aboutsummaryrefslogtreecommitdiff
path: root/drivers/infiniband/hw/mlx4/srq.c
blob: 818b7ecace5e3c812b7f7500ba44a13ac46ef397 (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
/*
 * Copyright (c) 2007 Cisco Systems, Inc. All rights reserved.
 * Copyright (c) 2007, 2008 Mellanox Technologies. All rights reserved.
 *
 * This software is available to you under a choice of one of two
 * licenses.  You may choose to be licensed under the terms of the GNU
 * General Public License (GPL) Version 2, available from the file
 * COPYING in the main directory of this source tree, or the
 * OpenIB.org BSD license below:
 *
 *     Redistribution and use in source and binary forms, with or
 *     without modification, are permitted provided that the following
 *     conditions are met:
 *
 *      - Redistributions of source code must retain the above
 *        copyright notice, this list of conditions and the following
 *        disclaimer.
 *
 *      - Redistributions in binary form must reproduce the above
 *        copyright notice, this list of conditions and the following
 *        disclaimer in the documentation and/or other materials
 *        provided with the distribution.
 *
 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
 * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
 * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
 * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
 * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
 * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
 * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
 * SOFTWARE.
 */

#include <linux/mlx4/qp.h>
#include <linux/mlx4/srq.h>
#include <linux/slab.h>

#include "mlx4_ib.h"
#include "user.h"

static void *get_wqe(struct mlx4_ib_srq *srq, int n)
{
	return mlx4_buf_offset(&srq->buf, n << srq->msrq.wqe_shift);
}

static void mlx4_ib_srq_event(struct mlx4_srq *srq, enum mlx4_event type)
{
	struct ib_event event;
	struct ib_srq *ibsrq = &to_mibsrq(srq)->ibsrq;

	if (ibsrq->event_handler) {
		event.device      = ibsrq->device;
		event.element.srq = ibsrq;
		switch (type) {
		case MLX4_EVENT_TYPE_SRQ_LIMIT:
			event.event = IB_EVENT_SRQ_LIMIT_REACHED;
			break;
		case MLX4_EVENT_TYPE_SRQ_CATAS_ERROR:
			event.event = IB_EVENT_SRQ_ERR;
			break;
		default:
			printk(KERN_WARNING "mlx4_ib: Unexpected event type %d "
			       "on SRQ %06x\n", type, srq->srqn);
			return;
		}

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

struct ib_srq *mlx4_ib_create_srq(struct ib_pd *pd,
				  struct ib_srq_init_attr *init_attr,
				  struct ib_udata *udata)
{
	struct mlx4_ib_dev *dev = to_mdev(pd->device);
	struct mlx4_ib_srq *srq;
	struct mlx4_wqe_srq_next_seg *next;
	struct mlx4_wqe_data_seg *scatter;
	int desc_size;
	int buf_size;
	int err;
	int i;

	/* Sanity check SRQ size before proceeding */
	if (init_attr->attr.max_wr  >= dev->dev->caps.max_srq_wqes ||
	    init_attr->attr.max_sge >  dev->dev->caps.max_srq_sge)
		return ERR_PTR(-EINVAL);

	srq = kmalloc(sizeof *srq, GFP_KERNEL);
	if (!srq)
		return ERR_PTR(-ENOMEM);

	mutex_init(&srq->mutex);
	spin_lock_init(&srq->lock);
	srq->msrq.max    = roundup_pow_of_two(init_attr->attr.max_wr + 1);
	srq->msrq.max_gs = init_attr->attr.max_sge;

	desc_size = max(32UL,
			roundup_pow_of_two(sizeof (struct mlx4_wqe_srq_next_seg) +
					   srq->msrq.max_gs *
					   sizeof (struct mlx4_wqe_data_seg)));
	srq->msrq.wqe_shift = ilog2(desc_size);

	buf_size = srq->msrq.max * desc_size;

	if (pd->uobject) {
		struct mlx4_ib_create_srq ucmd;

		if (ib_copy_from_udata(&ucmd, udata, sizeof ucmd)) {
			err = -EFAULT;
			goto err_srq;
		}

		srq->umem = ib_umem_get(pd->uobject->context, ucmd.buf_addr,
					buf_size, 0, 0);
		if (IS_ERR(srq->umem)) {
			err = PTR_ERR(srq->umem);
			goto err_srq;
		}

		err = mlx4_mtt_init(dev->dev, ib_umem_page_count(srq->umem),
				    ilog2(srq->umem->page_size), &srq->mtt);
		if (err)
			goto err_buf;

		err = mlx4_ib_umem_write_mtt(dev, &srq->mtt, srq->umem);
		if (err)
			goto err_mtt;

		err = mlx4_ib_db_map_user(to_mucontext(pd->uobject->context),
					  ucmd.db_addr, &srq->db);
		if (err)
			goto err_mtt;
	} else {
		err = mlx4_db_alloc(dev->dev, &srq->db, 0);
		if (err)
			goto err_srq;

		*srq->db.db = 0;

		if (mlx4_buf_alloc(dev->dev, buf_size, PAGE_SIZE * 2, &srq->buf)) {
			err = -ENOMEM;
			goto err_db;
		}

		srq->head    = 0;
		srq->tail    = srq->msrq.max - 1;
		srq->wqe_ctr = 0;

		for (i = 0; i < srq->msrq.max; ++i) {
			next = get_wqe(srq, i);
			next->next_wqe_index =
				cpu_to_be16((i + 1) & (srq->msrq.max - 1));

			for (scatter = (void *) (next + 1);
			     (void *) scatter < (void *) next + desc_size;
			     ++scatter)
				scatter->lkey = cpu_to_be32(MLX4_INVALID_LKEY);
		}

		err = mlx4_mtt_init(dev->dev, srq->buf.npages, srq->buf.page_shift,
				    &srq->mtt);
		if (err)
			goto err_buf;

		err = mlx4_buf_write_mtt(dev->dev, &srq->mtt, &srq->buf);
		if (err)
			goto err_mtt;

		srq->wrid = kmalloc(srq->msrq.max * sizeof (u64), GFP_KERNEL);
		if (!srq->wrid) {
			err = -ENOMEM;
			goto err_mtt;
		}
	}

	err = mlx4_srq_alloc(dev->dev, to_mpd(pd)->pdn, &srq->mtt,
			     srq->db.dma, &srq->msrq);
	if (err)
		goto err_wrid;

	srq->msrq.event = mlx4_ib_srq_event;

	if (pd->uobject)
		if (ib_copy_to_udata(udata, &srq->msrq.srqn, sizeof (__u32))) {
			err = -EFAULT;
			goto err_wrid;
		}

	init_attr->attr.max_wr = srq->msrq.max - 1;

	return &srq->ibsrq;

err_wrid:
	if (pd->uobject)
		mlx4_ib_db_unmap_user(to_mucontext(pd->uobject->context), &srq->db);
	else
		kfree(srq->wrid);

err_mtt:
	mlx4_mtt_cleanup(dev->dev, &srq->mtt);

err_buf:
	if (pd->uobject)
		ib_umem_release(srq->umem);
	else
		mlx4_buf_free(dev->dev, buf_size, &srq->buf);

err_db:
	if (!pd->uobject)
		mlx4_db_free(dev->dev, &srq->db);

err_srq:
	kfree(srq);

	return ERR_PTR(err);
}

int mlx4_ib_modify_srq(struct ib_srq *ibsrq, struct ib_srq_attr *attr,
		       enum ib_srq_attr_mask attr_mask, struct ib_udata *udata)
{
	struct mlx4_ib_dev *dev = to_mdev(ibsrq->device);
	struct mlx4_ib_srq *srq = to_msrq(ibsrq);
	int ret;

	/* We don't support resizing SRQs (yet?) */
	if (attr_mask & IB_SRQ_MAX_WR)
		return -EINVAL;

	if (attr_mask & IB_SRQ_LIMIT) {
		if (attr->srq_limit >= srq->msrq.max)
			return -EINVAL;

		mutex_lock(&srq->mutex);
		ret = mlx4_srq_arm(dev->dev, &srq->msrq, attr->srq_limit);
		mutex_unlock(&srq->mutex);

		if (ret)
			return ret;
	}

	return 0;
}

int mlx4_ib_query_srq(struct ib_srq *ibsrq, struct ib_srq_attr *srq_attr)
{
	struct mlx4_ib_dev *dev = to_mdev(ibsrq->device);
	struct mlx4_ib_srq *srq = to_msrq(ibsrq);
	int ret;
	int limit_watermark;

	ret = mlx4_srq_query(dev->dev, &srq->msrq, &limit_watermark);
	if (ret)
		return ret;

	srq_attr->srq_limit = limit_watermark;
	srq_attr->max_wr    = srq->msrq.max - 1;
	srq_attr->max_sge   = srq->msrq.max_gs;

	return 0;
}

int mlx4_ib_destroy_srq(struct ib_srq *srq)
{
	struct mlx4_ib_dev *dev = to_mdev(srq->device);
	struct mlx4_ib_srq *msrq = to_msrq(srq);

	mlx4_srq_free(dev->dev, &msrq->msrq);
	mlx4_mtt_cleanup(dev->dev, &msrq->mtt);

	if (srq->uobject) {
		mlx4_ib_db_unmap_user(to_mucontext(srq->uobject->context), &msrq->db);
		ib_umem_release(msrq->umem);
	} else {
		kfree(msrq->wrid);
		mlx4_buf_free(dev->dev, msrq->msrq.max << msrq->msrq.wqe_shift,
			      &msrq->buf);
		mlx4_db_free(dev->dev, &msrq->db);
	}

	kfree(msrq);

	return 0;
}

void mlx4_ib_free_srq_wqe(struct mlx4_ib_srq *srq, int wqe_index)
{
	struct mlx4_wqe_srq_next_seg *next;

	/* always called with interrupts disabled. */
	spin_lock(&srq->lock);

	next = get_wqe(srq, srq->tail);
	next->next_wqe_index = cpu_to_be16(wqe_index);
	srq->tail = wqe_index;

	spin_unlock(&srq->lock);
}

int mlx4_ib_post_srq_recv(struct ib_srq *ibsrq, struct ib_recv_wr *wr,
			  struct ib_recv_wr **bad_wr)
{
	struct mlx4_ib_srq *srq = to_msrq(ibsrq);
	struct mlx4_wqe_srq_next_seg *next;
	struct mlx4_wqe_data_seg *scat;
	unsigned long flags;
	int err = 0;
	int nreq;
	int i;

	spin_lock_irqsave(&srq->lock, flags);

	for (nreq = 0; wr; ++nreq, wr = wr->next) {
		if (unlikely(wr->num_sge > srq->msrq.max_gs)) {
			err = -EINVAL;
			*bad_wr = wr;
			break;
		}

		if (unlikely(srq->head == srq->tail)) {
			err = -ENOMEM;
			*bad_wr = wr;
			break;
		}

		srq->wrid[srq->head] = wr->wr_id;

		next      = get_wqe(srq, srq->head);
		srq->head = be16_to_cpu(next->next_wqe_index);
		scat      = (struct mlx4_wqe_data_seg *) (next + 1);

		for (i = 0; i < wr->num_sge; ++i) {
			scat[i].byte_count = cpu_to_be32(wr->sg_list[i].length);
			scat[i].lkey       = cpu_to_be32(wr->sg_list[i].lkey);
			scat[i].addr       = cpu_to_be64(wr->sg_list[i].addr);
		}

		if (i < srq->msrq.max_gs) {
			scat[i].byte_count = 0;
			scat[i].lkey       = cpu_to_be32(MLX4_INVALID_LKEY);
			scat[i].addr       = 0;
		}
	}

	if (likely(nreq)) {
		srq->wqe_ctr += nreq;

		/*
		 * Make sure that descriptors are written before
		 * doorbell record.
		 */
		wmb();

		*srq->db.db = cpu_to_be32(srq->wqe_ctr);
	}

	spin_unlock_irqrestore(&srq->lock, flags);

	return err;
}