qib_keys.c 9.55 KB
Newer Older
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
/*
 * Copyright (c) 2006, 2007, 2009 QLogic Corporation. All rights reserved.
 * Copyright (c) 2005, 2006 PathScale, Inc. 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 "qib.h"

/**
 * qib_alloc_lkey - allocate an lkey
 * @mr: memory region that this lkey protects
39 40 41 42
 * @dma_region: 0->normal key, 1->restricted DMA key
 *
 * Returns 0 if successful, otherwise returns -errno.
 *
43
 * Increments mr reference count as required.
44 45
 *
 * Sets the lkey field mr for non-dma regions.
46 47 48
 *
 */

49
int qib_alloc_lkey(struct qib_mregion *mr, int dma_region)
50 51 52 53
{
	unsigned long flags;
	u32 r;
	u32 n;
54 55 56
	int ret = 0;
	struct qib_ibdev *dev = to_idev(mr->pd->device);
	struct qib_lkey_table *rkt = &dev->lk_table;
57 58 59

	spin_lock_irqsave(&rkt->lock, flags);

60 61
	/* special case for dma_mr lkey == 0 */
	if (dma_region) {
62 63
		struct qib_mregion *tmr;

64
		tmr = rcu_access_pointer(dev->dma_mr);
65
		if (!tmr) {
66
			qib_get_mr(mr);
67
			rcu_assign_pointer(dev->dma_mr, mr);
68 69 70 71 72
			mr->lkey_published = 1;
		}
		goto success;
	}

73 74 75 76 77 78 79
	/* Find the next available LKEY */
	r = rkt->next;
	n = r;
	for (;;) {
		if (rkt->table[r] == NULL)
			break;
		r = (r + 1) & (rkt->max - 1);
80
		if (r == n)
81 82 83 84 85 86 87 88
			goto bail;
	}
	rkt->next = (r + 1) & (rkt->max - 1);
	/*
	 * Make sure lkey is never zero which is reserved to indicate an
	 * unrestricted LKEY.
	 */
	rkt->gen++;
89 90 91 92
	/*
	 * bits are capped in qib_verbs.c to insure enough bits
	 * for generation number
	 */
93 94 95 96 97 98 99
	mr->lkey = (r << (32 - ib_qib_lkey_table_size)) |
		((((1 << (24 - ib_qib_lkey_table_size)) - 1) & rkt->gen)
		 << 8);
	if (mr->lkey == 0) {
		mr->lkey |= 1 << 8;
		rkt->gen++;
	}
100
	qib_get_mr(mr);
101
	rcu_assign_pointer(rkt->table[r], mr);
102 103
	mr->lkey_published = 1;
success:
104
	spin_unlock_irqrestore(&rkt->lock, flags);
105
out:
106
	return ret;
107 108 109 110
bail:
	spin_unlock_irqrestore(&rkt->lock, flags);
	ret = -ENOMEM;
	goto out;
111 112 113 114
}

/**
 * qib_free_lkey - free an lkey
115
 * @mr: mr to free from tables
116
 */
117
void qib_free_lkey(struct qib_mregion *mr)
118 119 120 121
{
	unsigned long flags;
	u32 lkey = mr->lkey;
	u32 r;
122 123 124 125 126 127
	struct qib_ibdev *dev = to_idev(mr->pd->device);
	struct qib_lkey_table *rkt = &dev->lk_table;

	spin_lock_irqsave(&rkt->lock, flags);
	if (!mr->lkey_published)
		goto out;
128 129 130
	if (lkey == 0)
		rcu_assign_pointer(dev->dma_mr, NULL);
	else {
131
		r = lkey >> (32 - ib_qib_lkey_table_size);
132
		rcu_assign_pointer(rkt->table[r], NULL);
133
	}
134 135
	qib_put_mr(mr);
	mr->lkey_published = 0;
136
out:
137
	spin_unlock_irqrestore(&rkt->lock, flags);
138 139 140 141 142
}

/**
 * qib_lkey_ok - check IB SGE for validity and initialize
 * @rkt: table containing lkey to check SGE against
143
 * @pd: protection domain
144 145 146 147 148 149
 * @isge: outgoing internal SGE
 * @sge: SGE to check
 * @acc: access flags
 *
 * Return 1 if valid and successful, otherwise returns 0.
 *
150 151
 * increments the reference count upon success
 *
152 153 154 155 156 157 158 159 160 161 162 163 164 165
 * Check the IB SGE for validity and initialize our internal version
 * of it.
 */
int qib_lkey_ok(struct qib_lkey_table *rkt, struct qib_pd *pd,
		struct qib_sge *isge, struct ib_sge *sge, int acc)
{
	struct qib_mregion *mr;
	unsigned n, m;
	size_t off;

	/*
	 * We use LKEY == zero for kernel virtual addresses
	 * (see qib_get_dma_mr and qib_dma.c).
	 */
166
	rcu_read_lock();
167 168 169 170 171
	if (sge->lkey == 0) {
		struct qib_ibdev *dev = to_idev(pd->ibpd.device);

		if (pd->user)
			goto bail;
172 173 174 175
		mr = rcu_dereference(dev->dma_mr);
		if (!mr)
			goto bail;
		if (unlikely(!atomic_inc_not_zero(&mr->refcount)))
176
			goto bail;
177
		rcu_read_unlock();
178

179
		isge->mr = mr;
180 181 182 183 184 185 186
		isge->vaddr = (void *) sge->addr;
		isge->length = sge->length;
		isge->sge_length = sge->length;
		isge->m = 0;
		isge->n = 0;
		goto ok;
	}
187 188 189
	mr = rcu_dereference(
		rkt->table[(sge->lkey >> (32 - ib_qib_lkey_table_size))]);
	if (unlikely(!mr || mr->lkey != sge->lkey || mr->pd != &pd->ibpd))
190 191 192
		goto bail;

	off = sge->addr - mr->user_base;
193 194 195
	if (unlikely(sge->addr < mr->user_base ||
		     off + sge->length > mr->length ||
		     (mr->access_flags & acc) != acc))
196
		goto bail;
197 198 199
	if (unlikely(!atomic_inc_not_zero(&mr->refcount)))
		goto bail;
	rcu_read_unlock();
200 201

	off += mr->offset;
202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223
	if (mr->page_shift) {
		/*
		page sizes are uniform power of 2 so no loop is necessary
		entries_spanned_by_off is the number of times the loop below
		would have executed.
		*/
		size_t entries_spanned_by_off;

		entries_spanned_by_off = off >> mr->page_shift;
		off -= (entries_spanned_by_off << mr->page_shift);
		m = entries_spanned_by_off/QIB_SEGSZ;
		n = entries_spanned_by_off%QIB_SEGSZ;
	} else {
		m = 0;
		n = 0;
		while (off >= mr->map[m]->segs[n].length) {
			off -= mr->map[m]->segs[n].length;
			n++;
			if (n >= QIB_SEGSZ) {
				m++;
				n = 0;
			}
224 225 226 227 228 229 230 231 232
		}
	}
	isge->mr = mr;
	isge->vaddr = mr->map[m]->segs[n].vaddr + off;
	isge->length = mr->map[m]->segs[n].length - off;
	isge->sge_length = sge->length;
	isge->m = m;
	isge->n = n;
ok:
233
	return 1;
234
bail:
235
	rcu_read_unlock();
236
	return 0;
237 238 239 240
}

/**
 * qib_rkey_ok - check the IB virtual address, length, and RKEY
241 242
 * @qp: qp for validation
 * @sge: SGE state
243 244 245 246 247 248
 * @len: length of data
 * @vaddr: virtual address to place data
 * @rkey: rkey to check
 * @acc: access flags
 *
 * Return 1 if successful, otherwise 0.
249 250
 *
 * increments the reference count upon success
251 252 253 254 255 256 257 258 259 260 261 262 263
 */
int qib_rkey_ok(struct qib_qp *qp, struct qib_sge *sge,
		u32 len, u64 vaddr, u32 rkey, int acc)
{
	struct qib_lkey_table *rkt = &to_idev(qp->ibqp.device)->lk_table;
	struct qib_mregion *mr;
	unsigned n, m;
	size_t off;

	/*
	 * We use RKEY == zero for kernel virtual addresses
	 * (see qib_get_dma_mr and qib_dma.c).
	 */
264
	rcu_read_lock();
265 266 267 268 269 270
	if (rkey == 0) {
		struct qib_pd *pd = to_ipd(qp->ibqp.pd);
		struct qib_ibdev *dev = to_idev(pd->ibpd.device);

		if (pd->user)
			goto bail;
271 272 273 274
		mr = rcu_dereference(dev->dma_mr);
		if (!mr)
			goto bail;
		if (unlikely(!atomic_inc_not_zero(&mr->refcount)))
275
			goto bail;
276
		rcu_read_unlock();
277

278
		sge->mr = mr;
279 280 281 282 283 284 285 286
		sge->vaddr = (void *) vaddr;
		sge->length = len;
		sge->sge_length = len;
		sge->m = 0;
		sge->n = 0;
		goto ok;
	}

287 288 289
	mr = rcu_dereference(
		rkt->table[(rkey >> (32 - ib_qib_lkey_table_size))]);
	if (unlikely(!mr || mr->lkey != rkey || qp->ibqp.pd != mr->pd))
290 291 292 293 294
		goto bail;

	off = vaddr - mr->iova;
	if (unlikely(vaddr < mr->iova || off + len > mr->length ||
		     (mr->access_flags & acc) == 0))
295
		goto bail;
296 297 298
	if (unlikely(!atomic_inc_not_zero(&mr->refcount)))
		goto bail;
	rcu_read_unlock();
299 300

	off += mr->offset;
301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322
	if (mr->page_shift) {
		/*
		page sizes are uniform power of 2 so no loop is necessary
		entries_spanned_by_off is the number of times the loop below
		would have executed.
		*/
		size_t entries_spanned_by_off;

		entries_spanned_by_off = off >> mr->page_shift;
		off -= (entries_spanned_by_off << mr->page_shift);
		m = entries_spanned_by_off/QIB_SEGSZ;
		n = entries_spanned_by_off%QIB_SEGSZ;
	} else {
		m = 0;
		n = 0;
		while (off >= mr->map[m]->segs[n].length) {
			off -= mr->map[m]->segs[n].length;
			n++;
			if (n >= QIB_SEGSZ) {
				m++;
				n = 0;
			}
323 324 325 326 327 328 329 330 331
		}
	}
	sge->mr = mr;
	sge->vaddr = mr->map[m]->segs[n].vaddr + off;
	sge->length = mr->map[m]->segs[n].length - off;
	sge->sge_length = len;
	sge->m = m;
	sge->n = n;
ok:
332
	return 1;
333
bail:
334
	rcu_read_unlock();
335
	return 0;
336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356
}

/*
 * Initialize the memory region specified by the work reqeust.
 */
int qib_fast_reg_mr(struct qib_qp *qp, struct ib_send_wr *wr)
{
	struct qib_lkey_table *rkt = &to_idev(qp->ibqp.device)->lk_table;
	struct qib_pd *pd = to_ipd(qp->ibqp.pd);
	struct qib_mregion *mr;
	u32 rkey = wr->wr.fast_reg.rkey;
	unsigned i, n, m;
	int ret = -EINVAL;
	unsigned long flags;
	u64 *page_list;
	size_t ps;

	spin_lock_irqsave(&rkt->lock, flags);
	if (pd->user || rkey == 0)
		goto bail;

357 358 359
	mr = rcu_dereference_protected(
		rkt->table[(rkey >> (32 - ib_qib_lkey_table_size))],
		lockdep_is_held(&rkt->lock));
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
	if (unlikely(mr == NULL || qp->ibqp.pd != mr->pd))
		goto bail;

	if (wr->wr.fast_reg.page_list_len > mr->max_segs)
		goto bail;

	ps = 1UL << wr->wr.fast_reg.page_shift;
	if (wr->wr.fast_reg.length > ps * wr->wr.fast_reg.page_list_len)
		goto bail;

	mr->user_base = wr->wr.fast_reg.iova_start;
	mr->iova = wr->wr.fast_reg.iova_start;
	mr->lkey = rkey;
	mr->length = wr->wr.fast_reg.length;
	mr->access_flags = wr->wr.fast_reg.access_flags;
	page_list = wr->wr.fast_reg.page_list->page_list;
	m = 0;
	n = 0;
	for (i = 0; i < wr->wr.fast_reg.page_list_len; i++) {
		mr->map[m]->segs[n].vaddr = (void *) page_list[i];
		mr->map[m]->segs[n].length = ps;
		if (++n == QIB_SEGSZ) {
			m++;
			n = 0;
		}
	}

	ret = 0;
bail:
	spin_unlock_irqrestore(&rkt->lock, flags);
	return ret;
}