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
|
// SPDX-License-Identifier: (BSD-3-Clause OR GPL-2.0-only)
/* Copyright(c) 2014 - 2020 Intel Corporation */
#include <linux/delay.h>
#include "adf_accel_devices.h"
#include "adf_transport_internal.h"
#include "adf_transport_access_macros.h"
#include "adf_cfg.h"
#include "adf_common_drv.h"
static inline u32 adf_modulo(u32 data, u32 shift)
{
u32 div = data >> shift;
u32 mult = div << shift;
return data - mult;
}
static inline int adf_check_ring_alignment(u64 addr, u64 size)
{
if (((size - 1) & addr) != 0)
return -EFAULT;
return 0;
}
static int adf_verify_ring_size(u32 msg_size, u32 msg_num)
{
int i = ADF_MIN_RING_SIZE;
for (; i <= ADF_MAX_RING_SIZE; i++)
if ((msg_size * msg_num) == ADF_SIZE_TO_RING_SIZE_IN_BYTES(i))
return i;
return ADF_DEFAULT_RING_SIZE;
}
static int adf_reserve_ring(struct adf_etr_bank_data *bank, u32 ring)
{
spin_lock(&bank->lock);
if (bank->ring_mask & (1 << ring)) {
spin_unlock(&bank->lock);
return -EFAULT;
}
bank->ring_mask |= (1 << ring);
spin_unlock(&bank->lock);
return 0;
}
static void adf_unreserve_ring(struct adf_etr_bank_data *bank, u32 ring)
{
spin_lock(&bank->lock);
bank->ring_mask &= ~(1 << ring);
spin_unlock(&bank->lock);
}
static void adf_enable_ring_irq(struct adf_etr_bank_data *bank, u32 ring)
{
spin_lock_bh(&bank->lock);
bank->irq_mask |= (1 << ring);
spin_unlock_bh(&bank->lock);
WRITE_CSR_INT_COL_EN(bank->csr_addr, bank->bank_number, bank->irq_mask);
WRITE_CSR_INT_COL_CTL(bank->csr_addr, bank->bank_number,
bank->irq_coalesc_timer);
}
static void adf_disable_ring_irq(struct adf_etr_bank_data *bank, u32 ring)
{
spin_lock_bh(&bank->lock);
bank->irq_mask &= ~(1 << ring);
spin_unlock_bh(&bank->lock);
WRITE_CSR_INT_COL_EN(bank->csr_addr, bank->bank_number, bank->irq_mask);
}
int adf_send_message(struct adf_etr_ring_data *ring, u32 *msg)
{
if (atomic_add_return(1, ring->inflights) >
ADF_MAX_INFLIGHTS(ring->ring_size, ring->msg_size)) {
atomic_dec(ring->inflights);
return -EAGAIN;
}
spin_lock_bh(&ring->lock);
memcpy((void *)((uintptr_t)ring->base_addr + ring->tail), msg,
ADF_MSG_SIZE_TO_BYTES(ring->msg_size));
ring->tail = adf_modulo(ring->tail +
ADF_MSG_SIZE_TO_BYTES(ring->msg_size),
ADF_RING_SIZE_MODULO(ring->ring_size));
WRITE_CSR_RING_TAIL(ring->bank->csr_addr, ring->bank->bank_number,
ring->ring_number, ring->tail);
spin_unlock_bh(&ring->lock);
return 0;
}
static int adf_handle_response(struct adf_etr_ring_data *ring)
{
u32 msg_counter = 0;
u32 *msg = (u32 *)((uintptr_t)ring->base_addr + ring->head);
while (*msg != ADF_RING_EMPTY_SIG) {
ring->callback((u32 *)msg);
atomic_dec(ring->inflights);
*msg = ADF_RING_EMPTY_SIG;
ring->head = adf_modulo(ring->head +
ADF_MSG_SIZE_TO_BYTES(ring->msg_size),
ADF_RING_SIZE_MODULO(ring->ring_size));
msg_counter++;
msg = (u32 *)((uintptr_t)ring->base_addr + ring->head);
}
if (msg_counter > 0)
WRITE_CSR_RING_HEAD(ring->bank->csr_addr,
ring->bank->bank_number,
ring->ring_number, ring->head);
return 0;
}
static void adf_configure_tx_ring(struct adf_etr_ring_data *ring)
{
u32 ring_config = BUILD_RING_CONFIG(ring->ring_size);
WRITE_CSR_RING_CONFIG(ring->bank->csr_addr, ring->bank->bank_number,
ring->ring_number, ring_config);
}
static void adf_configure_rx_ring(struct adf_etr_ring_data *ring)
{
u32 ring_config =
BUILD_RESP_RING_CONFIG(ring->ring_size,
ADF_RING_NEAR_WATERMARK_512,
ADF_RING_NEAR_WATERMARK_0);
WRITE_CSR_RING_CONFIG(ring->bank->csr_addr, ring->bank->bank_number,
ring->ring_number, ring_config);
}
static int adf_init_ring(struct adf_etr_ring_data *ring)
{
struct adf_etr_bank_data *bank = ring->bank;
struct adf_accel_dev *accel_dev = bank->accel_dev;
struct adf_hw_device_data *hw_data = accel_dev->hw_device;
u64 ring_base;
u32 ring_size_bytes =
ADF_SIZE_TO_RING_SIZE_IN_BYTES(ring->ring_size);
ring_size_bytes = ADF_RING_SIZE_BYTES_MIN(ring_size_bytes);
ring->base_addr = dma_alloc_coherent(&GET_DEV(accel_dev),
ring_size_bytes, &ring->dma_addr,
GFP_KERNEL);
if (!ring->base_addr)
return -ENOMEM;
memset(ring->base_addr, 0x7F, ring_size_bytes);
/* The base_addr has to be aligned to the size of the buffer */
if (adf_check_ring_alignment(ring->dma_addr, ring_size_bytes)) {
dev_err(&GET_DEV(accel_dev), "Ring address not aligned\n");
dma_free_coherent(&GET_DEV(accel_dev), ring_size_bytes,
ring->base_addr, ring->dma_addr);
return -EFAULT;
}
if (hw_data->tx_rings_mask & (1 << ring->ring_number))
adf_configure_tx_ring(ring);
else
adf_configure_rx_ring(ring);
ring_base = BUILD_RING_BASE_ADDR(ring->dma_addr, ring->ring_size);
WRITE_CSR_RING_BASE(ring->bank->csr_addr, ring->bank->bank_number,
ring->ring_number, ring_base);
spin_lock_init(&ring->lock);
return 0;
}
static void adf_cleanup_ring(struct adf_etr_ring_data *ring)
{
u32 ring_size_bytes =
ADF_SIZE_TO_RING_SIZE_IN_BYTES(ring->ring_size);
ring_size_bytes = ADF_RING_SIZE_BYTES_MIN(ring_size_bytes);
if (ring->base_addr) {
memset(ring->base_addr, 0x7F, ring_size_bytes);
dma_free_coherent(&GET_DEV(ring->bank->accel_dev),
ring_size_bytes, ring->base_addr,
ring->dma_addr);
}
}
int adf_create_ring(struct adf_accel_dev *accel_dev, const char *section,
u32 bank_num, u32 num_msgs,
u32 msg_size, const char *ring_name,
adf_callback_fn callback, int poll_mode,
struct adf_etr_ring_data **ring_ptr)
{
struct adf_etr_data *transport_data = accel_dev->transport;
struct adf_etr_bank_data *bank;
struct adf_etr_ring_data *ring;
char val[ADF_CFG_MAX_VAL_LEN_IN_BYTES];
u32 ring_num;
int ret;
if (bank_num >= GET_MAX_BANKS(accel_dev)) {
dev_err(&GET_DEV(accel_dev), "Invalid bank number\n");
return -EFAULT;
}
if (msg_size > ADF_MSG_SIZE_TO_BYTES(ADF_MAX_MSG_SIZE)) {
dev_err(&GET_DEV(accel_dev), "Invalid msg size\n");
return -EFAULT;
}
if (ADF_MAX_INFLIGHTS(adf_verify_ring_size(msg_size, num_msgs),
ADF_BYTES_TO_MSG_SIZE(msg_size)) < 2) {
dev_err(&GET_DEV(accel_dev),
"Invalid ring size for given msg size\n");
return -EFAULT;
}
if (adf_cfg_get_param_value(accel_dev, section, ring_name, val)) {
dev_err(&GET_DEV(accel_dev), "Section %s, no such entry : %s\n",
section, ring_name);
return -EFAULT;
}
if (kstrtouint(val, 10, &ring_num)) {
dev_err(&GET_DEV(accel_dev), "Can't get ring number\n");
return -EFAULT;
}
if (ring_num >= ADF_ETR_MAX_RINGS_PER_BANK) {
dev_err(&GET_DEV(accel_dev), "Invalid ring number\n");
return -EFAULT;
}
bank = &transport_data->banks[bank_num];
if (adf_reserve_ring(bank, ring_num)) {
dev_err(&GET_DEV(accel_dev), "Ring %d, %s already exists.\n",
ring_num, ring_name);
return -EFAULT;
}
ring = &bank->rings[ring_num];
ring->ring_number = ring_num;
ring->bank = bank;
ring->callback = callback;
ring->msg_size = ADF_BYTES_TO_MSG_SIZE(msg_size);
ring->ring_size = adf_verify_ring_size(msg_size, num_msgs);
ring->head = 0;
ring->tail = 0;
atomic_set(ring->inflights, 0);
ret = adf_init_ring(ring);
if (ret)
goto err;
/* Enable HW arbitration for the given ring */
adf_update_ring_arb(ring);
if (adf_ring_debugfs_add(ring, ring_name)) {
dev_err(&GET_DEV(accel_dev),
"Couldn't add ring debugfs entry\n");
ret = -EFAULT;
goto err;
}
/* Enable interrupts if needed */
if (callback && (!poll_mode))
adf_enable_ring_irq(bank, ring->ring_number);
*ring_ptr = ring;
return 0;
err:
adf_cleanup_ring(ring);
adf_unreserve_ring(bank, ring_num);
adf_update_ring_arb(ring);
return ret;
}
void adf_remove_ring(struct adf_etr_ring_data *ring)
{
struct adf_etr_bank_data *bank = ring->bank;
/* Disable interrupts for the given ring */
adf_disable_ring_irq(bank, ring->ring_number);
/* Clear PCI config space */
WRITE_CSR_RING_CONFIG(bank->csr_addr, bank->bank_number,
ring->ring_number, 0);
WRITE_CSR_RING_BASE(bank->csr_addr, bank->bank_number,
ring->ring_number, 0);
adf_ring_debugfs_rm(ring);
adf_unreserve_ring(bank, ring->ring_number);
/* Disable HW arbitration for the given ring */
adf_update_ring_arb(ring);
adf_cleanup_ring(ring);
}
static void adf_ring_response_handler(struct adf_etr_bank_data *bank)
{
u32 empty_rings, i;
empty_rings = READ_CSR_E_STAT(bank->csr_addr, bank->bank_number);
empty_rings = ~empty_rings & bank->irq_mask;
for (i = 0; i < ADF_ETR_MAX_RINGS_PER_BANK; ++i) {
if (empty_rings & (1 << i))
adf_handle_response(&bank->rings[i]);
}
}
void adf_response_handler(uintptr_t bank_addr)
{
struct adf_etr_bank_data *bank = (void *)bank_addr;
/* Handle all the responses and reenable IRQs */
adf_ring_response_handler(bank);
WRITE_CSR_INT_FLAG_AND_COL(bank->csr_addr, bank->bank_number,
bank->irq_mask);
}
static inline int adf_get_cfg_int(struct adf_accel_dev *accel_dev,
const char *section, const char *format,
u32 key, u32 *value)
{
char key_buf[ADF_CFG_MAX_KEY_LEN_IN_BYTES];
char val_buf[ADF_CFG_MAX_VAL_LEN_IN_BYTES];
snprintf(key_buf, ADF_CFG_MAX_KEY_LEN_IN_BYTES, format, key);
if (adf_cfg_get_param_value(accel_dev, section, key_buf, val_buf))
return -EFAULT;
if (kstrtouint(val_buf, 10, value))
return -EFAULT;
return 0;
}
static void adf_get_coalesc_timer(struct adf_etr_bank_data *bank,
const char *section,
u32 bank_num_in_accel)
{
if (adf_get_cfg_int(bank->accel_dev, section,
ADF_ETRMGR_COALESCE_TIMER_FORMAT,
bank_num_in_accel, &bank->irq_coalesc_timer))
bank->irq_coalesc_timer = ADF_COALESCING_DEF_TIME;
if (ADF_COALESCING_MAX_TIME < bank->irq_coalesc_timer ||
ADF_COALESCING_MIN_TIME > bank->irq_coalesc_timer)
bank->irq_coalesc_timer = ADF_COALESCING_DEF_TIME;
}
static int adf_init_bank(struct adf_accel_dev *accel_dev,
struct adf_etr_bank_data *bank,
u32 bank_num, void __iomem *csr_addr)
{
struct adf_hw_device_data *hw_data = accel_dev->hw_device;
struct adf_etr_ring_data *ring;
struct adf_etr_ring_data *tx_ring;
u32 i, coalesc_enabled = 0;
memset(bank, 0, sizeof(*bank));
bank->bank_number = bank_num;
bank->csr_addr = csr_addr;
bank->accel_dev = accel_dev;
spin_lock_init(&bank->lock);
/* Enable IRQ coalescing always. This will allow to use
* the optimised flag and coalesc register.
* If it is disabled in the config file just use min time value */
if ((adf_get_cfg_int(accel_dev, "Accelerator0",
ADF_ETRMGR_COALESCING_ENABLED_FORMAT, bank_num,
&coalesc_enabled) == 0) && coalesc_enabled)
adf_get_coalesc_timer(bank, "Accelerator0", bank_num);
else
bank->irq_coalesc_timer = ADF_COALESCING_MIN_TIME;
for (i = 0; i < ADF_ETR_MAX_RINGS_PER_BANK; i++) {
WRITE_CSR_RING_CONFIG(csr_addr, bank_num, i, 0);
WRITE_CSR_RING_BASE(csr_addr, bank_num, i, 0);
ring = &bank->rings[i];
if (hw_data->tx_rings_mask & (1 << i)) {
ring->inflights =
kzalloc_node(sizeof(atomic_t),
GFP_KERNEL,
dev_to_node(&GET_DEV(accel_dev)));
if (!ring->inflights)
goto err;
} else {
if (i < hw_data->tx_rx_gap) {
dev_err(&GET_DEV(accel_dev),
"Invalid tx rings mask config\n");
goto err;
}
tx_ring = &bank->rings[i - hw_data->tx_rx_gap];
ring->inflights = tx_ring->inflights;
}
}
if (adf_bank_debugfs_add(bank)) {
dev_err(&GET_DEV(accel_dev),
"Failed to add bank debugfs entry\n");
goto err;
}
WRITE_CSR_INT_FLAG(csr_addr, bank_num, ADF_BANK_INT_FLAG_CLEAR_MASK);
WRITE_CSR_INT_SRCSEL(csr_addr, bank_num);
return 0;
err:
for (i = 0; i < ADF_ETR_MAX_RINGS_PER_BANK; i++) {
ring = &bank->rings[i];
if (hw_data->tx_rings_mask & (1 << i))
kfree(ring->inflights);
}
return -ENOMEM;
}
/**
* adf_init_etr_data() - Initialize transport rings for acceleration device
* @accel_dev: Pointer to acceleration device.
*
* Function is the initializes the communications channels (rings) to the
* acceleration device accel_dev.
* To be used by QAT device specific drivers.
*
* Return: 0 on success, error code otherwise.
*/
int adf_init_etr_data(struct adf_accel_dev *accel_dev)
{
struct adf_etr_data *etr_data;
struct adf_hw_device_data *hw_data = accel_dev->hw_device;
void __iomem *csr_addr;
u32 size;
u32 num_banks = 0;
int i, ret;
etr_data = kzalloc_node(sizeof(*etr_data), GFP_KERNEL,
dev_to_node(&GET_DEV(accel_dev)));
if (!etr_data)
return -ENOMEM;
num_banks = GET_MAX_BANKS(accel_dev);
size = num_banks * sizeof(struct adf_etr_bank_data);
etr_data->banks = kzalloc_node(size, GFP_KERNEL,
dev_to_node(&GET_DEV(accel_dev)));
if (!etr_data->banks) {
ret = -ENOMEM;
goto err_bank;
}
accel_dev->transport = etr_data;
i = hw_data->get_etr_bar_id(hw_data);
csr_addr = accel_dev->accel_pci_dev.pci_bars[i].virt_addr;
/* accel_dev->debugfs_dir should always be non-NULL here */
etr_data->debug = debugfs_create_dir("transport",
accel_dev->debugfs_dir);
for (i = 0; i < num_banks; i++) {
ret = adf_init_bank(accel_dev, &etr_data->banks[i], i,
csr_addr);
if (ret)
goto err_bank_all;
}
return 0;
err_bank_all:
debugfs_remove(etr_data->debug);
kfree(etr_data->banks);
err_bank:
kfree(etr_data);
accel_dev->transport = NULL;
return ret;
}
EXPORT_SYMBOL_GPL(adf_init_etr_data);
static void cleanup_bank(struct adf_etr_bank_data *bank)
{
u32 i;
for (i = 0; i < ADF_ETR_MAX_RINGS_PER_BANK; i++) {
struct adf_accel_dev *accel_dev = bank->accel_dev;
struct adf_hw_device_data *hw_data = accel_dev->hw_device;
struct adf_etr_ring_data *ring = &bank->rings[i];
if (bank->ring_mask & (1 << i))
adf_cleanup_ring(ring);
if (hw_data->tx_rings_mask & (1 << i))
kfree(ring->inflights);
}
adf_bank_debugfs_rm(bank);
memset(bank, 0, sizeof(*bank));
}
static void adf_cleanup_etr_handles(struct adf_accel_dev *accel_dev)
{
struct adf_etr_data *etr_data = accel_dev->transport;
u32 i, num_banks = GET_MAX_BANKS(accel_dev);
for (i = 0; i < num_banks; i++)
cleanup_bank(&etr_data->banks[i]);
}
/**
* adf_cleanup_etr_data() - Clear transport rings for acceleration device
* @accel_dev: Pointer to acceleration device.
*
* Function is the clears the communications channels (rings) of the
* acceleration device accel_dev.
* To be used by QAT device specific drivers.
*
* Return: void
*/
void adf_cleanup_etr_data(struct adf_accel_dev *accel_dev)
{
struct adf_etr_data *etr_data = accel_dev->transport;
if (etr_data) {
adf_cleanup_etr_handles(accel_dev);
debugfs_remove(etr_data->debug);
kfree(etr_data->banks);
kfree(etr_data);
accel_dev->transport = NULL;
}
}
EXPORT_SYMBOL_GPL(adf_cleanup_etr_data);
|