summaryrefslogtreecommitdiffstats
path: root/drivers/infiniband/hw/hns/hns_roce_srq.c
blob: f40a000e94ee7e550fa095f7d4f58baf74b5a1b2 (plain)
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
// SPDX-License-Identifier: GPL-2.0 OR Linux-OpenIB
/*
 * Copyright (c) 2018 Hisilicon Limited.
 */

#include <rdma/ib_umem.h>
#include <rdma/hns-abi.h>
#include "hns_roce_device.h"
#include "hns_roce_cmd.h"
#include "hns_roce_hem.h"

void hns_roce_srq_event(struct hns_roce_dev *hr_dev, u32 srqn, int event_type)
{
	struct hns_roce_srq_table *srq_table = &hr_dev->srq_table;
	struct hns_roce_srq *srq;

	xa_lock(&srq_table->xa);
	srq = xa_load(&srq_table->xa, srqn & (hr_dev->caps.num_srqs - 1));
	if (srq)
		atomic_inc(&srq->refcount);
	xa_unlock(&srq_table->xa);

	if (!srq) {
		dev_warn(hr_dev->dev, "Async event for bogus SRQ %08x\n", srqn);
		return;
	}

	srq->event(srq, event_type);

	if (atomic_dec_and_test(&srq->refcount))
		complete(&srq->free);
}

static void hns_roce_ib_srq_event(struct hns_roce_srq *srq,
				  enum hns_roce_event event_type)
{
	struct hns_roce_dev *hr_dev = to_hr_dev(srq->ibsrq.device);
	struct ib_srq *ibsrq = &srq->ibsrq;
	struct ib_event event;

	if (ibsrq->event_handler) {
		event.device      = ibsrq->device;
		event.element.srq = ibsrq;
		switch (event_type) {
		case HNS_ROCE_EVENT_TYPE_SRQ_LIMIT_REACH:
			event.event = IB_EVENT_SRQ_LIMIT_REACHED;
			break;
		case HNS_ROCE_EVENT_TYPE_SRQ_CATAS_ERROR:
			event.event = IB_EVENT_SRQ_ERR;
			break;
		default:
			dev_err(hr_dev->dev,
			   "hns_roce:Unexpected event type 0x%x on SRQ %06lx\n",
			   event_type, srq->srqn);
			return;
		}

		ibsrq->event_handler(&event, ibsrq->srq_context);
	}
}

static int hns_roce_hw_create_srq(struct hns_roce_dev *dev,
				  struct hns_roce_cmd_mailbox *mailbox,
				  unsigned long srq_num)
{
	return hns_roce_cmd_mbox(dev, mailbox->dma, 0, srq_num, 0,
				 HNS_ROCE_CMD_CREATE_SRQ,
				 HNS_ROCE_CMD_TIMEOUT_MSECS);
}

static int hns_roce_hw_destroy_srq(struct hns_roce_dev *dev,
				   struct hns_roce_cmd_mailbox *mailbox,
				   unsigned long srq_num)
{
	return hns_roce_cmd_mbox(dev, 0, mailbox ? mailbox->dma : 0, srq_num,
				 mailbox ? 0 : 1, HNS_ROCE_CMD_DESTROY_SRQ,
				 HNS_ROCE_CMD_TIMEOUT_MSECS);
}

static int alloc_srqc(struct hns_roce_dev *hr_dev, struct hns_roce_srq *srq,
		      u32 pdn, u32 cqn, u16 xrcd, u64 db_rec_addr)
{
	struct hns_roce_srq_table *srq_table = &hr_dev->srq_table;
	struct ib_device *ibdev = &hr_dev->ib_dev;
	struct hns_roce_cmd_mailbox *mailbox;
	u64 mtts_wqe[MTT_MIN_COUNT] = { 0 };
	u64 mtts_idx[MTT_MIN_COUNT] = { 0 };
	dma_addr_t dma_handle_wqe = 0;
	dma_addr_t dma_handle_idx = 0;
	int ret;

	/* Get the physical address of srq buf */
	ret = hns_roce_mtr_find(hr_dev, &srq->buf_mtr, 0, mtts_wqe,
				ARRAY_SIZE(mtts_wqe), &dma_handle_wqe);
	if (ret < 1) {
		ibdev_err(ibdev, "Failed to find mtr for SRQ WQE\n");
		return -ENOBUFS;
	}

	/* Get physical address of idx que buf */
	ret = hns_roce_mtr_find(hr_dev, &srq->idx_que.mtr, 0, mtts_idx,
				ARRAY_SIZE(mtts_idx), &dma_handle_idx);
	if (ret < 1) {
		ibdev_err(ibdev, "Failed to find mtr for SRQ idx\n");
		return -ENOBUFS;
	}

	ret = hns_roce_bitmap_alloc(&srq_table->bitmap, &srq->srqn);
	if (ret) {
		ibdev_err(ibdev, "Failed to alloc SRQ number, err %d\n", ret);
		return -ENOMEM;
	}

	ret = hns_roce_table_get(hr_dev, &srq_table->table, srq->srqn);
	if (ret) {
		ibdev_err(ibdev, "Failed to get SRQC table, err %d\n", ret);
		goto err_out;
	}

	ret = xa_err(xa_store(&srq_table->xa, srq->srqn, srq, GFP_KERNEL));
	if (ret) {
		ibdev_err(ibdev, "Failed to store SRQC, err %d\n", ret);
		goto err_put;
	}

	mailbox = hns_roce_alloc_cmd_mailbox(hr_dev);
	if (IS_ERR_OR_NULL(mailbox)) {
		ret = -ENOMEM;
		ibdev_err(ibdev, "Failed to alloc mailbox for SRQC\n");
		goto err_xa;
	}

	hr_dev->hw->write_srqc(hr_dev, srq, pdn, xrcd, cqn, mailbox->buf,
			       mtts_wqe, mtts_idx, dma_handle_wqe,
			       dma_handle_idx);

	ret = hns_roce_hw_create_srq(hr_dev, mailbox, srq->srqn);
	hns_roce_free_cmd_mailbox(hr_dev, mailbox);
	if (ret) {
		ibdev_err(ibdev, "Failed to config SRQC, err %d\n", ret);
		goto err_xa;
	}

	atomic_set(&srq->refcount, 1);
	init_completion(&srq->free);
	return ret;

err_xa:
	xa_erase(&srq_table->xa, srq->srqn);

err_put:
	hns_roce_table_put(hr_dev, &srq_table->table, srq->srqn);

err_out:
	hns_roce_bitmap_free(&srq_table->bitmap, srq->srqn, BITMAP_NO_RR);
	return ret;
}

static void free_srqc(struct hns_roce_dev *hr_dev, struct hns_roce_srq *srq)
{
	struct hns_roce_srq_table *srq_table = &hr_dev->srq_table;
	int ret;

	ret = hns_roce_hw_destroy_srq(hr_dev, NULL, srq->srqn);
	if (ret)
		dev_err(hr_dev->dev, "DESTROY_SRQ failed (%d) for SRQN %06lx\n",
			ret, srq->srqn);

	xa_erase(&srq_table->xa, srq->srqn);

	if (atomic_dec_and_test(&srq->refcount))
		complete(&srq->free);
	wait_for_completion(&srq->free);

	hns_roce_table_put(hr_dev, &srq_table->table, srq->srqn);
	hns_roce_bitmap_free(&srq_table->bitmap, srq->srqn, BITMAP_NO_RR);
}

static int alloc_srq_buf(struct hns_roce_dev *hr_dev, struct hns_roce_srq *srq,
			 struct ib_udata *udata, unsigned long addr)
{
	struct ib_device *ibdev = &hr_dev->ib_dev;
	struct hns_roce_buf_attr buf_attr = {};
	int err;

	srq->wqe_shift = ilog2(roundup_pow_of_two(max(HNS_ROCE_SGE_SIZE,
						      HNS_ROCE_SGE_SIZE *
						      srq->max_gs)));

	buf_attr.page_shift = hr_dev->caps.srqwqe_buf_pg_sz + HNS_HW_PAGE_SHIFT;
	buf_attr.region[0].size = to_hr_hem_entries_size(srq->wqe_cnt,
							 srq->wqe_shift);
	buf_attr.region[0].hopnum = hr_dev->caps.srqwqe_hop_num;
	buf_attr.region_count = 1;
	buf_attr.fixed_page = true;

	err = hns_roce_mtr_create(hr_dev, &srq->buf_mtr, &buf_attr,
				  hr_dev->caps.srqwqe_ba_pg_sz +
				  HNS_HW_PAGE_SHIFT, udata, addr);
	if (err)
		ibdev_err(ibdev, "Failed to alloc SRQ buf mtr, err %d\n", err);

	return err;
}

static void free_srq_buf(struct hns_roce_dev *hr_dev, struct hns_roce_srq *srq)
{
	hns_roce_mtr_destroy(hr_dev, &srq->buf_mtr);
}

static int alloc_srq_idx(struct hns_roce_dev *hr_dev, struct hns_roce_srq *srq,
			 struct ib_udata *udata, unsigned long addr)
{
	struct hns_roce_idx_que *idx_que = &srq->idx_que;
	struct ib_device *ibdev = &hr_dev->ib_dev;
	struct hns_roce_buf_attr buf_attr = {};
	int err;

	srq->idx_que.entry_shift = ilog2(HNS_ROCE_IDX_QUE_ENTRY_SZ);

	buf_attr.page_shift = hr_dev->caps.idx_buf_pg_sz + HNS_HW_PAGE_SHIFT;
	buf_attr.region[0].size = to_hr_hem_entries_size(srq->wqe_cnt,
					srq->idx_que.entry_shift);
	buf_attr.region[0].hopnum = hr_dev->caps.idx_hop_num;
	buf_attr.region_count = 1;
	buf_attr.fixed_page = true;

	err = hns_roce_mtr_create(hr_dev, &idx_que->mtr, &buf_attr,
				  hr_dev->caps.idx_ba_pg_sz + HNS_HW_PAGE_SHIFT,
				  udata, addr);
	if (err) {
		ibdev_err(ibdev, "Failed to alloc SRQ idx mtr, err %d\n", err);
		return err;
	}

	if (!udata) {
		idx_que->bitmap = bitmap_zalloc(srq->wqe_cnt, GFP_KERNEL);
		if (!idx_que->bitmap) {
			ibdev_err(ibdev, "Failed to alloc SRQ idx bitmap\n");
			err = -ENOMEM;
			goto err_idx_mtr;
		}

	}

	return 0;
err_idx_mtr:
	hns_roce_mtr_destroy(hr_dev, &idx_que->mtr);

	return err;
}

static void free_srq_idx(struct hns_roce_dev *hr_dev, struct hns_roce_srq *srq)
{
	struct hns_roce_idx_que *idx_que = &srq->idx_que;

	bitmap_free(idx_que->bitmap);
	idx_que->bitmap = NULL;
	hns_roce_mtr_destroy(hr_dev, &idx_que->mtr);
}

static int alloc_srq_wrid(struct hns_roce_dev *hr_dev, struct hns_roce_srq *srq)
{
	srq->head = 0;
	srq->tail = srq->wqe_cnt - 1;
	srq->wrid = kvmalloc_array(srq->wqe_cnt, sizeof(u64), GFP_KERNEL);
	if (!srq->wrid)
		return -ENOMEM;

	return 0;
}

static void free_srq_wrid(struct hns_roce_srq *srq)
{
	kfree(srq->wrid);
	srq->wrid = NULL;
}

int hns_roce_create_srq(struct ib_srq *ib_srq,
			struct ib_srq_init_attr *init_attr,
			struct ib_udata *udata)
{
	struct hns_roce_dev *hr_dev = to_hr_dev(ib_srq->device);
	struct hns_roce_ib_create_srq_resp resp = {};
	struct hns_roce_srq *srq = to_hr_srq(ib_srq);
	struct ib_device *ibdev = &hr_dev->ib_dev;
	struct hns_roce_ib_create_srq ucmd = {};
	int ret = 0;
	u32 cqn;

	/* Check the actual SRQ wqe and SRQ sge num */
	if (init_attr->attr.max_wr >= hr_dev->caps.max_srq_wrs ||
	    init_attr->attr.max_sge > hr_dev->caps.max_srq_sges)
		return -EINVAL;

	mutex_init(&srq->mutex);
	spin_lock_init(&srq->lock);

	srq->wqe_cnt = roundup_pow_of_two(init_attr->attr.max_wr + 1);
	srq->max_gs = init_attr->attr.max_sge + HNS_ROCE_RESERVED_SGE;

	if (udata) {
		ret = ib_copy_from_udata(&ucmd, udata, sizeof(ucmd));
		if (ret) {
			ibdev_err(ibdev, "Failed to copy SRQ udata, err %d\n",
				  ret);
			return ret;
		}
	}

	ret = alloc_srq_buf(hr_dev, srq, udata, ucmd.buf_addr);
	if (ret) {
		ibdev_err(ibdev, "Failed to alloc SRQ buffer, err %d\n", ret);
		return ret;
	}

	ret = alloc_srq_idx(hr_dev, srq, udata, ucmd.que_addr);
	if (ret) {
		ibdev_err(ibdev, "Failed to alloc SRQ idx, err %d\n", ret);
		goto err_buf_alloc;
	}

	if (!udata) {
		ret = alloc_srq_wrid(hr_dev, srq);
		if (ret) {
			ibdev_err(ibdev, "Failed to alloc SRQ wrid, err %d\n",
				  ret);
			goto err_idx_alloc;
		}
	}

	cqn = ib_srq_has_cq(init_attr->srq_type) ?
	      to_hr_cq(init_attr->ext.cq)->cqn : 0;
	srq->db_reg_l = hr_dev->reg_base + SRQ_DB_REG;

	ret = alloc_srqc(hr_dev, srq, to_hr_pd(ib_srq->pd)->pdn, cqn, 0, 0);
	if (ret) {
		ibdev_err(ibdev, "Failed to alloc SRQ context, err %d\n", ret);
		goto err_wrid_alloc;
	}

	srq->event = hns_roce_ib_srq_event;
	resp.srqn = srq->srqn;

	if (udata) {
		if (ib_copy_to_udata(udata, &resp,
				     min(udata->outlen, sizeof(resp)))) {
			ret = -EFAULT;
			goto err_srqc_alloc;
		}
	}

	return 0;

err_srqc_alloc:
	free_srqc(hr_dev, srq);
err_wrid_alloc:
	free_srq_wrid(srq);
err_idx_alloc:
	free_srq_idx(hr_dev, srq);
err_buf_alloc:
	free_srq_buf(hr_dev, srq);
	return ret;
}

void hns_roce_destroy_srq(struct ib_srq *ibsrq, struct ib_udata *udata)
{
	struct hns_roce_dev *hr_dev = to_hr_dev(ibsrq->device);
	struct hns_roce_srq *srq = to_hr_srq(ibsrq);

	free_srqc(hr_dev, srq);
	free_srq_idx(hr_dev, srq);
	free_srq_wrid(srq);
	free_srq_buf(hr_dev, srq);
}

int hns_roce_init_srq_table(struct hns_roce_dev *hr_dev)
{
	struct hns_roce_srq_table *srq_table = &hr_dev->srq_table;

	xa_init(&srq_table->xa);

	return hns_roce_bitmap_init(&srq_table->bitmap, hr_dev->caps.num_srqs,
				    hr_dev->caps.num_srqs - 1,
				    hr_dev->caps.reserved_srqs, 0);
}

void hns_roce_cleanup_srq_table(struct hns_roce_dev *hr_dev)
{
	hns_roce_bitmap_cleanup(&hr_dev->srq_table.bitmap);
}