Repository navigation
Expand file tree
/
Copy pathcore_common.c
More file actions
720 lines (605 loc) · 18.9 KB
/
Copy pathcore_common.c
File metadata and controls
720 lines (605 loc) · 18.9 KB
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
// SPDX-License-Identifier: <SPDX License Expression>
#include "mx_dma.h"
#ifndef MX_DMA_DISABLE_TRACE
#include "trace.h"
#else
#define trace_mx_dma_xfer_submit(xfer_id, no_completion) do { } while (0)
#define trace_mx_dma_xfer_post_submit(dev_id, pushed_count) do { } while (0)
#define trace_mx_dma_xfer_complete(xfer_id, status, result, is_zombie) do { } while (0)
#define trace_mx_dma_xfer_complete_orphan(xfer_id, status, result) do { } while (0)
#endif
int mx_get_list_count(size_t total_desc_cnt, int descs_per_list)
{
int list_cnt = 1;
while (total_desc_cnt > descs_per_list) {
total_desc_cnt -= (descs_per_list - 1);
list_cnt++;
}
return list_cnt;
}
int mx_sg_locate(struct sg_table *sgt, size_t byte_offset,
struct scatterlist **out_sg, size_t *out_intra)
{
struct scatterlist *sg;
size_t acc = 0;
int i;
for_each_sgtable_dma_sg(sgt, sg, i) {
size_t dlen = sg_dma_len(sg);
if (acc + dlen > byte_offset) {
*out_sg = sg;
*out_intra = byte_offset - acc;
return 0;
}
acc += dlen;
}
*out_sg = NULL;
*out_intra = 0;
return -EINVAL;
}
static size_t prp_chunk_len_at(dma_addr_t dma_addr, size_t dma_size)
{
size_t rem = (size_t)(dma_addr & (dma_size - 1));
return rem ? (dma_size - rem) : dma_size;
}
size_t mx_prp_first_chunk_len(struct scatterlist *sg, size_t intra_off, size_t dma_size)
{
size_t len = prp_chunk_len_at(sg_dma_address(sg) + intra_off, dma_size);
return min_t(size_t, len, sg_dma_len(sg) - intra_off);
}
int mx_get_total_desc_count(struct scatterlist *sg, size_t intra_off, size_t byte_size,
size_t dma_size, size_t *out_cnt)
{
size_t remaining = byte_size;
size_t total = 0;
dma_addr_t end;
*out_cnt = 0;
if (byte_size == 0)
return -EINVAL;
while (remaining > 0 && sg) {
size_t avail = sg_dma_len(sg) - intra_off;
size_t consumed = min(avail, remaining);
size_t first_len = mx_prp_first_chunk_len(sg, intra_off, dma_size);
first_len = min(first_len, consumed);
total += 1;
if (consumed > first_len)
total += DIV_ROUND_UP(consumed - first_len, dma_size);
remaining -= consumed;
if (remaining == 0)
break;
end = sg_dma_address(sg) + intra_off + consumed;
if (end & (dma_size - 1)) {
pr_warn_ratelimited("sg entry ends off a %zu-byte chunk boundary (end=%pad)\n",
dma_size, &end);
return -EINVAL;
}
sg = sg_next(sg);
intra_off = 0;
if (sg && (sg_dma_address(sg) & (dma_size - 1))) {
pr_warn_ratelimited("sg entry starts off a %zu-byte chunk boundary (dma=%pad)\n",
dma_size, &sg->dma_address);
return -EINVAL;
}
}
if (remaining) {
pr_warn_ratelimited("sg mapping short by %zu bytes\n", remaining);
return -EINVAL;
}
*out_cnt = total;
return 0;
}
int mx_desc_list_init(struct mx_pci_dev *mx_pdev,
struct mx_transfer *transfer, size_t dma_size,
int descs_per_list, bool skip_first_entry,
size_t desc_cnt, uint64_t *out_ba)
{
struct sg_table *sgt = &transfer->sg_ctx->sgt;
size_t byte_offset = transfer->sg_byte_offset;
size_t remaining = transfer->size;
struct scatterlist *sg = NULL;
size_t intra_off = 0;
dma_addr_t dma_addr;
size_t entry_avail;
size_t len;
uint64_t *desc;
size_t total_desc_cnt;
int list_cnt, list_idx, desc_idx;
int ret;
*out_ba = 0;
ret = mx_sg_locate(sgt, byte_offset, &sg, &intra_off);
if (ret) {
pr_warn("Failed to locate sg slice (byte_offset=%zu)\n", byte_offset);
return ret;
}
total_desc_cnt = desc_cnt;
if (total_desc_cnt == 0) {
pr_warn("desc count is 0 (byte_size=%zu, skip_first=%d)\n", remaining, skip_first_entry);
return -EINVAL;
}
list_cnt = mx_get_list_count(total_desc_cnt, descs_per_list);
ret = desc_list_alloc(mx_pdev, transfer, list_cnt);
if (ret) {
pr_warn("Failed to desc_list_alloc (err=%d)\n", ret);
return ret;
}
list_idx = 0;
desc_idx = 0;
desc = (uint64_t *)transfer->desc_list_va[list_idx];
dma_addr = sg_dma_address(sg) + intra_off;
entry_avail = sg_dma_len(sg) - intra_off;
len = mx_prp_first_chunk_len(sg, intra_off, dma_size);
len = min3(len, entry_avail, remaining);
if (skip_first_entry) {
dma_addr += len;
entry_avail -= len;
remaining -= len;
if (entry_avail == 0 && remaining > 0) {
sg = sg_next(sg);
if (!sg) {
pr_warn("sg_next NULL after skip_first\n");
desc_list_free(mx_pdev, transfer);
return -EINVAL;
}
dma_addr = sg_dma_address(sg);
entry_avail = sg_dma_len(sg);
}
if (dma_addr & (dma_size - 1))
goto misaligned;
len = min3(dma_size, entry_avail, remaining);
}
while (remaining > 0) {
if (desc_idx == descs_per_list - 1 && total_desc_cnt > 1) {
if (list_idx + 1 >= list_cnt)
goto overrun;
desc[desc_idx] = (uint64_t)transfer->desc_list_ba[++list_idx];
desc = (uint64_t *)transfer->desc_list_va[list_idx];
desc_idx = 0;
}
if (desc_idx >= descs_per_list || total_desc_cnt == 0)
goto overrun;
desc[desc_idx++] = dma_addr;
dma_addr += len;
entry_avail -= len;
remaining -= len;
total_desc_cnt--;
if (remaining == 0)
break;
if (entry_avail == 0) {
sg = sg_next(sg);
if (!sg) {
pr_warn("sg_next NULL mid-walk (remaining=%zu)\n", remaining);
desc_list_free(mx_pdev, transfer);
return -EINVAL;
}
dma_addr = sg_dma_address(sg);
entry_avail = sg_dma_len(sg);
}
if (dma_addr & (dma_size - 1))
goto misaligned;
len = min3(dma_size, entry_avail, remaining);
}
*out_ba = transfer->desc_list_ba[0];
return 0;
misaligned:
pr_warn("desc walk left a %zu-byte chunk boundary (dma=%pad, remaining=%zu)\n",
dma_size, &dma_addr, remaining);
desc_list_free(mx_pdev, transfer);
return -EINVAL;
overrun:
pr_warn("desc count disagrees with emit walk (remaining=%zu, list=%d/%d, idx=%d)\n",
remaining, list_idx, list_cnt, desc_idx);
desc_list_free(mx_pdev, transfer);
return -EINVAL;
}
#define BACKOFF_SPIN_ITERS 100
#define BACKOFF_BASE_SLEEP_US 8
#define BACKOFF_MAX_SLEEP_US 16000
#define BACKOFF_TICKS_PER_LEVEL 4
#define BACKOFF_MAX_SHIFT (ilog2(BACKOFF_MAX_SLEEP_US / BACKOFF_BASE_SLEEP_US) + 1)
static inline void poll_backoff(unsigned int *idle_count)
{
unsigned int count = min(*idle_count + 1, 255u);
unsigned int shift, sleep_us;
*idle_count = count;
BUILD_BUG_ON((BACKOFF_BASE_SLEEP_US << BACKOFF_MAX_SHIFT) < BACKOFF_MAX_SLEEP_US);
if (count <= BACKOFF_SPIN_ITERS) {
cond_resched();
return;
}
shift = min_t(unsigned int, (count - BACKOFF_SPIN_ITERS - 1) / BACKOFF_TICKS_PER_LEVEL,
BACKOFF_MAX_SHIFT);
sleep_us = min_t(unsigned int, BACKOFF_BASE_SLEEP_US << shift, BACKOFF_MAX_SLEEP_US);
usleep_range_state(sleep_us, sleep_us + (sleep_us >> 3), TASK_INTERRUPTIBLE);
}
void mx_queue_common_init(struct mx_queue *q, struct mx_pci_dev *mx_pdev, const struct mx_queue_ops *ops)
{
q->dev = &mx_pdev->pdev->dev;
q->mx_pdev = mx_pdev;
q->ops = ops;
spin_lock_init(&q->sq_lock);
INIT_LIST_HEAD(&q->sq_list);
init_swait_queue_head(&q->sq_wait);
init_swait_queue_head(&q->cq_wait);
atomic_set(&q->wait_count, 0);
atomic_set(&q->zombie_wait_count, 0);
atomic_set(&q->lv_health, MX_LIVENESS_ALIVE);
q->lv_progress_jiffies = jiffies;
}
int mx_start_io_queue(struct mx_pci_dev *mx_pdev, struct mx_queue *q)
{
int ret;
mx_pdev->submit_thread = kthread_run(mx_submit_handler, q, "mx_submit_thd%d", mx_pdev->dev_id);
if (IS_ERR(mx_pdev->submit_thread)) {
ret = PTR_ERR(mx_pdev->submit_thread);
pr_err("Failed to create submit thread (err=%d)\n", ret);
mx_pdev->submit_thread = NULL;
return ret;
}
sched_set_fifo_low(mx_pdev->submit_thread);
mx_pdev->complete_thread = kthread_run(mx_complete_handler, q, "mx_complete_thd%d", mx_pdev->dev_id);
if (IS_ERR(mx_pdev->complete_thread)) {
ret = PTR_ERR(mx_pdev->complete_thread);
pr_err("Failed to create complete thread (err=%d)\n", ret);
kthread_stop(mx_pdev->submit_thread);
mx_pdev->submit_thread = NULL;
mx_pdev->complete_thread = NULL;
return ret;
}
sched_set_fifo_low(mx_pdev->complete_thread);
mx_pdev->io_queue = q;
mx_bind_handlers_to_numa(mx_pdev);
return 0;
}
void mx_stop_queue_threads(struct mx_pci_dev *mx_pdev)
{
int ret;
if (!IS_ERR_OR_NULL(mx_pdev->submit_thread)) {
ret = kthread_stop(mx_pdev->submit_thread);
if (ret)
pr_err("submit_thread thread doesn't stop properly (err=%d)\n", ret);
}
mx_pdev->submit_thread = NULL;
if (!IS_ERR_OR_NULL(mx_pdev->complete_thread)) {
ret = kthread_stop(mx_pdev->complete_thread);
if (ret)
pr_err("complete_thread thread doesn't stop properly (err=%d)\n", ret);
}
mx_pdev->complete_thread = NULL;
}
static unsigned long mx_liveness_stalled_ms(struct mx_queue *q)
{
unsigned long progress = READ_ONCE(q->lv_progress_jiffies);
unsigned long now = jiffies;
return time_after(now, progress) ? jiffies_to_msecs(now - progress) : 0;
}
static int mx_queue_live_count(struct mx_queue *q)
{
return atomic_read(&q->wait_count) - atomic_read(&q->zombie_wait_count);
}
static long mx_sq_idle_timeout(struct mx_queue *q, bool lv_on)
{
if (lv_on && (mx_queue_live_count(q) > 0 || atomic_read(&q->lv_inflight)))
return msecs_to_jiffies(POLLING_INTERVAL_MSEC);
return MAX_SCHEDULE_TIMEOUT;
}
static long mx_cq_idle_timeout(struct mx_queue *q)
{
if (mx_queue_live_count(q) > 0 || atomic_read(&q->lv_inflight))
return msecs_to_jiffies(POLLING_INTERVAL_MSEC);
return msecs_to_jiffies(ZOMBIE_POLL_INTERVAL_MSEC);
}
static void mx_liveness_watchdog(struct mx_queue *q)
{
struct mx_pci_dev *mx_pdev = q->mx_pdev;
int outstanding = atomic_read(&q->wait_count) - atomic_read(&q->zombie_wait_count);
unsigned int dead_ms = max(READ_ONCE(mx_pdev->liveness_dead_ms), 2u);
unsigned int stall_ms = clamp(READ_ONCE(mx_pdev->liveness_stall_ms), 1u, dead_ms - 1);
unsigned long stalled_ms;
u64 pong_wait_ns;
pong_wait_ns = ktime_get_ns() - READ_ONCE(q->lv_sent_ns);
if (atomic_read(&q->lv_inflight) &&
pong_wait_ns > (u64)dead_ms * NSEC_PER_MSEC) {
atomic_cmpxchg(&q->lv_health, MX_LIVENESS_SUSPECT, MX_LIVENESS_DEAD);
if (q->ops->is_pushable(q)) {
dev_warn_ratelimited(q->dev,
"liveness: no pong for %llu ms (outstanding=%d), re-probing\n",
pong_wait_ns / NSEC_PER_MSEC, outstanding);
WRITE_ONCE(q->lv_sent_ns, ktime_get_ns());
q->ops->build_ping_command(q->lv_ping_cmd);
q->ops->push_command(q, q->lv_ping_cmd);
}
return;
}
if (outstanding <= 0)
return;
stalled_ms = mx_liveness_stalled_ms(q);
if (stalled_ms > dead_ms && atomic_read(&q->lv_inflight) == 0 &&
mx_liveness_stalled_ms(q) > dead_ms)
atomic_set(&q->lv_health, MX_LIVENESS_DEAD);
if (stalled_ms > stall_ms && atomic_read(&q->lv_inflight) == 0 &&
q->ops->is_pushable(q) &&
atomic_cmpxchg(&q->lv_inflight, 0, 1) == 0) {
atomic_cmpxchg(&q->lv_health, MX_LIVENESS_ALIVE, MX_LIVENESS_SUSPECT);
WRITE_ONCE(q->lv_sent_ns, ktime_get_ns());
q->ops->build_ping_command(q->lv_ping_cmd);
q->ops->push_command(q, q->lv_ping_cmd);
}
}
static unsigned int mx_sq_push_locked(struct mx_queue *q, bool lv_on, bool *went_live)
{
const struct mx_queue_ops *ops = q->ops;
struct mx_transfer *transfer, *tmp;
unsigned int pushed_count = 0;
bool no_completion;
void *command;
lockdep_assert_held(&q->sq_lock);
*went_live = false;
list_for_each_entry_safe(transfer, tmp, &q->sq_list, entry) {
if (lv_on && atomic_read(&q->lv_inflight))
break;
if (!ops->is_pushable(q))
break;
no_completion = transfer->no_completion;
command = transfer->command;
list_del_init(&transfer->entry);
trace_mx_dma_xfer_submit((u32)transfer->id, no_completion);
if (!no_completion &&
atomic_inc_return(&q->wait_count) -
atomic_read(&q->zombie_wait_count) <= 1) {
WRITE_ONCE(q->lv_progress_jiffies, jiffies);
*went_live = true;
}
/* INTENTIONAL: a completion may free transfer once pushed; read fields first. */
ops->push_command(q, command);
pushed_count++;
if (no_completion) {
complete(&transfer->done);
} else {
swake_up_one(&q->cq_wait);
}
}
return pushed_count;
}
void mx_queue_submit(struct mx_queue *q, struct mx_transfer **transfers, int count)
{
unsigned int pushed_count;
unsigned long flags;
bool lv_on, went_live, backlog;
int i;
spin_lock_irqsave(&q->sq_lock, flags);
for (i = 0; i < count; i++)
list_add_tail(&transfers[i]->entry, &q->sq_list);
lv_on = READ_ONCE(q->mx_pdev->liveness_enable);
pushed_count = mx_sq_push_locked(q, lv_on, &went_live);
if (pushed_count && q->ops->post_submit)
q->ops->post_submit(q);
if (pushed_count)
trace_mx_dma_xfer_post_submit(q->mx_pdev->dev_id, pushed_count);
backlog = !list_empty(&q->sq_list);
spin_unlock_irqrestore(&q->sq_lock, flags);
if (backlog || (lv_on && went_live))
swake_up_one(&q->sq_wait);
}
int mx_submit_handler(void *arg)
{
struct mx_queue *q = (struct mx_queue *)arg;
const struct mx_queue_ops *ops = q->ops;
unsigned long flags;
unsigned int idle_count = 0;
unsigned int pushed_count;
bool lv_on, rearm, went_live;
long timeout;
while (!kthread_should_stop()) {
lv_on = READ_ONCE(q->mx_pdev->liveness_enable);
timeout = mx_sq_idle_timeout(q, lv_on);
rearm = lv_on && timeout == MAX_SCHEDULE_TIMEOUT;
__swait_event_interruptible_timeout(q->sq_wait,
!list_empty(&q->sq_list) ||
READ_ONCE(q->mx_pdev->liveness_enable) != lv_on ||
(rearm && mx_queue_live_count(q) > 0) || kthread_should_stop(),
timeout);
lv_on = READ_ONCE(q->mx_pdev->liveness_enable);
spin_lock_irqsave(&q->sq_lock, flags);
pushed_count = mx_sq_push_locked(q, lv_on, &went_live);
if (lv_on)
mx_liveness_watchdog(q);
if (ops->post_submit)
ops->post_submit(q);
if (pushed_count)
trace_mx_dma_xfer_post_submit(q->mx_pdev->dev_id,
pushed_count);
spin_unlock_irqrestore(&q->sq_lock, flags);
if (pushed_count || list_empty(&q->sq_list))
idle_count = 0;
else
poll_backoff(&idle_count);
}
return 0;
}
int mx_complete_handler(void *arg)
{
struct mx_queue *q = (struct mx_queue *)arg;
const struct mx_queue_ops *ops = q->ops;
struct mx_transfer *transfer;
struct mx_completion_info info;
unsigned long id_flags;
unsigned int idle_count = 0;
while (!kthread_should_stop()) {
bool popped_any = false;
__swait_event_interruptible_timeout(q->cq_wait,
mx_queue_live_count(q) > 0 || kthread_should_stop(),
mx_cq_idle_timeout(q));
while (ops->is_popable(q)) {
popped_any = true;
ops->pop_completion(q, &info);
WRITE_ONCE(q->lv_progress_jiffies, jiffies);
atomic_set(&q->lv_health, MX_LIVENESS_ALIVE);
if (info.id == MX_PING_ID) {
if (atomic_xchg(&q->lv_inflight, 0) == 1)
WRITE_ONCE(q->lv_rtt_ns,
ktime_get_ns() - READ_ONCE(q->lv_sent_ns));
continue;
}
atomic_set(&q->lv_inflight, 0);
transfer = transfer_id_claim_completion(info.id, &id_flags);
if (!transfer) {
trace_mx_dma_xfer_complete_orphan((u32)info.id, info.status, info.result);
dev_warn_ratelimited(q->dev,
"Completion for unknown transfer (id=%d)\n", info.id);
continue;
}
trace_mx_dma_xfer_complete((u32)info.id, info.status, info.result,
READ_ONCE(transfer->is_zombie));
atomic_dec(&q->wait_count);
if (READ_ONCE(transfer->is_zombie)) {
transfer_id_complete_unlock(id_flags);
continue;
}
transfer->result = info.result;
transfer->status = info.status;
complete(&transfer->done);
transfer_id_complete_unlock(id_flags);
}
if (ops->post_complete)
ops->post_complete(q);
if (popped_any || mx_queue_live_count(q) <= 0)
idle_count = 0;
else
poll_backoff(&idle_count);
}
return 0;
}
static void mx_bar_map_release(struct kref *kref)
{
struct mx_bar_map *bar_map = container_of(kref, struct mx_bar_map, kref);
mutex_destroy(&bar_map->lock);
kfree(bar_map);
}
struct mx_bar_map *mx_bar_map_alloc(void)
{
struct mx_bar_map *bar_map;
bar_map = kzalloc(sizeof(*bar_map), GFP_KERNEL);
if (!bar_map)
return NULL;
kref_init(&bar_map->kref);
mutex_init(&bar_map->lock);
return bar_map;
}
void mx_bar_map_put(struct mx_bar_map *bar_map)
{
kref_put(&bar_map->kref, mx_bar_map_release);
}
static void mx_bar_vma_open(struct vm_area_struct *vma)
{
struct mx_bar_vma *bar_vma = vma->vm_private_data;
struct mx_bar_map *bar_map = bar_vma->bar_map;
mutex_lock(&bar_map->lock);
kref_get(&bar_map->kref);
bar_map->count++;
refcount_inc(&bar_vma->refs);
mutex_unlock(&bar_map->lock);
}
static void mx_bar_vma_close(struct vm_area_struct *vma)
{
struct mx_bar_vma *bar_vma = vma->vm_private_data;
struct mx_bar_map *bar_map = bar_vma->bar_map;
struct mx_pci_dev *mx_pdev = bar_vma->mx_pdev;
struct mx_file_ctx *owner_ctx = bar_vma->owner_ctx;
bool last;
mutex_lock(&bar_map->lock);
if (!WARN_ON_ONCE(!bar_map->count)) {
bar_map->count--;
if (!bar_map->count)
bar_map->mapping = NULL;
}
last = refcount_dec_and_test(&bar_vma->refs);
mutex_unlock(&bar_map->lock);
if (last) {
mutex_lock(&mx_pdev->lease.lock);
if (!WARN_ON_ONCE(!owner_ctx->bar_mapping_count))
owner_ctx->bar_mapping_count--;
mutex_unlock(&mx_pdev->lease.lock);
kfree(bar_vma);
mx_file_ctx_put(owner_ctx);
mx_pdev_put(mx_pdev);
}
mx_bar_map_put(bar_map);
}
static const struct vm_operations_struct mx_bar_vm_ops = {
.open = mx_bar_vma_open,
.close = mx_bar_vma_close,
};
int mx_bar_mmap(struct mx_pci_dev *mx_pdev, struct mx_file_ctx *ctx,
struct vm_area_struct *vma)
{
struct mx_bar_map *bar_map = mx_pdev->bar_map;
struct mx_bar_vma *bar_vma;
resource_size_t vm_size;
unsigned long pfn;
uint32_t qid;
int ret;
bar_vma = kzalloc(sizeof(*bar_vma), GFP_KERNEL);
if (!bar_vma)
return -ENOMEM;
mutex_lock(&bar_map->lock);
if (!mx_pdev->enabled) {
ret = -ENODEV;
goto out_unlock;
}
for (qid = 0; qid < MAX_NUM_OF_MBOX; qid++) {
if (mx_pdev->sq_mbox_list[qid]) {
ret = -EBUSY;
goto out_unlock;
}
}
if (vma->vm_pgoff != 0) {
ret = -EINVAL;
goto out_unlock;
}
if (!(vma->vm_flags & VM_SHARED)) {
ret = -EINVAL;
goto out_unlock;
}
vm_size = vma->vm_end - vma->vm_start;
if (vm_size != mx_pdev->bar_mapped_size) {
ret = -EINVAL;
goto out_unlock;
}
vma->vm_page_prot = pgprot_noncached(vma->vm_page_prot);
#if LINUX_VERSION_CODE >= KERNEL_VERSION(6, 3, 0) || RHEL_RELEASE_CODE >= RHEL_RELEASE_VERSION(9, 6)
vm_flags_set(vma, VM_IO | VM_PFNMAP | VM_DONTEXPAND | VM_DONTDUMP);
#else
vma->vm_flags |= (VM_IO | VM_PFNMAP | VM_DONTEXPAND | VM_DONTDUMP);
#endif
/* INTENTIONAL: full-BAR mappers must leave driver-owned regions (v1 qid 48) alone. */
pfn = pci_resource_start(mx_pdev->pdev, MXDMA_BAR_INDEX) >> PAGE_SHIFT;
mutex_lock(&mx_pdev->lease.lock);
if (ctx->bar_mapping_count == U32_MAX) {
mutex_unlock(&mx_pdev->lease.lock);
ret = -EOVERFLOW;
goto out_unlock;
}
mutex_unlock(&mx_pdev->lease.lock);
ret = io_remap_pfn_range(vma, vma->vm_start, pfn, vm_size,
vma->vm_page_prot);
if (!ret) {
bar_map->mapping = vma->vm_file->f_mapping;
mx_file_ctx_get(ctx);
bar_vma->owner_ctx = ctx;
mutex_lock(&mx_pdev->lease.lock);
ctx->bar_mapping_count++;
mutex_unlock(&mx_pdev->lease.lock);
kref_get(&mx_pdev->ref);
bar_vma->mx_pdev = mx_pdev;
bar_vma->bar_map = bar_map;
refcount_set(&bar_vma->refs, 1);
vma->vm_private_data = bar_vma;
vma->vm_ops = &mx_bar_vm_ops;
kref_get(&bar_map->kref);
bar_map->count++;
bar_vma = NULL;
}
out_unlock:
mutex_unlock(&bar_map->lock);
kfree(bar_vma);
return ret;
}