qib_qp.c 34.6 KB
Newer Older
1
/*
2
 * Copyright (c) 2012, 2013 Intel Corporation.  All rights reserved.
3
 * Copyright (c) 2006 - 2012 QLogic Corporation.  * All rights reserved.
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
 * 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 <linux/err.h>
#include <linux/vmalloc.h>
M
Mike Marciniszyn 已提交
37
#include <linux/jhash.h>
38
#include <rdma/rdma_vt.h>
39 40 41
#ifdef CONFIG_DEBUG_FS
#include <linux/seq_file.h>
#endif
42 43 44 45 46 47 48 49 50 51 52 53 54 55

#include "qib.h"

#define BITS_PER_PAGE           (PAGE_SIZE*BITS_PER_BYTE)
#define BITS_PER_PAGE_MASK      (BITS_PER_PAGE-1)

static inline unsigned mk_qpn(struct qib_qpn_table *qpt,
			      struct qpn_map *map, unsigned off)
{
	return (map - qpt->map) * BITS_PER_PAGE + off;
}

static inline unsigned find_next_offset(struct qib_qpn_table *qpt,
					struct qpn_map *map, unsigned off,
56
					unsigned n)
57 58 59
{
	if (qpt->mask) {
		off++;
60 61
		if (((off & qpt->mask) >> 1) >= n)
			off = (off | qpt->mask) + 2;
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
	} else
		off = find_next_zero_bit(map->page, BITS_PER_PAGE, off);
	return off;
}

/*
 * Convert the AETH credit code into the number of credits.
 */
static u32 credit_table[31] = {
	0,                      /* 0 */
	1,                      /* 1 */
	2,                      /* 2 */
	3,                      /* 3 */
	4,                      /* 4 */
	6,                      /* 5 */
	8,                      /* 6 */
	12,                     /* 7 */
	16,                     /* 8 */
	24,                     /* 9 */
	32,                     /* A */
	48,                     /* B */
	64,                     /* C */
	96,                     /* D */
	128,                    /* E */
	192,                    /* F */
	256,                    /* 10 */
	384,                    /* 11 */
	512,                    /* 12 */
	768,                    /* 13 */
	1024,                   /* 14 */
	1536,                   /* 15 */
	2048,                   /* 16 */
	3072,                   /* 17 */
	4096,                   /* 18 */
	6144,                   /* 19 */
	8192,                   /* 1A */
	12288,                  /* 1B */
	16384,                  /* 1C */
	24576,                  /* 1D */
	32768                   /* 1E */
};

104 105
static void get_map_page(struct qib_qpn_table *qpt, struct qpn_map *map,
			 gfp_t gfp)
106
{
107
	unsigned long page = get_zeroed_page(gfp);
108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125

	/*
	 * Free the page if someone raced with us installing it.
	 */

	spin_lock(&qpt->lock);
	if (map->page)
		free_page(page);
	else
		map->page = (void *)page;
	spin_unlock(&qpt->lock);
}

/*
 * Allocate the next available QPN or
 * zero/one for QP type IB_QPT_SMI/IB_QPT_GSI.
 */
static int alloc_qpn(struct qib_devdata *dd, struct qib_qpn_table *qpt,
126
		     enum ib_qp_type type, u8 port, gfp_t gfp)
127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145
{
	u32 i, offset, max_scan, qpn;
	struct qpn_map *map;
	u32 ret;

	if (type == IB_QPT_SMI || type == IB_QPT_GSI) {
		unsigned n;

		ret = type == IB_QPT_GSI;
		n = 1 << (ret + 2 * (port - 1));
		spin_lock(&qpt->lock);
		if (qpt->flags & n)
			ret = -EINVAL;
		else
			qpt->flags |= n;
		spin_unlock(&qpt->lock);
		goto bail;
	}

M
Mike Marciniszyn 已提交
146
	qpn = qpt->last + 2;
147 148
	if (qpn >= QPN_MAX)
		qpn = 2;
149 150
	if (qpt->mask && ((qpn & qpt->mask) >> 1) >= dd->n_krcv_queues)
		qpn = (qpn | qpt->mask) + 2;
151 152 153 154 155
	offset = qpn & BITS_PER_PAGE_MASK;
	map = &qpt->map[qpn / BITS_PER_PAGE];
	max_scan = qpt->nmaps - !offset;
	for (i = 0;;) {
		if (unlikely(!map->page)) {
156
			get_map_page(qpt, map, gfp);
157 158 159 160 161 162 163 164 165
			if (unlikely(!map->page))
				break;
		}
		do {
			if (!test_and_set_bit(offset, map->page)) {
				qpt->last = qpn;
				ret = qpn;
				goto bail;
			}
166 167
			offset = find_next_offset(qpt, map, offset,
				dd->n_krcv_queues);
168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186
			qpn = mk_qpn(qpt, map, offset);
			/*
			 * This test differs from alloc_pidmap().
			 * If find_next_offset() does find a zero
			 * bit, we don't need to check for QPN
			 * wrapping around past our starting QPN.
			 * We just need to be sure we don't loop
			 * forever.
			 */
		} while (offset < BITS_PER_PAGE && qpn < QPN_MAX);
		/*
		 * In order to keep the number of pages allocated to a
		 * minimum, we scan the all existing pages before increasing
		 * the size of the bitmap table.
		 */
		if (++i > max_scan) {
			if (qpt->nmaps == QPNMAP_ENTRIES)
				break;
			map = &qpt->map[qpt->nmaps++];
187
			offset = 0;
188 189
		} else if (map < &qpt->map[qpt->nmaps]) {
			++map;
190
			offset = 0;
191 192
		} else {
			map = &qpt->map[0];
193
			offset = 2;
194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212
		}
		qpn = mk_qpn(qpt, map, offset);
	}

	ret = -ENOMEM;

bail:
	return ret;
}

static void free_qpn(struct qib_qpn_table *qpt, u32 qpn)
{
	struct qpn_map *map;

	map = qpt->map + qpn / BITS_PER_PAGE;
	if (map->page)
		clear_bit(qpn & BITS_PER_PAGE_MASK, map->page);
}

M
Mike Marciniszyn 已提交
213 214 215 216 217 218 219
static inline unsigned qpn_hash(struct qib_ibdev *dev, u32 qpn)
{
	return jhash_1word(qpn, dev->qp_rnd) &
		(dev->qp_table_size - 1);
}


220 221 222 223
/*
 * Put the QP into the hash table.
 * The hash table holds a reference to the QP.
 */
224
static void insert_qp(struct qib_ibdev *dev, struct rvt_qp *qp)
225 226 227
{
	struct qib_ibport *ibp = to_iport(qp->ibqp.device, qp->port_num);
	unsigned long flags;
M
Mike Marciniszyn 已提交
228
	unsigned n = qpn_hash(dev, qp->ibqp.qp_num);
229

M
Mike Marciniszyn 已提交
230
	atomic_inc(&qp->refcount);
231
	spin_lock_irqsave(&dev->qpt_lock, flags);
232 233

	if (qp->ibqp.qp_num == 0)
234
		rcu_assign_pointer(ibp->rvp.qp[0], qp);
235
	else if (qp->ibqp.qp_num == 1)
236
		rcu_assign_pointer(ibp->rvp.qp[1], qp);
237 238
	else {
		qp->next = dev->qp_table[n];
M
Mike Marciniszyn 已提交
239
		rcu_assign_pointer(dev->qp_table[n], qp);
240 241 242 243 244 245 246 247 248
	}

	spin_unlock_irqrestore(&dev->qpt_lock, flags);
}

/*
 * Remove the QP from the table so it can't be found asynchronously by
 * the receive interrupt routine.
 */
249
static void remove_qp(struct qib_ibdev *dev, struct rvt_qp *qp)
250 251
{
	struct qib_ibport *ibp = to_iport(qp->ibqp.device, qp->port_num);
M
Mike Marciniszyn 已提交
252
	unsigned n = qpn_hash(dev, qp->ibqp.qp_num);
253
	unsigned long flags;
254
	int removed = 1;
255 256 257

	spin_lock_irqsave(&dev->qpt_lock, flags);

258 259 260 261
	if (rcu_dereference_protected(ibp->rvp.qp[0],
				      lockdep_is_held(&dev->qpt_lock)) == qp) {
		RCU_INIT_POINTER(ibp->rvp.qp[0], NULL);
	} else if (rcu_dereference_protected(ibp->rvp.qp[1],
262
			lockdep_is_held(&dev->qpt_lock)) == qp) {
263
		RCU_INIT_POINTER(ibp->rvp.qp[1], NULL);
M
Mike Marciniszyn 已提交
264
	} else {
265 266
		struct rvt_qp *q;
		struct rvt_qp __rcu **qpp;
M
Mike Marciniszyn 已提交
267

268
		removed = 0;
M
Mike Marciniszyn 已提交
269
		qpp = &dev->qp_table[n];
270 271 272
		for (; (q = rcu_dereference_protected(*qpp,
				lockdep_is_held(&dev->qpt_lock))) != NULL;
				qpp = &q->next)
273
			if (q == qp) {
274
				RCU_INIT_POINTER(*qpp,
275 276
					rcu_dereference_protected(qp->next,
					 lockdep_is_held(&dev->qpt_lock)));
277
				removed = 1;
278 279
				break;
			}
M
Mike Marciniszyn 已提交
280
	}
281 282

	spin_unlock_irqrestore(&dev->qpt_lock, flags);
283 284 285 286
	if (removed) {
		synchronize_rcu();
		atomic_dec(&qp->refcount);
	}
287 288 289 290 291 292 293 294 295 296 297 298 299
}

/**
 * qib_free_all_qps - check for QPs still in use
 * @qpt: the QP table to empty
 *
 * There should not be any QPs still in use.
 * Free memory for table.
 */
unsigned qib_free_all_qps(struct qib_devdata *dd)
{
	struct qib_ibdev *dev = &dd->verbs_dev;
	unsigned long flags;
300
	struct rvt_qp *qp;
301 302 303 304 305 306 307
	unsigned n, qp_inuse = 0;

	for (n = 0; n < dd->num_pports; n++) {
		struct qib_ibport *ibp = &dd->pport[n].ibport_data;

		if (!qib_mcast_tree_empty(ibp))
			qp_inuse++;
M
Mike Marciniszyn 已提交
308
		rcu_read_lock();
309
		if (rcu_dereference(ibp->rvp.qp[0]))
310
			qp_inuse++;
311
		if (rcu_dereference(ibp->rvp.qp[1]))
312
			qp_inuse++;
M
Mike Marciniszyn 已提交
313
		rcu_read_unlock();
314 315 316 317
	}

	spin_lock_irqsave(&dev->qpt_lock, flags);
	for (n = 0; n < dev->qp_table_size; n++) {
318 319
		qp = rcu_dereference_protected(dev->qp_table[n],
			lockdep_is_held(&dev->qpt_lock));
320
		RCU_INIT_POINTER(dev->qp_table[n], NULL);
321

322 323
		for (; qp; qp = rcu_dereference_protected(qp->next,
					lockdep_is_held(&dev->qpt_lock)))
324 325 326
			qp_inuse++;
	}
	spin_unlock_irqrestore(&dev->qpt_lock, flags);
M
Mike Marciniszyn 已提交
327
	synchronize_rcu();
328 329 330 331 332 333 334 335 336 337 338 339

	return qp_inuse;
}

/**
 * qib_lookup_qpn - return the QP with the given QPN
 * @qpt: the QP table
 * @qpn: the QP number to look up
 *
 * The caller is responsible for decrementing the QP reference count
 * when done.
 */
340
struct rvt_qp *qib_lookup_qpn(struct qib_ibport *ibp, u32 qpn)
341
{
342
	struct rvt_qp *qp = NULL;
343

344
	rcu_read_lock();
M
Mike Marciniszyn 已提交
345 346
	if (unlikely(qpn <= 1)) {
		if (qpn == 0)
347
			qp = rcu_dereference(ibp->rvp.qp[0]);
M
Mike Marciniszyn 已提交
348
		else
349
			qp = rcu_dereference(ibp->rvp.qp[1]);
350 351
		if (qp)
			atomic_inc(&qp->refcount);
M
Mike Marciniszyn 已提交
352 353 354
	} else {
		struct qib_ibdev *dev = &ppd_from_ibp(ibp)->dd->verbs_dev;
		unsigned n = qpn_hash(dev, qpn);
355

356 357
		for (qp = rcu_dereference(dev->qp_table[n]); qp;
			qp = rcu_dereference(qp->next))
358 359
			if (qp->ibqp.qp_num == qpn) {
				atomic_inc(&qp->refcount);
360
				break;
361
			}
M
Mike Marciniszyn 已提交
362 363
	}
	rcu_read_unlock();
364 365 366 367 368 369 370 371
	return qp;
}

/**
 * qib_reset_qp - initialize the QP state to the reset state
 * @qp: the QP to reset
 * @type: the QP type
 */
372
static void qib_reset_qp(struct rvt_qp *qp, enum ib_qp_type type)
373
{
374
	struct qib_qp_priv *priv = qp->priv;
375 376 377
	qp->remote_qpn = 0;
	qp->qkey = 0;
	qp->qp_access_flags = 0;
378
	atomic_set(&priv->s_dma_busy, 0);
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
	qp->s_flags &= QIB_S_SIGNAL_REQ_WR;
	qp->s_hdrwords = 0;
	qp->s_wqe = NULL;
	qp->s_draining = 0;
	qp->s_next_psn = 0;
	qp->s_last_psn = 0;
	qp->s_sending_psn = 0;
	qp->s_sending_hpsn = 0;
	qp->s_psn = 0;
	qp->r_psn = 0;
	qp->r_msn = 0;
	if (type == IB_QPT_RC) {
		qp->s_state = IB_OPCODE_RC_SEND_LAST;
		qp->r_state = IB_OPCODE_RC_SEND_LAST;
	} else {
		qp->s_state = IB_OPCODE_UC_SEND_LAST;
		qp->r_state = IB_OPCODE_UC_SEND_LAST;
	}
	qp->s_ack_state = IB_OPCODE_RC_ACKNOWLEDGE;
	qp->r_nak_state = 0;
	qp->r_aflags = 0;
	qp->r_flags = 0;
	qp->s_head = 0;
	qp->s_tail = 0;
	qp->s_cur = 0;
	qp->s_acked = 0;
	qp->s_last = 0;
	qp->s_ssn = 1;
	qp->s_lsn = 0;
	qp->s_mig_state = IB_MIG_MIGRATED;
	memset(qp->s_ack_queue, 0, sizeof(qp->s_ack_queue));
	qp->r_head_ack_queue = 0;
	qp->s_tail_ack_queue = 0;
	qp->s_num_rd_atomic = 0;
	if (qp->r_rq.wq) {
		qp->r_rq.wq->head = 0;
		qp->r_rq.wq->tail = 0;
	}
	qp->r_sge.num_sge = 0;
}

420
static void clear_mr_refs(struct rvt_qp *qp, int clr_sends)
421 422 423 424
{
	unsigned n;

	if (test_and_clear_bit(QIB_R_REWIND_SGE, &qp->r_aflags))
425
		qib_put_ss(&qp->s_rdma_read_sge);
426

427
	qib_put_ss(&qp->r_sge);
428 429 430

	if (clr_sends) {
		while (qp->s_last != qp->s_head) {
431
			struct rvt_swqe *wqe = get_swqe_ptr(qp, qp->s_last);
432 433 434
			unsigned i;

			for (i = 0; i < wqe->wr.num_sge; i++) {
435
				struct rvt_sge *sge = &wqe->sg_list[i];
436

437
				rvt_put_mr(sge->mr);
438 439 440 441
			}
			if (qp->ibqp.qp_type == IB_QPT_UD ||
			    qp->ibqp.qp_type == IB_QPT_SMI ||
			    qp->ibqp.qp_type == IB_QPT_GSI)
442 443
				atomic_dec(
				 &ibah_to_rvtah(wqe->ud_wr.ah)->refcount);
444 445 446 447
			if (++qp->s_last >= qp->s_size)
				qp->s_last = 0;
		}
		if (qp->s_rdma_mr) {
448
			rvt_put_mr(qp->s_rdma_mr);
449 450 451 452 453 454 455 456
			qp->s_rdma_mr = NULL;
		}
	}

	if (qp->ibqp.qp_type != IB_QPT_RC)
		return;

	for (n = 0; n < ARRAY_SIZE(qp->s_ack_queue); n++) {
457
		struct rvt_ack_entry *e = &qp->s_ack_queue[n];
458 459 460

		if (e->opcode == IB_OPCODE_RC_RDMA_READ_REQUEST &&
		    e->rdma_sge.mr) {
461
			rvt_put_mr(e->rdma_sge.mr);
462 463 464 465 466 467 468 469 470 471 472 473
			e->rdma_sge.mr = NULL;
		}
	}
}

/**
 * qib_error_qp - put a QP into the error state
 * @qp: the QP to put into the error state
 * @err: the receive completion error to signal if a RWQE is active
 *
 * Flushes both send and receive work queues.
 * Returns true if last WQE event should be generated.
474
 * The QP r_lock and s_lock should be held and interrupts disabled.
475 476
 * If we are already in error state, just return.
 */
477
int qib_error_qp(struct rvt_qp *qp, enum ib_wc_status err)
478
{
479
	struct qib_qp_priv *priv = qp->priv;
480 481 482 483 484 485 486 487 488 489 490 491 492
	struct qib_ibdev *dev = to_idev(qp->ibqp.device);
	struct ib_wc wc;
	int ret = 0;

	if (qp->state == IB_QPS_ERR || qp->state == IB_QPS_RESET)
		goto bail;

	qp->state = IB_QPS_ERR;

	if (qp->s_flags & (QIB_S_TIMER | QIB_S_WAIT_RNR)) {
		qp->s_flags &= ~(QIB_S_TIMER | QIB_S_WAIT_RNR);
		del_timer(&qp->s_timer);
	}
493 494 495 496

	if (qp->s_flags & QIB_S_ANY_WAIT_SEND)
		qp->s_flags &= ~QIB_S_ANY_WAIT_SEND;

497
	spin_lock(&dev->pending_lock);
498
	if (!list_empty(&priv->iowait) && !(qp->s_flags & QIB_S_BUSY)) {
499
		qp->s_flags &= ~QIB_S_ANY_WAIT_IO;
500
		list_del_init(&priv->iowait);
501 502 503 504 505 506
	}
	spin_unlock(&dev->pending_lock);

	if (!(qp->s_flags & QIB_S_BUSY)) {
		qp->s_hdrwords = 0;
		if (qp->s_rdma_mr) {
507
			rvt_put_mr(qp->s_rdma_mr);
508 509
			qp->s_rdma_mr = NULL;
		}
510 511 512
		if (priv->s_tx) {
			qib_put_txreq(priv->s_tx);
			priv->s_tx = NULL;
513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533
		}
	}

	/* Schedule the sending tasklet to drain the send work queue. */
	if (qp->s_last != qp->s_head)
		qib_schedule_send(qp);

	clear_mr_refs(qp, 0);

	memset(&wc, 0, sizeof(wc));
	wc.qp = &qp->ibqp;
	wc.opcode = IB_WC_RECV;

	if (test_and_clear_bit(QIB_R_WRID_VALID, &qp->r_aflags)) {
		wc.wr_id = qp->r_wr_id;
		wc.status = err;
		qib_cq_enter(to_icq(qp->ibqp.recv_cq), &wc, 1);
	}
	wc.status = IB_WC_WR_FLUSH_ERR;

	if (qp->r_rq.wq) {
534
		struct rvt_rwq *wq;
535 536 537 538 539 540 541 542 543 544 545 546 547 548 549 550 551 552 553 554 555 556 557 558 559 560 561 562 563 564 565 566 567 568 569 570 571 572 573 574 575 576
		u32 head;
		u32 tail;

		spin_lock(&qp->r_rq.lock);

		/* sanity check pointers before trusting them */
		wq = qp->r_rq.wq;
		head = wq->head;
		if (head >= qp->r_rq.size)
			head = 0;
		tail = wq->tail;
		if (tail >= qp->r_rq.size)
			tail = 0;
		while (tail != head) {
			wc.wr_id = get_rwqe_ptr(&qp->r_rq, tail)->wr_id;
			if (++tail >= qp->r_rq.size)
				tail = 0;
			qib_cq_enter(to_icq(qp->ibqp.recv_cq), &wc, 1);
		}
		wq->tail = tail;

		spin_unlock(&qp->r_rq.lock);
	} else if (qp->ibqp.event_handler)
		ret = 1;

bail:
	return ret;
}

/**
 * qib_modify_qp - modify the attributes of a queue pair
 * @ibqp: the queue pair who's attributes we're modifying
 * @attr: the new attributes
 * @attr_mask: the mask of attributes to modify
 * @udata: user data for libibverbs.so
 *
 * Returns 0 on success, otherwise returns an errno.
 */
int qib_modify_qp(struct ib_qp *ibqp, struct ib_qp_attr *attr,
		  int attr_mask, struct ib_udata *udata)
{
	struct qib_ibdev *dev = to_idev(ibqp->device);
577
	struct rvt_qp *qp = to_iqp(ibqp);
578
	struct qib_qp_priv *priv = qp->priv;
579 580 581 582 583 584 585 586 587 588 589 590 591 592 593
	enum ib_qp_state cur_state, new_state;
	struct ib_event ev;
	int lastwqe = 0;
	int mig = 0;
	int ret;
	u32 pmtu = 0; /* for gcc warning only */

	spin_lock_irq(&qp->r_lock);
	spin_lock(&qp->s_lock);

	cur_state = attr_mask & IB_QP_CUR_STATE ?
		attr->cur_qp_state : qp->state;
	new_state = attr_mask & IB_QP_STATE ? attr->qp_state : cur_state;

	if (!ib_modify_qp_is_ok(cur_state, new_state, ibqp->qp_type,
594
				attr_mask, IB_LINK_LAYER_UNSPECIFIED))
595 596 597
		goto inval;

	if (attr_mask & IB_QP_AV) {
598
		if (attr->ah_attr.dlid >= be16_to_cpu(IB_MULTICAST_LID_BASE))
599
			goto inval;
600
		if (rvt_check_ah(qp->ibqp.device, &attr->ah_attr))
601 602 603 604
			goto inval;
	}

	if (attr_mask & IB_QP_ALT_PATH) {
605 606
		if (attr->alt_ah_attr.dlid >=
		    be16_to_cpu(IB_MULTICAST_LID_BASE))
607
			goto inval;
608
		if (rvt_check_ah(qp->ibqp.device, &attr->alt_ah_attr))
609 610 611 612 613 614 615 616 617 618 619 620 621 622 623 624 625 626 627 628 629 630 631 632 633 634 635 636 637 638 639 640 641 642 643 644 645 646 647 648 649 650 651 652 653 654 655 656 657 658 659 660 661 662 663 664 665 666 667 668 669 670 671 672 673 674 675 676 677 678 679 680 681 682 683 684 685 686 687 688 689 690 691 692 693 694 695 696 697 698 699 700 701 702 703 704 705
			goto inval;
		if (attr->alt_pkey_index >= qib_get_npkeys(dd_from_dev(dev)))
			goto inval;
	}

	if (attr_mask & IB_QP_PKEY_INDEX)
		if (attr->pkey_index >= qib_get_npkeys(dd_from_dev(dev)))
			goto inval;

	if (attr_mask & IB_QP_MIN_RNR_TIMER)
		if (attr->min_rnr_timer > 31)
			goto inval;

	if (attr_mask & IB_QP_PORT)
		if (qp->ibqp.qp_type == IB_QPT_SMI ||
		    qp->ibqp.qp_type == IB_QPT_GSI ||
		    attr->port_num == 0 ||
		    attr->port_num > ibqp->device->phys_port_cnt)
			goto inval;

	if (attr_mask & IB_QP_DEST_QPN)
		if (attr->dest_qp_num > QIB_QPN_MASK)
			goto inval;

	if (attr_mask & IB_QP_RETRY_CNT)
		if (attr->retry_cnt > 7)
			goto inval;

	if (attr_mask & IB_QP_RNR_RETRY)
		if (attr->rnr_retry > 7)
			goto inval;

	/*
	 * Don't allow invalid path_mtu values.  OK to set greater
	 * than the active mtu (or even the max_cap, if we have tuned
	 * that to a small mtu.  We'll set qp->path_mtu
	 * to the lesser of requested attribute mtu and active,
	 * for packetizing messages.
	 * Note that the QP port has to be set in INIT and MTU in RTR.
	 */
	if (attr_mask & IB_QP_PATH_MTU) {
		struct qib_devdata *dd = dd_from_dev(dev);
		int mtu, pidx = qp->port_num - 1;

		mtu = ib_mtu_enum_to_int(attr->path_mtu);
		if (mtu == -1)
			goto inval;
		if (mtu > dd->pport[pidx].ibmtu) {
			switch (dd->pport[pidx].ibmtu) {
			case 4096:
				pmtu = IB_MTU_4096;
				break;
			case 2048:
				pmtu = IB_MTU_2048;
				break;
			case 1024:
				pmtu = IB_MTU_1024;
				break;
			case 512:
				pmtu = IB_MTU_512;
				break;
			case 256:
				pmtu = IB_MTU_256;
				break;
			default:
				pmtu = IB_MTU_2048;
			}
		} else
			pmtu = attr->path_mtu;
	}

	if (attr_mask & IB_QP_PATH_MIG_STATE) {
		if (attr->path_mig_state == IB_MIG_REARM) {
			if (qp->s_mig_state == IB_MIG_ARMED)
				goto inval;
			if (new_state != IB_QPS_RTS)
				goto inval;
		} else if (attr->path_mig_state == IB_MIG_MIGRATED) {
			if (qp->s_mig_state == IB_MIG_REARM)
				goto inval;
			if (new_state != IB_QPS_RTS && new_state != IB_QPS_SQD)
				goto inval;
			if (qp->s_mig_state == IB_MIG_ARMED)
				mig = 1;
		} else
			goto inval;
	}

	if (attr_mask & IB_QP_MAX_DEST_RD_ATOMIC)
		if (attr->max_dest_rd_atomic > QIB_MAX_RDMA_ATOMIC)
			goto inval;

	switch (new_state) {
	case IB_QPS_RESET:
		if (qp->state != IB_QPS_RESET) {
			qp->state = IB_QPS_RESET;
			spin_lock(&dev->pending_lock);
706 707
			if (!list_empty(&priv->iowait))
				list_del_init(&priv->iowait);
708 709 710 711 712
			spin_unlock(&dev->pending_lock);
			qp->s_flags &= ~(QIB_S_TIMER | QIB_S_ANY_WAIT);
			spin_unlock(&qp->s_lock);
			spin_unlock_irq(&qp->r_lock);
			/* Stop the sending work queue and retry timer */
713
			cancel_work_sync(&priv->s_work);
714
			del_timer_sync(&qp->s_timer);
715 716 717 718 719
			wait_event(priv->wait_dma,
				   !atomic_read(&priv->s_dma_busy));
			if (priv->s_tx) {
				qib_put_txreq(priv->s_tx);
				priv->s_tx = NULL;
720 721 722 723 724 725 726 727 728 729 730 731 732 733 734 735 736 737 738 739 740 741 742 743 744 745 746 747 748 749 750 751 752 753 754 755 756 757 758 759 760 761 762 763 764 765 766 767 768 769 770 771 772 773 774 775 776 777 778 779 780 781 782 783 784 785 786 787 788 789 790 791 792 793 794 795 796 797
			}
			remove_qp(dev, qp);
			wait_event(qp->wait, !atomic_read(&qp->refcount));
			spin_lock_irq(&qp->r_lock);
			spin_lock(&qp->s_lock);
			clear_mr_refs(qp, 1);
			qib_reset_qp(qp, ibqp->qp_type);
		}
		break;

	case IB_QPS_RTR:
		/* Allow event to retrigger if QP set to RTR more than once */
		qp->r_flags &= ~QIB_R_COMM_EST;
		qp->state = new_state;
		break;

	case IB_QPS_SQD:
		qp->s_draining = qp->s_last != qp->s_cur;
		qp->state = new_state;
		break;

	case IB_QPS_SQE:
		if (qp->ibqp.qp_type == IB_QPT_RC)
			goto inval;
		qp->state = new_state;
		break;

	case IB_QPS_ERR:
		lastwqe = qib_error_qp(qp, IB_WC_WR_FLUSH_ERR);
		break;

	default:
		qp->state = new_state;
		break;
	}

	if (attr_mask & IB_QP_PKEY_INDEX)
		qp->s_pkey_index = attr->pkey_index;

	if (attr_mask & IB_QP_PORT)
		qp->port_num = attr->port_num;

	if (attr_mask & IB_QP_DEST_QPN)
		qp->remote_qpn = attr->dest_qp_num;

	if (attr_mask & IB_QP_SQ_PSN) {
		qp->s_next_psn = attr->sq_psn & QIB_PSN_MASK;
		qp->s_psn = qp->s_next_psn;
		qp->s_sending_psn = qp->s_next_psn;
		qp->s_last_psn = qp->s_next_psn - 1;
		qp->s_sending_hpsn = qp->s_last_psn;
	}

	if (attr_mask & IB_QP_RQ_PSN)
		qp->r_psn = attr->rq_psn & QIB_PSN_MASK;

	if (attr_mask & IB_QP_ACCESS_FLAGS)
		qp->qp_access_flags = attr->qp_access_flags;

	if (attr_mask & IB_QP_AV) {
		qp->remote_ah_attr = attr->ah_attr;
		qp->s_srate = attr->ah_attr.static_rate;
	}

	if (attr_mask & IB_QP_ALT_PATH) {
		qp->alt_ah_attr = attr->alt_ah_attr;
		qp->s_alt_pkey_index = attr->alt_pkey_index;
	}

	if (attr_mask & IB_QP_PATH_MIG_STATE) {
		qp->s_mig_state = attr->path_mig_state;
		if (mig) {
			qp->remote_ah_attr = qp->alt_ah_attr;
			qp->port_num = qp->alt_ah_attr.port_num;
			qp->s_pkey_index = qp->s_alt_pkey_index;
		}
	}

798
	if (attr_mask & IB_QP_PATH_MTU) {
799
		qp->path_mtu = pmtu;
800 801
		qp->pmtu = ib_mtu_enum_to_int(pmtu);
	}
802 803 804 805 806 807 808 809 810 811 812 813 814 815

	if (attr_mask & IB_QP_RETRY_CNT) {
		qp->s_retry_cnt = attr->retry_cnt;
		qp->s_retry = attr->retry_cnt;
	}

	if (attr_mask & IB_QP_RNR_RETRY) {
		qp->s_rnr_retry_cnt = attr->rnr_retry;
		qp->s_rnr_retry = attr->rnr_retry;
	}

	if (attr_mask & IB_QP_MIN_RNR_TIMER)
		qp->r_min_rnr_timer = attr->min_rnr_timer;

816
	if (attr_mask & IB_QP_TIMEOUT) {
817
		qp->timeout = attr->timeout;
818 819 820 821
		qp->timeout_jiffies =
			usecs_to_jiffies((4096UL * (1UL << qp->timeout)) /
				1000UL);
	}
822 823 824 825 826 827 828 829 830 831 832 833 834 835 836 837 838 839 840 841 842 843 844 845 846 847 848 849 850 851 852 853 854 855 856 857 858 859 860 861 862 863 864

	if (attr_mask & IB_QP_QKEY)
		qp->qkey = attr->qkey;

	if (attr_mask & IB_QP_MAX_DEST_RD_ATOMIC)
		qp->r_max_rd_atomic = attr->max_dest_rd_atomic;

	if (attr_mask & IB_QP_MAX_QP_RD_ATOMIC)
		qp->s_max_rd_atomic = attr->max_rd_atomic;

	spin_unlock(&qp->s_lock);
	spin_unlock_irq(&qp->r_lock);

	if (cur_state == IB_QPS_RESET && new_state == IB_QPS_INIT)
		insert_qp(dev, qp);

	if (lastwqe) {
		ev.device = qp->ibqp.device;
		ev.element.qp = &qp->ibqp;
		ev.event = IB_EVENT_QP_LAST_WQE_REACHED;
		qp->ibqp.event_handler(&ev, qp->ibqp.qp_context);
	}
	if (mig) {
		ev.device = qp->ibqp.device;
		ev.element.qp = &qp->ibqp;
		ev.event = IB_EVENT_PATH_MIG;
		qp->ibqp.event_handler(&ev, qp->ibqp.qp_context);
	}
	ret = 0;
	goto bail;

inval:
	spin_unlock(&qp->s_lock);
	spin_unlock_irq(&qp->r_lock);
	ret = -EINVAL;

bail:
	return ret;
}

int qib_query_qp(struct ib_qp *ibqp, struct ib_qp_attr *attr,
		 int attr_mask, struct ib_qp_init_attr *init_attr)
{
865
	struct rvt_qp *qp = to_iqp(ibqp);
866 867 868 869 870 871 872 873 874 875 876 877 878 879 880 881 882 883 884 885 886 887 888 889 890 891 892 893 894 895 896 897 898 899 900 901 902 903 904 905 906 907 908 909 910 911 912 913 914 915 916 917

	attr->qp_state = qp->state;
	attr->cur_qp_state = attr->qp_state;
	attr->path_mtu = qp->path_mtu;
	attr->path_mig_state = qp->s_mig_state;
	attr->qkey = qp->qkey;
	attr->rq_psn = qp->r_psn & QIB_PSN_MASK;
	attr->sq_psn = qp->s_next_psn & QIB_PSN_MASK;
	attr->dest_qp_num = qp->remote_qpn;
	attr->qp_access_flags = qp->qp_access_flags;
	attr->cap.max_send_wr = qp->s_size - 1;
	attr->cap.max_recv_wr = qp->ibqp.srq ? 0 : qp->r_rq.size - 1;
	attr->cap.max_send_sge = qp->s_max_sge;
	attr->cap.max_recv_sge = qp->r_rq.max_sge;
	attr->cap.max_inline_data = 0;
	attr->ah_attr = qp->remote_ah_attr;
	attr->alt_ah_attr = qp->alt_ah_attr;
	attr->pkey_index = qp->s_pkey_index;
	attr->alt_pkey_index = qp->s_alt_pkey_index;
	attr->en_sqd_async_notify = 0;
	attr->sq_draining = qp->s_draining;
	attr->max_rd_atomic = qp->s_max_rd_atomic;
	attr->max_dest_rd_atomic = qp->r_max_rd_atomic;
	attr->min_rnr_timer = qp->r_min_rnr_timer;
	attr->port_num = qp->port_num;
	attr->timeout = qp->timeout;
	attr->retry_cnt = qp->s_retry_cnt;
	attr->rnr_retry = qp->s_rnr_retry_cnt;
	attr->alt_port_num = qp->alt_ah_attr.port_num;
	attr->alt_timeout = qp->alt_timeout;

	init_attr->event_handler = qp->ibqp.event_handler;
	init_attr->qp_context = qp->ibqp.qp_context;
	init_attr->send_cq = qp->ibqp.send_cq;
	init_attr->recv_cq = qp->ibqp.recv_cq;
	init_attr->srq = qp->ibqp.srq;
	init_attr->cap = attr->cap;
	if (qp->s_flags & QIB_S_SIGNAL_REQ_WR)
		init_attr->sq_sig_type = IB_SIGNAL_REQ_WR;
	else
		init_attr->sq_sig_type = IB_SIGNAL_ALL_WR;
	init_attr->qp_type = qp->ibqp.qp_type;
	init_attr->port_num = qp->port_num;
	return 0;
}

/**
 * qib_compute_aeth - compute the AETH (syndrome + MSN)
 * @qp: the queue pair to compute the AETH for
 *
 * Returns the AETH.
 */
918
__be32 qib_compute_aeth(struct rvt_qp *qp)
919 920 921 922 923 924 925 926 927 928 929 930
{
	u32 aeth = qp->r_msn & QIB_MSN_MASK;

	if (qp->ibqp.srq) {
		/*
		 * Shared receive queues don't generate credits.
		 * Set the credit field to the invalid value.
		 */
		aeth |= QIB_AETH_CREDIT_INVAL << QIB_AETH_CREDIT_SHIFT;
	} else {
		u32 min, max, x;
		u32 credits;
931
		struct rvt_rwq *wq = qp->r_rq.wq;
932 933 934 935 936 937 938 939 940 941 942 943 944 945 946 947 948 949 950 951 952 953 954 955 956 957 958 959 960 961 962 963 964 965 966 967 968 969 970 971 972 973 974 975 976 977 978 979 980 981 982 983 984 985
		u32 head;
		u32 tail;

		/* sanity check pointers before trusting them */
		head = wq->head;
		if (head >= qp->r_rq.size)
			head = 0;
		tail = wq->tail;
		if (tail >= qp->r_rq.size)
			tail = 0;
		/*
		 * Compute the number of credits available (RWQEs).
		 * XXX Not holding the r_rq.lock here so there is a small
		 * chance that the pair of reads are not atomic.
		 */
		credits = head - tail;
		if ((int)credits < 0)
			credits += qp->r_rq.size;
		/*
		 * Binary search the credit table to find the code to
		 * use.
		 */
		min = 0;
		max = 31;
		for (;;) {
			x = (min + max) / 2;
			if (credit_table[x] == credits)
				break;
			if (credit_table[x] > credits)
				max = x;
			else if (min == x)
				break;
			else
				min = x;
		}
		aeth |= x << QIB_AETH_CREDIT_SHIFT;
	}
	return cpu_to_be32(aeth);
}

/**
 * qib_create_qp - create a queue pair for a device
 * @ibpd: the protection domain who's device we create the queue pair for
 * @init_attr: the attributes of the queue pair
 * @udata: user data for libibverbs.so
 *
 * Returns the queue pair on success, otherwise returns an errno.
 *
 * Called by the ib_create_qp() core verbs function.
 */
struct ib_qp *qib_create_qp(struct ib_pd *ibpd,
			    struct ib_qp_init_attr *init_attr,
			    struct ib_udata *udata)
{
986
	struct rvt_qp *qp;
987
	int err;
988
	struct rvt_swqe *swq = NULL;
989 990 991 992 993
	struct qib_ibdev *dev;
	struct qib_devdata *dd;
	size_t sz;
	size_t sg_list_sz;
	struct ib_qp *ret;
994
	gfp_t gfp;
995
	struct qib_qp_priv *priv;
996 997

	if (init_attr->cap.max_send_sge > ib_qib_max_sges ||
998
	    init_attr->cap.max_send_wr > ib_qib_max_qp_wrs ||
999 1000 1001 1002 1003 1004 1005 1006 1007 1008
	    init_attr->create_flags & ~(IB_QP_CREATE_USE_GFP_NOIO))
		return ERR_PTR(-EINVAL);

	/* GFP_NOIO is applicable in RC QPs only */
	if (init_attr->create_flags & IB_QP_CREATE_USE_GFP_NOIO &&
	    init_attr->qp_type != IB_QPT_RC)
		return ERR_PTR(-EINVAL);

	gfp = init_attr->create_flags & IB_QP_CREATE_USE_GFP_NOIO ?
			GFP_NOIO : GFP_KERNEL;
1009 1010 1011 1012 1013 1014 1015 1016 1017 1018 1019 1020 1021 1022 1023 1024 1025 1026 1027 1028 1029 1030 1031 1032 1033 1034 1035 1036

	/* Check receive queue parameters if no SRQ is specified. */
	if (!init_attr->srq) {
		if (init_attr->cap.max_recv_sge > ib_qib_max_sges ||
		    init_attr->cap.max_recv_wr > ib_qib_max_qp_wrs) {
			ret = ERR_PTR(-EINVAL);
			goto bail;
		}
		if (init_attr->cap.max_send_sge +
		    init_attr->cap.max_send_wr +
		    init_attr->cap.max_recv_sge +
		    init_attr->cap.max_recv_wr == 0) {
			ret = ERR_PTR(-EINVAL);
			goto bail;
		}
	}

	switch (init_attr->qp_type) {
	case IB_QPT_SMI:
	case IB_QPT_GSI:
		if (init_attr->port_num == 0 ||
		    init_attr->port_num > ibpd->device->phys_port_cnt) {
			ret = ERR_PTR(-EINVAL);
			goto bail;
		}
	case IB_QPT_UC:
	case IB_QPT_RC:
	case IB_QPT_UD:
1037
		sz = sizeof(struct rvt_sge) *
1038
			init_attr->cap.max_send_sge +
1039
			sizeof(struct rvt_swqe);
1040 1041
		swq = __vmalloc((init_attr->cap.max_send_wr + 1) * sz,
				gfp, PAGE_KERNEL);
1042 1043 1044 1045 1046 1047 1048
		if (swq == NULL) {
			ret = ERR_PTR(-ENOMEM);
			goto bail;
		}
		sz = sizeof(*qp);
		sg_list_sz = 0;
		if (init_attr->srq) {
D
Dennis Dalessandro 已提交
1049
			struct rvt_srq *srq = ibsrq_to_rvtsrq(init_attr->srq);
1050 1051 1052 1053 1054 1055 1056

			if (srq->rq.max_sge > 1)
				sg_list_sz = sizeof(*qp->r_sg_list) *
					(srq->rq.max_sge - 1);
		} else if (init_attr->cap.max_recv_sge > 1)
			sg_list_sz = sizeof(*qp->r_sg_list) *
				(init_attr->cap.max_recv_sge - 1);
1057
		qp = kzalloc(sz + sg_list_sz, gfp);
1058 1059 1060 1061
		if (!qp) {
			ret = ERR_PTR(-ENOMEM);
			goto bail_swq;
		}
M
Mike Marciniszyn 已提交
1062
		RCU_INIT_POINTER(qp->next, NULL);
1063 1064 1065 1066 1067 1068 1069 1070
		priv = kzalloc(sizeof(*priv), gfp);
		if (!priv) {
			ret = ERR_PTR(-ENOMEM);
			goto bail_qp_hdr;
		}
		priv->owner = qp;
		priv->s_hdr = kzalloc(sizeof(*priv->s_hdr), gfp);
		if (!priv->s_hdr) {
1071 1072 1073
			ret = ERR_PTR(-ENOMEM);
			goto bail_qp;
		}
1074
		qp->priv = priv;
1075 1076 1077
		qp->timeout_jiffies =
			usecs_to_jiffies((4096UL * (1UL << qp->timeout)) /
				1000UL);
1078 1079 1080 1081 1082 1083
		if (init_attr->srq)
			sz = 0;
		else {
			qp->r_rq.size = init_attr->cap.max_recv_wr + 1;
			qp->r_rq.max_sge = init_attr->cap.max_recv_sge;
			sz = (sizeof(struct ib_sge) * qp->r_rq.max_sge) +
1084
				sizeof(struct rvt_rwqe);
1085 1086
			if (gfp != GFP_NOIO)
				qp->r_rq.wq = vmalloc_user(
1087
						sizeof(struct rvt_rwq) +
1088 1089 1090
						qp->r_rq.size * sz);
			else
				qp->r_rq.wq = __vmalloc(
1091
						sizeof(struct rvt_rwq) +
1092 1093 1094
						qp->r_rq.size * sz,
						gfp, PAGE_KERNEL);

1095 1096 1097 1098 1099 1100 1101 1102 1103 1104 1105 1106 1107 1108 1109
			if (!qp->r_rq.wq) {
				ret = ERR_PTR(-ENOMEM);
				goto bail_qp;
			}
		}

		/*
		 * ib_create_qp() will initialize qp->ibqp
		 * except for qp->ibqp.qp_num.
		 */
		spin_lock_init(&qp->r_lock);
		spin_lock_init(&qp->s_lock);
		spin_lock_init(&qp->r_rq.lock);
		atomic_set(&qp->refcount, 0);
		init_waitqueue_head(&qp->wait);
1110
		init_waitqueue_head(&priv->wait_dma);
1111 1112
		init_timer(&qp->s_timer);
		qp->s_timer.data = (unsigned long)qp;
1113 1114
		INIT_WORK(&priv->s_work, qib_do_send);
		INIT_LIST_HEAD(&priv->iowait);
1115 1116 1117 1118 1119 1120 1121 1122 1123 1124
		INIT_LIST_HEAD(&qp->rspwait);
		qp->state = IB_QPS_RESET;
		qp->s_wq = swq;
		qp->s_size = init_attr->cap.max_send_wr + 1;
		qp->s_max_sge = init_attr->cap.max_send_sge;
		if (init_attr->sq_sig_type == IB_SIGNAL_REQ_WR)
			qp->s_flags = QIB_S_SIGNAL_REQ_WR;
		dev = to_idev(ibpd->device);
		dd = dd_from_dev(dev);
		err = alloc_qpn(dd, &dev->qpn_table, init_attr->qp_type,
1125
				init_attr->port_num, gfp);
1126 1127 1128 1129 1130 1131 1132 1133 1134 1135 1136 1137 1138 1139 1140 1141 1142 1143 1144 1145 1146 1147 1148 1149 1150 1151 1152 1153 1154 1155 1156 1157 1158
		if (err < 0) {
			ret = ERR_PTR(err);
			vfree(qp->r_rq.wq);
			goto bail_qp;
		}
		qp->ibqp.qp_num = err;
		qp->port_num = init_attr->port_num;
		qib_reset_qp(qp, init_attr->qp_type);
		break;

	default:
		/* Don't support raw QPs */
		ret = ERR_PTR(-ENOSYS);
		goto bail;
	}

	init_attr->cap.max_inline_data = 0;

	/*
	 * Return the address of the RWQ as the offset to mmap.
	 * See qib_mmap() for details.
	 */
	if (udata && udata->outlen >= sizeof(__u64)) {
		if (!qp->r_rq.wq) {
			__u64 offset = 0;

			err = ib_copy_to_udata(udata, &offset,
					       sizeof(offset));
			if (err) {
				ret = ERR_PTR(err);
				goto bail_ip;
			}
		} else {
1159
			u32 s = sizeof(struct rvt_rwq) + qp->r_rq.size * sz;
1160 1161 1162 1163 1164 1165 1166 1167 1168 1169 1170 1171 1172 1173 1174 1175 1176 1177 1178 1179 1180 1181 1182 1183 1184 1185 1186 1187 1188 1189 1190 1191 1192 1193 1194 1195 1196 1197 1198 1199 1200 1201 1202 1203

			qp->ip = qib_create_mmap_info(dev, s,
						      ibpd->uobject->context,
						      qp->r_rq.wq);
			if (!qp->ip) {
				ret = ERR_PTR(-ENOMEM);
				goto bail_ip;
			}

			err = ib_copy_to_udata(udata, &(qp->ip->offset),
					       sizeof(qp->ip->offset));
			if (err) {
				ret = ERR_PTR(err);
				goto bail_ip;
			}
		}
	}

	spin_lock(&dev->n_qps_lock);
	if (dev->n_qps_allocated == ib_qib_max_qps) {
		spin_unlock(&dev->n_qps_lock);
		ret = ERR_PTR(-ENOMEM);
		goto bail_ip;
	}

	dev->n_qps_allocated++;
	spin_unlock(&dev->n_qps_lock);

	if (qp->ip) {
		spin_lock_irq(&dev->pending_lock);
		list_add(&qp->ip->pending_mmaps, &dev->pending_mmaps);
		spin_unlock_irq(&dev->pending_lock);
	}

	ret = &qp->ibqp;
	goto bail;

bail_ip:
	if (qp->ip)
		kref_put(&qp->ip->ref, qib_release_mmap_info);
	else
		vfree(qp->r_rq.wq);
	free_qpn(&dev->qpn_table, qp->ibqp.qp_num);
bail_qp:
1204 1205 1206
	kfree(priv->s_hdr);
	kfree(priv);
bail_qp_hdr:
1207 1208 1209 1210 1211 1212 1213 1214 1215 1216 1217 1218 1219 1220 1221 1222 1223 1224
	kfree(qp);
bail_swq:
	vfree(swq);
bail:
	return ret;
}

/**
 * qib_destroy_qp - destroy a queue pair
 * @ibqp: the queue pair to destroy
 *
 * Returns 0 on success.
 *
 * Note that this can be called while the QP is actively sending or
 * receiving!
 */
int qib_destroy_qp(struct ib_qp *ibqp)
{
1225
	struct rvt_qp *qp = to_iqp(ibqp);
1226
	struct qib_ibdev *dev = to_idev(ibqp->device);
1227
	struct qib_qp_priv *priv = qp->priv;
1228 1229 1230 1231 1232 1233

	/* Make sure HW and driver activity is stopped. */
	spin_lock_irq(&qp->s_lock);
	if (qp->state != IB_QPS_RESET) {
		qp->state = IB_QPS_RESET;
		spin_lock(&dev->pending_lock);
1234 1235
		if (!list_empty(&priv->iowait))
			list_del_init(&priv->iowait);
1236 1237 1238
		spin_unlock(&dev->pending_lock);
		qp->s_flags &= ~(QIB_S_TIMER | QIB_S_ANY_WAIT);
		spin_unlock_irq(&qp->s_lock);
1239
		cancel_work_sync(&priv->s_work);
1240
		del_timer_sync(&qp->s_timer);
1241 1242 1243 1244
		wait_event(priv->wait_dma, !atomic_read(&priv->s_dma_busy));
		if (priv->s_tx) {
			qib_put_txreq(priv->s_tx);
			priv->s_tx = NULL;
1245 1246 1247 1248 1249 1250 1251 1252 1253 1254 1255 1256 1257 1258 1259 1260 1261 1262
		}
		remove_qp(dev, qp);
		wait_event(qp->wait, !atomic_read(&qp->refcount));
		clear_mr_refs(qp, 1);
	} else
		spin_unlock_irq(&qp->s_lock);

	/* all user's cleaned up, mark it available */
	free_qpn(&dev->qpn_table, qp->ibqp.qp_num);
	spin_lock(&dev->n_qps_lock);
	dev->n_qps_allocated--;
	spin_unlock(&dev->n_qps_lock);

	if (qp->ip)
		kref_put(&qp->ip->ref, qib_release_mmap_info);
	else
		vfree(qp->r_rq.wq);
	vfree(qp->s_wq);
1263 1264
	kfree(priv->s_hdr);
	kfree(priv);
1265 1266 1267 1268 1269 1270 1271 1272 1273 1274 1275 1276 1277 1278 1279 1280 1281 1282 1283 1284 1285 1286 1287 1288 1289 1290 1291 1292 1293 1294 1295 1296 1297 1298 1299 1300
	kfree(qp);
	return 0;
}

/**
 * qib_init_qpn_table - initialize the QP number table for a device
 * @qpt: the QPN table
 */
void qib_init_qpn_table(struct qib_devdata *dd, struct qib_qpn_table *qpt)
{
	spin_lock_init(&qpt->lock);
	qpt->last = 1;          /* start with QPN 2 */
	qpt->nmaps = 1;
	qpt->mask = dd->qpn_mask;
}

/**
 * qib_free_qpn_table - free the QP number table for a device
 * @qpt: the QPN table
 */
void qib_free_qpn_table(struct qib_qpn_table *qpt)
{
	int i;

	for (i = 0; i < ARRAY_SIZE(qpt->map); i++)
		if (qpt->map[i].page)
			free_page((unsigned long) qpt->map[i].page);
}

/**
 * qib_get_credit - flush the send work queue of a QP
 * @qp: the qp who's send work queue to flush
 * @aeth: the Acknowledge Extended Transport Header
 *
 * The QP s_lock should be held.
 */
1301
void qib_get_credit(struct rvt_qp *qp, u32 aeth)
1302 1303 1304 1305 1306 1307 1308 1309 1310 1311 1312 1313 1314 1315 1316 1317 1318 1319 1320 1321 1322 1323 1324 1325 1326 1327 1328 1329
{
	u32 credit = (aeth >> QIB_AETH_CREDIT_SHIFT) & QIB_AETH_CREDIT_MASK;

	/*
	 * If the credit is invalid, we can send
	 * as many packets as we like.  Otherwise, we have to
	 * honor the credit field.
	 */
	if (credit == QIB_AETH_CREDIT_INVAL) {
		if (!(qp->s_flags & QIB_S_UNLIMITED_CREDIT)) {
			qp->s_flags |= QIB_S_UNLIMITED_CREDIT;
			if (qp->s_flags & QIB_S_WAIT_SSN_CREDIT) {
				qp->s_flags &= ~QIB_S_WAIT_SSN_CREDIT;
				qib_schedule_send(qp);
			}
		}
	} else if (!(qp->s_flags & QIB_S_UNLIMITED_CREDIT)) {
		/* Compute new LSN (i.e., MSN + credit) */
		credit = (aeth + credit_table[credit]) & QIB_MSN_MASK;
		if (qib_cmp24(credit, qp->s_lsn) > 0) {
			qp->s_lsn = credit;
			if (qp->s_flags & QIB_S_WAIT_SSN_CREDIT) {
				qp->s_flags &= ~QIB_S_WAIT_SSN_CREDIT;
				qib_schedule_send(qp);
			}
		}
	}
}
1330 1331 1332 1333 1334

#ifdef CONFIG_DEBUG_FS

struct qib_qp_iter {
	struct qib_ibdev *dev;
1335
	struct rvt_qp *qp;
1336 1337 1338 1339 1340 1341 1342 1343 1344 1345 1346 1347 1348 1349 1350 1351 1352 1353 1354 1355 1356 1357 1358 1359 1360
	int n;
};

struct qib_qp_iter *qib_qp_iter_init(struct qib_ibdev *dev)
{
	struct qib_qp_iter *iter;

	iter = kzalloc(sizeof(*iter), GFP_KERNEL);
	if (!iter)
		return NULL;

	iter->dev = dev;
	if (qib_qp_iter_next(iter)) {
		kfree(iter);
		return NULL;
	}

	return iter;
}

int qib_qp_iter_next(struct qib_qp_iter *iter)
{
	struct qib_ibdev *dev = iter->dev;
	int n = iter->n;
	int ret = 1;
1361 1362
	struct rvt_qp *pqp = iter->qp;
	struct rvt_qp *qp;
1363 1364 1365 1366 1367 1368 1369 1370 1371 1372 1373 1374 1375 1376 1377 1378 1379 1380 1381 1382 1383 1384

	for (; n < dev->qp_table_size; n++) {
		if (pqp)
			qp = rcu_dereference(pqp->next);
		else
			qp = rcu_dereference(dev->qp_table[n]);
		pqp = qp;
		if (qp) {
			iter->qp = qp;
			iter->n = n;
			return 0;
		}
	}
	return ret;
}

static const char * const qp_type_str[] = {
	"SMI", "GSI", "RC", "UC", "UD",
};

void qib_qp_iter_print(struct seq_file *s, struct qib_qp_iter *iter)
{
1385 1386
	struct rvt_swqe *wqe;
	struct rvt_qp *qp = iter->qp;
1387
	struct qib_qp_priv *priv = qp->priv;
1388 1389 1390 1391 1392 1393 1394 1395 1396 1397 1398

	wqe = get_swqe_ptr(qp, qp->s_last);
	seq_printf(s,
		   "N %d QP%u %s %u %u %u f=%x %u %u %u %u %u PSN %x %x %x %x %x (%u %u %u %u %u %u) QP%u LID %x\n",
		   iter->n,
		   qp->ibqp.qp_num,
		   qp_type_str[qp->ibqp.qp_type],
		   qp->state,
		   wqe->wr.opcode,
		   qp->s_hdrwords,
		   qp->s_flags,
1399 1400
		   atomic_read(&priv->s_dma_busy),
		   !list_empty(&priv->iowait),
1401 1402 1403 1404 1405 1406 1407 1408 1409 1410 1411 1412 1413
		   qp->timeout,
		   wqe->ssn,
		   qp->s_lsn,
		   qp->s_last_psn,
		   qp->s_psn, qp->s_next_psn,
		   qp->s_sending_psn, qp->s_sending_hpsn,
		   qp->s_last, qp->s_acked, qp->s_cur,
		   qp->s_tail, qp->s_head, qp->s_size,
		   qp->remote_qpn,
		   qp->remote_ah_attr.dlid);
}

#endif