resp *rsp)
{
struct pvrdma_cmd_create_cq *cmd = &req->create_cq;
struct pvrdma_cmd_create_cq_resp *resp = &rsp->create_cq_resp;
PvrdmaRing *ring = NULL;
int rc;
memset(resp, 0, sizeof(*resp));
resp->cqe = cmd->cqe;
rc = create_cq_ring(PCI_DEVICE(dev), &ring, cmd->pdir_dma, cmd->nchunks,
cmd->cqe);
if (rc) {
return rc;
}
rc = rdma_rm_alloc_cq(&dev->rdma_dev_res, &dev->backend_dev, cmd->cqe,
&resp->cq_handle, ring);
if (rc) {
destroy_cq_ring(ring);
}
resp->cqe = cmd->cqe;
return rc;
}
static int destroy_cq(PVRDMADev *dev, union pvrdma_cmd_req *req,
union pvrdma_cmd_resp *rsp)
{
struct pvrdma_cmd_destroy_cq *cmd = &req->destroy_cq;
RdmaRmCQ *cq;
PvrdmaRing *ring;
cq = rdma_rm_get_cq(&dev->rdma_dev_res, cmd->cq_handle);
if (!cq) {
rdma_error_report("Got invalid CQ handle");
return -EINVAL;
}
ring = (PvrdmaRing *)cq->opaque;
destroy_cq_ring(ring);
rdma_rm_dealloc_cq(&dev->rdma_dev_res, cmd->cq_handle);
return 0;
}
static int create_qp_rings(PCIDevice *pci_dev, uint64_t pdir_dma,
PvrdmaRing **rings, uint32_t scqe, uint32_t smax_sge,
uint32_t spages, uint32_t rcqe, uint32_t rmax_sge,
uint32_t rpages, uint8_t is_srq)
{
uint64_t *dir = NULL, *tbl = NULL;
PvrdmaRing *sr, *rr;
int rc = -EINVAL;
char ring_name[MAX_RING_NAME_SZ];
uint32_t wqe_sz;
if (!spages || spages > PVRDMA_MAX_FAST_REG_PAGES) {
rdma_error_report("Got invalid send page count for QP ring: %d",
spages);
return rc;
}
if (!is_srq && (!rpages || rpages > PVRDMA_MAX_FAST_REG_PAGES)) {
rdma_error_report("Got invalid recv page count for QP ring: %d",
rpages);
return rc;
}
dir = rdma_pci_dma_map(pci_dev, pdir_dma, TARGET_PAGE_SIZE);
if (!dir) {
rdma_error_report("Failed to map to QP page directory");
goto out;
}
tbl = rdma_pci_dma_map(pci_dev, dir[0], TARGET_PAGE_SIZE);
if (!tbl) {
rdma_error_report("Failed to map to QP page table");
goto out;
}
if (!is_srq) {
sr = g_malloc(2 * sizeof(*rr));
rr = &sr[1];
} else {
sr = g_malloc(sizeof(*sr));
}
*rings = sr;
/* Create send ring */
sr->ring_state = (struct pvrdma_ring *)
rdma_pci_dma_map(pci_dev, tbl[0], TARGET_PAGE_SIZE);
if (!sr->ring_state) {
rdma_error_report("Failed to map to QP ring state");
goto out_free_sr_mem;
}
wqe_sz = pow2ceil(sizeof(struct pvrdma_sq_wqe_hdr) +
sizeof(struct pvrdma_sge) * smax_sge - 1);
sprintf(ring_name, "qp_sring_%" PRIx64, pdir_dma);
rc = pvrdma_ring_init(sr, ring_name, pci_dev, sr->ring_state,
scqe, wqe_sz, (dma_addr_t *)&tbl[1], spages);
if (rc) {
goto out_unmap_ring_state;
}
if (!is_srq) {
/* Create recv ring */
rr->ring_state = &sr->ring_state[1];
wqe_sz = pow2ceil(sizeof(struct pvrdma_rq_wqe_hdr) +
sizeof(struct pvrdma_sge) * rmax_sge - 1);
sprintf(ring_name, "qp_rring_%" PRIx64, pdir_dma);
rc = pvrdma_ring_init(rr, ring_name, pci_dev, rr->ring_state,
rcqe, wqe_sz, (dma_addr_t *)&tbl[1 + spages],
rpages);
if (rc) {
goto out_free_sr;
}
}
goto out;
out_free_sr:
pvrdma_ring_free(sr);
out_unmap_ring_state:
rdma_pci_dma_unmap(pci_dev, sr->ring_state, TARGET_PAGE_SIZE);
out_free_sr_mem:
g_free(sr);
out:
rdma_pci_dma_unmap(pci_dev, tbl, TARGET_PAGE_SIZE);
rdma_pci_dma_unmap(pci_dev, dir, TARGET_PAGE_SIZE);
return rc;
}
static void destroy_qp_rings(PvrdmaRing *ring, uint8_t is_srq)
{
pvrdma_ring_free(&ring[0]);
if (!is_srq) {
pvrdma_ring_free(&ring[1]);
}
rdma_pci_dma_unmap(ring->dev, ring->ring_state, TARGET_PAGE_SIZE);
g_free(ring);
}
static int create_qp(PVRDMADev *dev, union pvrdma_cmd_req *req,
union pvrdma_cmd_resp *rsp)
{
struct pvrdma_cmd_create_qp *cmd = &req->create_qp;
struct pvrdma_cmd_create_qp_resp *resp = &rsp->create_qp_resp;
PvrdmaRing *rings = NULL;
int rc;
memset(resp, 0, sizeof(*resp));
rc = create_qp_rings(PCI_DEVICE(dev), cmd->pdir_dma, &rings,
cmd->max_send_wr, cmd->max_send_sge, cmd->send_chunks,
cmd->max_recv_wr, cmd->max_recv_sge,
cmd->total_chunks - cmd->send_chunks - 1, cmd->is_srq);
if (rc) {
return rc;
}
rc = rdma_rm_alloc_qp(&dev->rdma_dev_res, cmd->pd_handle, cmd->qp_type,
cmd->max_send_wr, cmd->max_send_sge,
cmd->send_cq_handle, cmd->max_recv_wr,
cmd->max_recv_sge, cmd->recv_cq_handle, rings,
&resp->qpn, cmd->is_srq, cmd->srq_handle);
if (rc) {
destroy_qp_rings(rings, cmd->is_srq);
return rc;
}
resp->max_send_wr = cmd->max_send_wr;
resp->max_recv_wr = cmd->max_recv_wr;
resp->max_send_sge = cmd->max_send_sge;
resp->max_recv_sge = cmd->max_recv_sge;
resp->max_inline_data = cmd->max_inline_data;
return 0;
}
static int modify_qp(PVRDMADev *dev, union pvrdma_cmd_req *req,
union pvrdma_cmd_resp *rsp)
{
struct pvrdma_cmd_modify_qp *cmd = &req->modify_qp;
int rc;
/* No need to verify sgid_index since it is u8 */
rc = rdma_rm_modify_qp(&dev->rdma_dev_res, &dev->backend_dev,
cmd->qp_handle, cmd->attr_mask,
cmd->attrs.ah_attr.grh.sgid_index,
(union ibv_gid *)&cmd->attrs.ah_attr.grh.dgid,
cmd->attrs.dest_qp_num,
(enum ibv_qp_state)cmd->attrs.qp_state,
cmd->attrs.qkey, cmd->attrs.rq_psn,
cmd->attrs.sq_psn);
return rc;
}
static int query_qp(PVRDMADev *dev, union pvrdma_cmd_req *req,
union pvrdma_cmd_resp *rsp)
{
struct pvrdma_cmd_query_qp *cmd = &req->query_qp;
struct pvrdma_cmd_query_qp_resp *resp = &rsp->query_qp_resp;
struct ibv_qp_init_attr init_attr;
int rc;
memset(resp, 0, sizeof(*resp));
rc = rdma_rm_query_qp(&dev->rdma_dev_res, &dev->backend_dev, cmd->qp_handle,
(struct ibv_qp_attr *)&resp->attrs, cmd->attr_mask,
&init_attr);
return rc;
}
static int destroy_qp(PVRDMADev *dev, union pvrdma_cmd_req *req,
union pvrdma_cmd_resp *rsp)
{
struct pvrdma_cmd_destroy_qp *cmd = &req->destroy_qp;
RdmaRmQP *qp;
PvrdmaRing *ring;
qp = rdma_rm_get_qp(&dev->rdma_dev_res, cmd->qp_handle);
if (!qp) {
return -EINVAL;
}
ring = (PvrdmaRing *)qp->opaque;
destroy_qp_rings(ring, qp->is_srq);
rdma_rm_dealloc_qp(&dev->rdma_dev_res, cmd->qp_handle);
return 0;
}
static int create_bind(PVRDMADev *dev, union pvrdma_cmd_req *req,
union pvrdma_cmd_resp *rsp)
{
struct pvrdma_cmd_create_bind *cmd = &req->create_bind;
int rc;
union ibv_gid *gid = (union ibv_gid *)&cmd->new_gid;
if (cmd->index >= MAX_PORT_GIDS) {
return -EINVAL;
}
rc = rdma_rm_add_gid(&dev->rdma_dev_res, &dev->backend_dev,
dev->backend_eth_device_name, gid, cmd->index);
return rc;
}
static int destroy_bind(PVRDMADev *dev, union pvrdma_cmd_req *req,
union pvrdma_cmd_resp *rsp)
{
int rc;
struct pvrdma_cmd_destroy_bind *cmd = &req->destroy_bind;
if (cmd->index >= MAX_PORT_GIDS) {
return -EINVAL;
}
rc = rdma_rm_del_gid(&dev->rdma_dev_res, &dev->backend_dev,
dev->backend_eth_device_name, cmd->index);
return rc;
}
static int create_uc(PVRDMADev *dev, union pvrdma_cmd_req *req,
union pvrdma_cmd_resp *rsp)
{
struct pvrdma_cmd_create_uc *cmd = &req->create_uc;
struct pvrdma_cmd_create_uc_resp *resp = &rsp->create_uc_resp;
int rc;
memset(resp, 0, sizeof(*resp));
rc = rdma_rm_alloc_uc(&dev->rdma_dev_res, cmd->pfn, &resp->ctx_handle);
return rc;
}
static int destroy_uc(PVRDMADev *dev, union pvrdma_cmd_req *req,
union pvrdma_cmd_resp *rsp)
{
struct pvrdma_cmd_destroy_uc *cmd = &req->destroy_uc;
rdma_rm_dealloc_uc(&dev->rdma_dev_res, cmd->ctx_handle);
return 0;
}
static int create_srq_ring(PCIDevice *pci_dev, PvrdmaRing **ring,
uint64_t pdir_dma, uint32_t max_wr,
uint32_t max_sge, uint32_t nchunks)
{
uint64_t *dir = NULL, *tbl = NULL;
PvrdmaRing *r;
int rc = -EINVAL;
char ring_name[MAX_RING_NAME_SZ];
uint32_t wqe_sz;
if (!nchunks || nchunks > PVRDMA_MAX_FAST_REG_PAGES) {
rdma_error_report("Got invalid page count for SRQ ring: %d",
nchunks);
return rc;
}
dir = rdma_pci_dma_map(pci_dev, pdir_dma, TARGET_PAGE_SIZE);
if (!dir) {
rdma_error_report("Failed to map to SRQ page directory");
goto out;
}
tbl = rdma_pci_dma_map(pci_dev, dir[0], TARGET_PAGE_SIZE);
if (!tbl) {
rdma_error_report("Failed to map to SRQ page table");
goto out;
}
r = g_malloc(sizeof(*r));
*ring = r;
r->ring_state = (struct pvrdma_ring *)
rdma_pci_dma_map(pci_dev, tbl[0], TARGET_PAGE_SIZE);
if (!r->ring_state) {
rdma_error_report("Failed to map tp SRQ ring state");
goto out_free_ring_mem;
}
wqe_sz = pow2ceil(sizeof(struct pvrdma_rq_wqe_hdr) +
sizeof(struct pvrdma_sge) * max_sge - 1);
sprintf(ring_name, "srq_ring_%" PRIx64, pdir_dma);
rc = pvrdma_ring_init(r, ring_name, pci_dev, &r->ring_state[1], max_wr,
wqe_sz, (dma_addr_t *)&tbl[1], nchunks - 1);
if (rc) {
goto out_unmap_ring_state;
}
goto out;
out_unmap_ring_state:
rdma_pci_dma_unmap(pci_dev, r->ring_state, TARGET_PAGE_SIZE);
out_free_ring_mem:
g_free(r);
out:
rdma_pci_dma_unmap(pci_dev, tbl, TARGET_PAGE_SIZE);
rdma_pci_dma_unmap(pci_dev, dir, TARGET_PAGE_SIZE);
return rc;
}
static void destroy_srq_ring(PvrdmaRing *ring)
{
pvrdma_ring_free(ring);
rdma_pci_dma_unmap(ring->dev, ring->ring_state, TARGET_PAGE_SIZE);
g_free(ring);
}
static int create_srq(PVRDMADev *dev, union pvrdma_cmd_req *req,
union pvrdma_cmd_resp *rsp)
{
struct pvrdma_cmd_create_srq *cmd = &req->create_srq;
struct pvrdma_cmd_create_srq_resp *resp = &rsp->create_srq_resp;
PvrdmaRing *ring = NULL;
int rc;
memset(resp, 0, sizeof(*resp));
rc = create_srq_ring(PCI_DEVICE(dev), &ring, cmd->pdir_dma,
cmd->attrs.max_wr, cmd->attrs.max_sge,
cmd->nchunks);
if (rc) {
return rc;
}
rc = rdma_rm_alloc_srq(&dev->rdma_dev_res, cmd->pd_handle,
cmd->attrs.max_wr, cmd->attrs.max_sge,
cmd->attrs.srq_limit, &resp->srqn, ring);
if (rc) {
destroy_srq_ring(ring);
return rc;
}
return 0;
}
static int query_srq(PVRDMADev *dev, union pvrdma_cmd_req *req,
union pvrdma_cmd_resp *rsp)
{
struct pvrdma_cmd_query_srq *cmd = &req->query_srq;
struct pvrdma_cmd_query_srq_resp *resp = &rsp->query_srq_resp;
memset(resp, 0, sizeof(*resp));
return rdma_rm_query_srq(&dev->rdma_dev_res, cmd->srq_handle,
(struct ibv_srq_attr *)&resp->attrs);
}
static int modify_srq(PVRDMADev *dev, union pvrdma_cmd_req *req,
union pvrdma_cmd_resp *rsp)
{
struct pvrdma_cmd_modify_srq *cmd = &req->modify_srq;
/* Only support SRQ limit */
if (!(cmd->attr_mask & IBV_SRQ_LIMIT) ||
(cmd->attr_mask & IBV_SRQ_MAX_WR))
return -EINVAL;
return rdma_rm_modify_srq(&dev->rdma_dev_res, cmd->srq_handle,
(struct ibv_srq_attr *)&cmd->attrs,
cmd->attr_mask);
}
static int destroy_srq(PVRDMADev *dev, union pvrdma_cmd_req *req,
union pvrdma_cmd_resp *rsp)
{
struct pvrdma_cmd_destroy_srq *cmd = &req->destroy_srq;
RdmaRmSRQ *srq;
PvrdmaRing *ring;
srq = rdma_rm_get_srq(&dev->rdma_dev_res, cmd->srq_handle);
if (!srq) {
return -EINVAL;
}
ring = (PvrdmaRing *)srq->opaque;
destroy_srq_ring(ring);
rdma_rm_dealloc_srq(&dev->rdma_dev_res, cmd->srq_handle);
return 0;
}
struct cmd_handler {
uint32_t cmd;
uint32_t ack;
int (*exec)(PVRDMADev *dev, union pvrdma_cmd_req *req,
union pvrdma_cmd_resp *rsp);
};
static struct cmd_handler cmd_handlers[] = {
{PVRDMA_CMD_QUERY_PORT, PVRDMA_CMD_QUERY_PORT_RESP, query_port},
{PVRDMA_CMD_QUERY_PKEY, PVRDMA_CMD_QUERY_PKEY_RESP, query_pkey},
{PVRDMA_CMD_CREATE_PD, PVRDMA_CMD_CREATE_PD_RESP, create_pd},
{PVRDMA_CMD_DESTROY_PD, PVRDMA_CMD_DESTROY_PD_RESP_NOOP, destroy_pd},
{PVRDMA_CMD_CREATE_MR, PVRDMA_CMD_CREATE_MR_RESP, create_mr},
{PVRDMA_CMD_DESTROY_MR, PVRDMA_CMD_DESTROY_MR_RESP_NOOP, destroy_mr},
{PVRDMA_CMD_CREATE_CQ, PVRDMA_CMD_CREATE_CQ_RESP, create_cq},
{PVRDMA_CMD_RESIZE_CQ, PVRDMA_CMD_RESIZE_CQ_RESP, NULL},
{PVRDMA_CMD_DESTROY_CQ, PVRDMA_CMD_DESTROY_CQ_RESP_NOOP, destroy_cq},
{PVRDMA_CMD_CREATE_QP, PVRDMA_CMD_CREATE_QP_RESP, create_qp},
{PVRDMA_CMD_MODIFY_QP, PVRDMA_CMD_MODIFY_QP_RESP, modify_qp},
{PVRDMA_CMD_QUERY_QP, PVRDMA_CMD_QUERY_QP_RESP, query_qp},
{PVRDMA_CMD_DESTROY_QP, PVRDMA_CMD_DESTROY_QP_RESP, destroy_qp},
{PVRDMA_CMD_CREATE_UC, PVRDMA_CMD_CREATE_UC_RESP, create_uc},
{PVRDMA_CMD_DESTROY_UC, PVRDMA_CMD_DESTROY_UC_RESP_NOOP, destroy_uc},
{PVRDMA_CMD_CREATE_BIND, PVRDMA_CMD_CREATE_BIND_RESP_NOOP, create_bind},
{PVRDMA_CMD_DESTROY_BIND, PVRDMA_CMD_DESTROY_BIND_RESP_NOOP, destroy_bind},
{PVRDMA_CMD_CREATE_SRQ, PVRDMA_CMD_CREATE_SRQ_RESP, create_srq},
{PVRDMA_CMD_QUERY_SRQ, PVRDMA_CMD_QUERY_SRQ_RESP, query_srq},
{PVRDMA_CMD_MODIFY_SRQ, PVRDMA_CMD_MODIFY_SRQ_RESP, modify_srq},
{PVRDMA_CMD_DESTROY_SRQ, PVRDMA_CMD_DESTROY_SRQ_RESP, destroy_srq},
};
int pvrdma_exec_cmd(PVRDMADev *dev)
{
int err = 0xFFFF;
DSRInfo *dsr_info;
dsr_info = &dev->dsr_info;
if (dsr_info->req->hdr.cmd >= sizeof(cmd_handlers) /
sizeof(struct cmd_handler)) {
rdma_error_report("Unsupported command");
goto out;
}
if (!cmd_handlers[dsr_info->req->hdr.cmd].exec) {
rdma_error_report("Unsupported command (not implemented yet)");
goto out;
}
err = cmd_handlers[dsr_info->req->hdr.cmd].exec(dev, dsr_info->req,
dsr_info->rsp);
dsr_info->rsp->hdr.response = dsr_info->req->hdr.response;
dsr_info->rsp->hdr.ack = cmd_handlers[dsr_info->req->hdr.cmd].ack;
dsr_info->rsp->hdr.err = err < 0 ? -err : 0;
trace_pvrdma_exec_cmd(dsr_info->req->hdr.cmd, dsr_info->rsp->hdr.err);
dev->stats.commands++;
out:
set_reg_val(dev, PVRDMA_REG_ERR, err);
post_interrupt(dev, INTR_VEC_CMD_RING);
return (err == 0) ? 0 : -EINVAL;
}
|