aboutsummaryrefslogtreecommitdiff
path: root/net/rds/ib_frmr.c
blob: 48332a6ed7383c51def7402dcfef1e581fa677f7 (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
/*
 * Copyright (c) 2016 Oracle.  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 "ib_mr.h"

static struct rds_ib_mr *rds_ib_alloc_frmr(struct rds_ib_device *rds_ibdev,
					   int npages)
{
	struct rds_ib_mr_pool *pool;
	struct rds_ib_mr *ibmr = NULL;
	struct rds_ib_frmr *frmr;
	int err = 0;

	if (npages <= RDS_MR_8K_MSG_SIZE)
		pool = rds_ibdev->mr_8k_pool;
	else
		pool = rds_ibdev->mr_1m_pool;

	ibmr = rds_ib_try_reuse_ibmr(pool);
	if (ibmr)
		return ibmr;

	ibmr = kzalloc_node(sizeof(*ibmr), GFP_KERNEL,
			    rdsibdev_to_node(rds_ibdev));
	if (!ibmr) {
		err = -ENOMEM;
		goto out_no_cigar;
	}

	frmr = &ibmr->u.frmr;
	frmr->mr = ib_alloc_mr(rds_ibdev->pd, IB_MR_TYPE_MEM_REG,
			 pool->fmr_attr.max_pages);
	if (IS_ERR(frmr->mr)) {
		pr_warn("RDS/IB: %s failed to allocate MR", __func__);
		goto out_no_cigar;
	}

	ibmr->pool = pool;
	if (pool->pool_type == RDS_IB_MR_8K_POOL)
		rds_ib_stats_inc(s_ib_rdma_mr_8k_alloc);
	else
		rds_ib_stats_inc(s_ib_rdma_mr_1m_alloc);

	if (atomic_read(&pool->item_count) > pool->max_items_soft)
		pool->max_items_soft = pool->max_items;

	frmr->fr_state = FRMR_IS_FREE;
	return ibmr;

out_no_cigar:
	kfree(ibmr);
	atomic_dec(&pool->item_count);
	return ERR_PTR(err);
}

static void rds_ib_free_frmr(struct rds_ib_mr *ibmr, bool drop)
{
	struct rds_ib_mr_pool *pool = ibmr->pool;

	if (drop)
		llist_add(&ibmr->llnode, &pool->drop_list);
	else
		llist_add(&ibmr->llnode, &pool->free_list);
	atomic_add(ibmr->sg_len, &pool->free_pinned);
	atomic_inc(&pool->dirty_count);

	/* If we've pinned too many pages, request a flush */
	if (atomic_read(&pool->free_pinned) >= pool->max_free_pinned ||
	    atomic_read(&pool->dirty_count) >= pool->max_items / 5)
		queue_delayed_work(rds_ib_mr_wq, &pool->flush_worker, 10);
}

static int rds_ib_post_reg_frmr(struct rds_ib_mr *ibmr)
{
	struct rds_ib_frmr *frmr = &ibmr->u.frmr;
	struct ib_send_wr *failed_wr;
	struct ib_reg_wr reg_wr;
	int ret, off = 0;

	while (atomic_dec_return(&ibmr->ic->i_fastreg_wrs) <= 0) {
		atomic_inc(&ibmr->ic->i_fastreg_wrs);
		cpu_relax();
	}

	ret = ib_map_mr_sg_zbva(frmr->mr, ibmr->sg, ibmr->sg_len,
				&off, PAGE_SIZE);
	if (unlikely(ret != ibmr->sg_len))
		return ret < 0 ? ret : -EINVAL;

	/* Perform a WR for the fast_reg_mr. Each individual page
	 * in the sg list is added to the fast reg page list and placed
	 * inside the fast_reg_mr WR.  The key used is a rolling 8bit
	 * counter, which should guarantee uniqueness.
	 */
	ib_update_fast_reg_key(frmr->mr, ibmr->remap_count++);
	frmr->fr_state = FRMR_IS_INUSE;

	memset(&reg_wr, 0, sizeof(reg_wr));
	reg_wr.wr.wr_id = (unsigned long)(void *)ibmr;
	reg_wr.wr.opcode = IB_WR_REG_MR;
	reg_wr.wr.num_sge = 0;
	reg_wr.mr = frmr->mr;
	reg_wr.key = frmr->mr->rkey;
	reg_wr.access = IB_ACCESS_LOCAL_WRITE |
			IB_ACCESS_REMOTE_READ |
			IB_ACCESS_REMOTE_WRITE;
	reg_wr.wr.send_flags = IB_SEND_SIGNALED;

	failed_wr = &reg_wr.wr;
	ret = ib_post_send(ibmr->ic->i_cm_id->qp, &reg_wr.wr, &failed_wr);
	WARN_ON(failed_wr != &reg_wr.wr);
	if (unlikely(ret)) {
		/* Failure here can be because of -ENOMEM as well */
		frmr->fr_state = FRMR_IS_STALE;
		atomic_inc(&ibmr->ic->i_fastreg_wrs);
		if (printk_ratelimit())
			pr_warn("RDS/IB: %s returned error(%d)\n",
				__func__, ret);
	}
	return ret;
}

static int rds_ib_map_frmr(struct rds_ib_device *rds_ibdev,
			   struct rds_ib_mr_pool *pool,
			   struct rds_ib_mr *ibmr,
			   struct scatterlist *sg, unsigned int sg_len)
{
	struct ib_device *dev = rds_ibdev->dev;
	struct rds_ib_frmr *frmr = &ibmr->u.frmr;
	int i;
	u32 len;
	int ret = 0;

	/* We want to teardown old ibmr values here and fill it up with
	 * new sg values
	 */
	rds_ib_teardown_mr(ibmr);

	ibmr->sg = sg;
	ibmr->sg_len = sg_len;
	ibmr->sg_dma_len = 0;
	frmr->sg_byte_len = 0;
	WARN_ON(ibmr->sg_dma_len);
	ibmr->sg_dma_len = ib_dma_map_sg(dev, ibmr->sg, ibmr->sg_len,
					 DMA_BIDIRECTIONAL);
	if (unlikely(!ibmr->sg_dma_len)) {
		pr_warn("RDS/IB: %s failed!\n", __func__);
		return -EBUSY;
	}

	frmr->sg_byte_len = 0;
	frmr->dma_npages = 0;
	len = 0;

	ret = -EINVAL;
	for (i = 0; i < ibmr->sg_dma_len; ++i) {
		unsigned int dma_len = ib_sg_dma_len(dev, &ibmr->sg[i]);
		u64 dma_addr = ib_sg_dma_address(dev, &ibmr->sg[i]);

		frmr->sg_byte_len += dma_len;
		if (dma_addr & ~PAGE_MASK) {
			if (i > 0)
				goto out_unmap;
			else
				++frmr->dma_npages;
		}

		if ((dma_addr + dma_len) & ~PAGE_MASK) {
			if (i < ibmr->sg_dma_len - 1)
				goto out_unmap;
			else
				++frmr->dma_npages;
		}

		len += dma_len;
	}
	frmr->dma_npages += len >> PAGE_SHIFT;

	if (frmr->dma_npages > ibmr->pool->fmr_attr.max_pages) {
		ret = -EMSGSIZE;
		goto out_unmap;
	}

	ret = rds_ib_post_reg_frmr(ibmr);
	if (ret)
		goto out_unmap;

	if (ibmr->pool->pool_type == RDS_IB_MR_8K_POOL)
		rds_ib_stats_inc(s_ib_rdma_mr_8k_used);
	else
		rds_ib_stats_inc(s_ib_rdma_mr_1m_used);

	return ret;

out_unmap:
	ib_dma_unmap_sg(rds_ibdev->dev, ibmr->sg, ibmr->sg_len,
			DMA_BIDIRECTIONAL);
	ibmr->sg_dma_len = 0;
	return ret;
}

static int rds_ib_post_inv(struct rds_ib_mr *ibmr)
{
	struct ib_send_wr *s_wr, *failed_wr;
	struct rds_ib_frmr *frmr = &ibmr->u.frmr;
	struct rdma_cm_id *i_cm_id = ibmr->ic->i_cm_id;
	int ret = -EINVAL;

	if (!i_cm_id || !i_cm_id->qp || !frmr->mr)
		goto out;

	if (frmr->fr_state != FRMR_IS_INUSE)
		goto out;

	while (atomic_dec_return(&ibmr->ic->i_fastunreg_wrs) <= 0) {
		atomic_inc(&ibmr->ic->i_fastunreg_wrs);
		cpu_relax();
	}

	frmr->fr_inv = true;
	s_wr = &frmr->fr_wr;

	memset(s_wr, 0, sizeof(*s_wr));
	s_wr->wr_id = (unsigned long)(void *)ibmr;
	s_wr->opcode = IB_WR_LOCAL_INV;
	s_wr->ex.invalidate_rkey = frmr->mr->rkey;
	s_wr->send_flags = IB_SEND_SIGNALED;

	failed_wr = s_wr;
	ret = ib_post_send(i_cm_id->qp, s_wr, &failed_wr);
	WARN_ON(failed_wr != s_wr);
	if (unlikely(ret)) {
		frmr->fr_state = FRMR_IS_STALE;
		frmr->fr_inv = false;
		atomic_inc(&ibmr->ic->i_fastunreg_wrs);
		pr_err("RDS/IB: %s returned error(%d)\n", __func__, ret);
		goto out;
	}
out:
	return ret;
}

void rds_ib_mr_cqe_handler(struct rds_ib_connection *ic, struct ib_wc *wc)
{
	struct rds_ib_mr *ibmr = (void *)(unsigned long)wc->wr_id;
	struct rds_ib_frmr *frmr = &ibmr->u.frmr;

	if (wc->status != IB_WC_SUCCESS) {
		frmr->fr_state = FRMR_IS_STALE;
		if (rds_conn_up(ic->conn))
			rds_ib_conn_error(ic->conn,
					  "frmr completion <%pI4,%pI4> status %u(%s), vendor_err 0x%x, disconnecting and reconnecting\n",
					  &ic->conn->c_laddr,
					  &ic->conn->c_faddr,
					  wc->status,
					  ib_wc_status_msg(wc->status),
					  wc->vendor_err);
	}

	if (frmr->fr_inv) {
		frmr->fr_state = FRMR_IS_FREE;
		frmr->fr_inv = false;
		atomic_inc(&ic->i_fastreg_wrs);
	} else {
		atomic_inc(&ic->i_fastunreg_wrs);
	}
}

void rds_ib_unreg_frmr(struct list_head *list, unsigned int *nfreed,
		       unsigned long *unpinned, unsigned int goal)
{
	struct rds_ib_mr *ibmr, *next;
	struct rds_ib_frmr *frmr;
	int ret = 0;
	unsigned int freed = *nfreed;

	/* String all ib_mr's onto one list and hand them to ib_unmap_fmr */
	list_for_each_entry(ibmr, list, unmap_list) {
		if (ibmr->sg_dma_len)
			ret |= rds_ib_post_inv(ibmr);
	}
	if (ret)
		pr_warn("RDS/IB: %s failed (err=%d)\n", __func__, ret);

	/* Now we can destroy the DMA mapping and unpin any pages */
	list_for_each_entry_safe(ibmr, next, list, unmap_list) {
		*unpinned += ibmr->sg_len;
		frmr = &ibmr->u.frmr;
		__rds_ib_teardown_mr(ibmr);
		if (freed < goal || frmr->fr_state == FRMR_IS_STALE) {
			/* Don't de-allocate if the MR is not free yet */
			if (frmr->fr_state == FRMR_IS_INUSE)
				continue;

			if (ibmr->pool->pool_type == RDS_IB_MR_8K_POOL)
				rds_ib_stats_inc(s_ib_rdma_mr_8k_free);
			else
				rds_ib_stats_inc(s_ib_rdma_mr_1m_free);
			list_del(&ibmr->unmap_list);
			if (frmr->mr)
				ib_dereg_mr(frmr->mr);
			kfree(ibmr);
			freed++;
		}
	}
	*nfreed = freed;
}

struct rds_ib_mr *rds_ib_reg_frmr(struct rds_ib_device *rds_ibdev,
				  struct rds_ib_connection *ic,
				  struct scatterlist *sg,
				  unsigned long nents, u32 *key)
{
	struct rds_ib_mr *ibmr = NULL;
	struct rds_ib_frmr *frmr;
	int ret;

	do {
		if (ibmr)
			rds_ib_free_frmr(ibmr, true);
		ibmr = rds_ib_alloc_frmr(rds_ibdev, nents);
		if (IS_ERR(ibmr))
			return ibmr;
		frmr = &ibmr->u.frmr;
	} while (frmr->fr_state != FRMR_IS_FREE);

	ibmr->ic = ic;
	ibmr->device = rds_ibdev;
	ret = rds_ib_map_frmr(rds_ibdev, ibmr->pool, ibmr, sg, nents);
	if (ret == 0) {
		*key = frmr->mr->rkey;
	} else {
		rds_ib_free_frmr(ibmr, false);
		ibmr = ERR_PTR(ret);
	}

	return ibmr;
}

void rds_ib_free_frmr_list(struct rds_ib_mr *ibmr)
{
	struct rds_ib_mr_pool *pool = ibmr->pool;
	struct rds_ib_frmr *frmr = &ibmr->u.frmr;

	if (frmr->fr_state == FRMR_IS_STALE)
		llist_add(&ibmr->llnode, &pool->drop_list);
	else
		llist_add(&ibmr->llnode, &pool->free_list);
}