[deliverable/linux.git] / drivers / infiniband / hw / qib / qib_keys.c

/*
 * 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
 * @dma_region: 0->normal key, 1->restricted DMA key
 *
 * Returns 0 if successful, otherwise returns -errno.
 *
 * Increments mr reference count as required.
 *
 * Sets the lkey field mr for non-dma regions.
 *
 */

int qib_alloc_lkey(struct qib_mregion *mr, int dma_region)
{
	unsigned long flags;
	u32 r;
	u32 n;
	int ret = 0;
	struct qib_ibdev *dev = to_idev(mr->pd->device);
	struct qib_lkey_table *rkt = &dev->lk_table;

	spin_lock_irqsave(&rkt->lock, flags);

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

		tmr = rcu_dereference(dev->dma_mr);
		if (!tmr) {
			qib_get_mr(mr);
			rcu_assign_pointer(dev->dma_mr, mr);
			mr->lkey_published = 1;
		}
		goto success;
	}

	/* Find the next available LKEY */
	r = rkt->next;
	n = r;
	for (;;) {
		if (rkt->table[r] == NULL)
			break;
		r = (r + 1) & (rkt->max - 1);
		if (r == n)
			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++;
	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++;
	}
	qib_get_mr(mr);
	rcu_assign_pointer(rkt->table[r], mr);
	mr->lkey_published = 1;
success:
	spin_unlock_irqrestore(&rkt->lock, flags);
out:
	return ret;
bail:
	spin_unlock_irqrestore(&rkt->lock, flags);
	ret = -ENOMEM;
	goto out;
}

/**
 * qib_free_lkey - free an lkey
 * @mr: mr to free from tables
 */
void qib_free_lkey(struct qib_mregion *mr)
{
	unsigned long flags;
	u32 lkey = mr->lkey;
	u32 r;
	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;
	if (lkey == 0)
		rcu_assign_pointer(dev->dma_mr, NULL);
	else {
		r = lkey >> (32 - ib_qib_lkey_table_size);
		rcu_assign_pointer(rkt->table[r], NULL);
	}
	qib_put_mr(mr);
	mr->lkey_published = 0;
out:
	spin_unlock_irqrestore(&rkt->lock, flags);
}

/**
 * qib_lkey_ok - check IB SGE for validity and initialize
 * @rkt: table containing lkey to check SGE against
 * @pd: protection domain
 * @isge: outgoing internal SGE
 * @sge: SGE to check
 * @acc: access flags
 *
 * Return 1 if valid and successful, otherwise returns 0.
 *
 * increments the reference count upon success
 *
 * 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).
	 */
	rcu_read_lock();
	if (sge->lkey == 0) {
		struct qib_ibdev *dev = to_idev(pd->ibpd.device);

		if (pd->user)
			goto bail;
		mr = rcu_dereference(dev->dma_mr);
		if (!mr)
			goto bail;
		if (unlikely(!atomic_inc_not_zero(&mr->refcount)))
			goto bail;
		rcu_read_unlock();

		isge->mr = mr;
		isge->vaddr = (void *) sge->addr;
		isge->length = sge->length;
		isge->sge_length = sge->length;
		isge->m = 0;
		isge->n = 0;
		goto ok;
	}
	mr = rcu_dereference(
		rkt->table[(sge->lkey >> (32 - ib_qib_lkey_table_size))]);
	if (unlikely(!mr || mr->lkey != sge->lkey || mr->pd != &pd->ibpd))
		goto bail;

	off = sge->addr - mr->user_base;
	if (unlikely(sge->addr < mr->iova || off + sge->length > mr->length ||
		     (mr->access_flags & acc) == 0))
		goto bail;
	if (unlikely(!atomic_inc_not_zero(&mr->refcount)))
		goto bail;
	rcu_read_unlock();

	off += mr->offset;
	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;
			}
		}
	}
	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:
	return 1;
bail:
	rcu_read_unlock();
	return 0;
}

/**
 * qib_rkey_ok - check the IB virtual address, length, and RKEY
 * @qp: qp for validation
 * @sge: SGE state
 * @len: length of data
 * @vaddr: virtual address to place data
 * @rkey: rkey to check
 * @acc: access flags
 *
 * Return 1 if successful, otherwise 0.
 *
 * increments the reference count upon success
 */
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).
	 */
	rcu_read_lock();
	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;
		mr = rcu_dereference(dev->dma_mr);
		if (!mr)
			goto bail;
		if (unlikely(!atomic_inc_not_zero(&mr->refcount)))
			goto bail;
		rcu_read_unlock();

		sge->mr = mr;
		sge->vaddr = (void *) vaddr;
		sge->length = len;
		sge->sge_length = len;
		sge->m = 0;
		sge->n = 0;
		goto ok;
	}

	mr = rcu_dereference(
		rkt->table[(rkey >> (32 - ib_qib_lkey_table_size))]);
	if (unlikely(!mr || mr->lkey != rkey || qp->ibqp.pd != mr->pd))
		goto bail;

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

	off += mr->offset;
	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;
			}
		}
	}
	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:
	return 1;
bail:
	rcu_read_unlock();
	return 0;
}

/*
 * 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;

	mr = rcu_dereference_protected(
		rkt->table[(rkey >> (32 - ib_qib_lkey_table_size))],
		lockdep_is_held(&rkt->lock));
	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;
}
Commit	Line	Data
f931551b RC	1	/*
	2	* Copyright (c) 2006, 2007, 2009 QLogic Corporation. All rights reserved.
	3	* Copyright (c) 2005, 2006 PathScale, Inc. All rights reserved.
	4	*
	5	* This software is available to you under a choice of one of two
	6	* licenses. You may choose to be licensed under the terms of the GNU
	7	* General Public License (GPL) Version 2, available from the file
	8	* COPYING in the main directory of this source tree, or the
	9	* OpenIB.org BSD license below:
	10	*
	11	* Redistribution and use in source and binary forms, with or
	12	* without modification, are permitted provided that the following
	13	* conditions are met:
	14	*
	15	* - Redistributions of source code must retain the above
	16	* copyright notice, this list of conditions and the following
	17	* disclaimer.
	18	*
	19	* - Redistributions in binary form must reproduce the above
	20	* copyright notice, this list of conditions and the following
	21	* disclaimer in the documentation and/or other materials
	22	* provided with the distribution.
	23	*
	24	* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
	25	* EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
	26	* MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
	27	* NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
	28	* BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
	29	* ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
	30	* CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
	31	* SOFTWARE.
	32	*/
	33
	34	#include "qib.h"
	35
	36	/**
	37	* qib_alloc_lkey - allocate an lkey
f931551b	38	* @mr: memory region that this lkey protects
6a82649f MM	39	* @dma_region: 0->normal key, 1->restricted DMA key
	40	*
	41	* Returns 0 if successful, otherwise returns -errno.
	42	*
8aac4cc3	43	* Increments mr reference count as required.
6a82649f MM	44	*
6a82649f MM	45	* Sets the lkey field mr for non-dma regions.
f931551b	46	*
f931551b RC	47	*/
f931551b RC	48
6a82649f	49	int qib_alloc_lkey(struct qib_mregion *mr, int dma_region)
f931551b RC	50	{
	51	unsigned long flags;
	52	u32 r;
	53	u32 n;
6a82649f MM	54	int ret = 0;
	55	struct qib_ibdev *dev = to_idev(mr->pd->device);
	56	struct qib_lkey_table *rkt = &dev->lk_table;
f931551b RC	57
	58	spin_lock_irqsave(&rkt->lock, flags);
	59
6a82649f MM	60	/* special case for dma_mr lkey == 0 */
6a82649f MM	61	if (dma_region) {
8aac4cc3 MM	62	struct qib_mregion *tmr;
	63
	64	tmr = rcu_dereference(dev->dma_mr);
	65	if (!tmr) {
6a82649f	66	qib_get_mr(mr);
8aac4cc3	67	rcu_assign_pointer(dev->dma_mr, mr);
6a82649f MM	68	mr->lkey_published = 1;
	69	}
	70	goto success;
	71	}
	72
f931551b RC	73	/* Find the next available LKEY */
	74	r = rkt->next;
	75	n = r;
	76	for (;;) {
	77	if (rkt->table[r] == NULL)
	78	break;
	79	r = (r + 1) & (rkt->max - 1);
6a82649f	80	if (r == n)
f931551b	81	goto bail;
f931551b RC	82	}
	83	rkt->next = (r + 1) & (rkt->max - 1);
	84	/*
	85	* Make sure lkey is never zero which is reserved to indicate an
	86	* unrestricted LKEY.
	87	*/
	88	rkt->gen++;
	89	mr->lkey = (r << (32 - ib_qib_lkey_table_size)) \|
	90	((((1 << (24 - ib_qib_lkey_table_size)) - 1) & rkt->gen)
	91	<< 8);
	92	if (mr->lkey == 0) {
	93	mr->lkey \|= 1 << 8;
	94	rkt->gen++;
	95	}
6a82649f	96	qib_get_mr(mr);
8aac4cc3	97	rcu_assign_pointer(rkt->table[r], mr);
6a82649f MM	98	mr->lkey_published = 1;
6a82649f MM	99	success:
f931551b	100	spin_unlock_irqrestore(&rkt->lock, flags);
6a82649f	101	out:
f931551b	102	return ret;
6a82649f MM	103	bail:
	104	spin_unlock_irqrestore(&rkt->lock, flags);
	105	ret = -ENOMEM;
	106	goto out;
f931551b RC	107	}
	108
	109	/**
	110	* qib_free_lkey - free an lkey
6a82649f	111	* @mr: mr to free from tables
f931551b	112	*/
6a82649f	113	void qib_free_lkey(struct qib_mregion *mr)
f931551b RC	114	{
	115	unsigned long flags;
	116	u32 lkey = mr->lkey;
	117	u32 r;
6a82649f MM	118	struct qib_ibdev *dev = to_idev(mr->pd->device);
	119	struct qib_lkey_table *rkt = &dev->lk_table;
	120
	121	spin_lock_irqsave(&rkt->lock, flags);
	122	if (!mr->lkey_published)
	123	goto out;
8aac4cc3 MM	124	if (lkey == 0)
	125	rcu_assign_pointer(dev->dma_mr, NULL);
	126	else {
f931551b	127	r = lkey >> (32 - ib_qib_lkey_table_size);
8aac4cc3	128	rcu_assign_pointer(rkt->table[r], NULL);
f931551b	129	}
8aac4cc3 MM	130	qib_put_mr(mr);
8aac4cc3 MM	131	mr->lkey_published = 0;
6a82649f	132	out:
8aac4cc3	133	spin_unlock_irqrestore(&rkt->lock, flags);
f931551b RC	134	}
	135
	136	/**
	137	* qib_lkey_ok - check IB SGE for validity and initialize
	138	* @rkt: table containing lkey to check SGE against
8aac4cc3	139	* @pd: protection domain
f931551b RC	140	* @isge: outgoing internal SGE
	141	* @sge: SGE to check
	142	* @acc: access flags
	143	*
	144	* Return 1 if valid and successful, otherwise returns 0.
	145	*
8aac4cc3 MM	146	* increments the reference count upon success
8aac4cc3 MM	147	*
f931551b RC	148	* Check the IB SGE for validity and initialize our internal version
	149	* of it.
	150	*/
	151	int qib_lkey_ok(struct qib_lkey_table rkt, struct qib_pd pd,
	152	struct qib_sge isge, struct ib_sge sge, int acc)
	153	{
	154	struct qib_mregion *mr;
	155	unsigned n, m;
	156	size_t off;
f931551b RC	157
	158	/*
	159	* We use LKEY == zero for kernel virtual addresses
	160	* (see qib_get_dma_mr and qib_dma.c).
	161	*/
8aac4cc3	162	rcu_read_lock();
f931551b RC	163	if (sge->lkey == 0) {
	164	struct qib_ibdev *dev = to_idev(pd->ibpd.device);
	165
	166	if (pd->user)
	167	goto bail;
8aac4cc3 MM	168	mr = rcu_dereference(dev->dma_mr);
	169	if (!mr)
	170	goto bail;
	171	if (unlikely(!atomic_inc_not_zero(&mr->refcount)))
f931551b	172	goto bail;
8aac4cc3	173	rcu_read_unlock();
4db62d47	174
8aac4cc3	175	isge->mr = mr;
f931551b RC	176	isge->vaddr = (void *) sge->addr;
	177	isge->length = sge->length;
	178	isge->sge_length = sge->length;
	179	isge->m = 0;
	180	isge->n = 0;
	181	goto ok;
	182	}
8aac4cc3 MM	183	mr = rcu_dereference(
	184	rkt->table[(sge->lkey >> (32 - ib_qib_lkey_table_size))]);
	185	if (unlikely(!mr \|\| mr->lkey != sge->lkey \|\| mr->pd != &pd->ibpd))
f931551b RC	186	goto bail;
	187
	188	off = sge->addr - mr->user_base;
8aac4cc3 MM	189	if (unlikely(sge->addr < mr->iova \|\| off + sge->length > mr->length \|\|
8aac4cc3 MM	190	(mr->access_flags & acc) == 0))
4db62d47	191	goto bail;
8aac4cc3 MM	192	if (unlikely(!atomic_inc_not_zero(&mr->refcount)))
	193	goto bail;
	194	rcu_read_unlock();
f931551b RC	195
f931551b RC	196	off += mr->offset;
2a600f14 MM	197	if (mr->page_shift) {
	198	/*
	199	page sizes are uniform power of 2 so no loop is necessary
	200	entries_spanned_by_off is the number of times the loop below
	201	would have executed.
	202	*/
	203	size_t entries_spanned_by_off;
	204
	205	entries_spanned_by_off = off >> mr->page_shift;
	206	off -= (entries_spanned_by_off << mr->page_shift);
	207	m = entries_spanned_by_off/QIB_SEGSZ;
	208	n = entries_spanned_by_off%QIB_SEGSZ;
	209	} else {
	210	m = 0;
	211	n = 0;
	212	while (off >= mr->map[m]->segs[n].length) {
	213	off -= mr->map[m]->segs[n].length;
	214	n++;
	215	if (n >= QIB_SEGSZ) {
	216	m++;
	217	n = 0;
	218	}
f931551b RC	219	}
f931551b RC	220	}
f931551b RC	221	isge->mr = mr;
	222	isge->vaddr = mr->map[m]->segs[n].vaddr + off;
	223	isge->length = mr->map[m]->segs[n].length - off;
	224	isge->sge_length = sge->length;
	225	isge->m = m;
	226	isge->n = n;
	227	ok:
4db62d47	228	return 1;
f931551b	229	bail:
8aac4cc3	230	rcu_read_unlock();
4db62d47	231	return 0;
f931551b RC	232	}
	233
	234	/**
	235	* qib_rkey_ok - check the IB virtual address, length, and RKEY
8aac4cc3 MM	236	* @qp: qp for validation
8aac4cc3 MM	237	* @sge: SGE state
f931551b RC	238	* @len: length of data
	239	* @vaddr: virtual address to place data
	240	* @rkey: rkey to check
	241	* @acc: access flags
	242	*
	243	* Return 1 if successful, otherwise 0.
8aac4cc3 MM	244	*
8aac4cc3 MM	245	* increments the reference count upon success
f931551b RC	246	*/
	247	int qib_rkey_ok(struct qib_qp qp, struct qib_sge sge,
	248	u32 len, u64 vaddr, u32 rkey, int acc)
	249	{
	250	struct qib_lkey_table *rkt = &to_idev(qp->ibqp.device)->lk_table;
	251	struct qib_mregion *mr;
	252	unsigned n, m;
	253	size_t off;
f931551b RC	254
	255	/*
	256	* We use RKEY == zero for kernel virtual addresses
	257	* (see qib_get_dma_mr and qib_dma.c).
	258	*/
8aac4cc3	259	rcu_read_lock();
f931551b RC	260	if (rkey == 0) {
	261	struct qib_pd *pd = to_ipd(qp->ibqp.pd);
	262	struct qib_ibdev *dev = to_idev(pd->ibpd.device);
	263
	264	if (pd->user)
	265	goto bail;
8aac4cc3 MM	266	mr = rcu_dereference(dev->dma_mr);
	267	if (!mr)
	268	goto bail;
	269	if (unlikely(!atomic_inc_not_zero(&mr->refcount)))
f931551b	270	goto bail;
8aac4cc3	271	rcu_read_unlock();
4db62d47	272
8aac4cc3	273	sge->mr = mr;
f931551b RC	274	sge->vaddr = (void *) vaddr;
	275	sge->length = len;
	276	sge->sge_length = len;
	277	sge->m = 0;
	278	sge->n = 0;
	279	goto ok;
	280	}
	281
8aac4cc3 MM	282	mr = rcu_dereference(
	283	rkt->table[(rkey >> (32 - ib_qib_lkey_table_size))]);
	284	if (unlikely(!mr \|\| mr->lkey != rkey \|\| qp->ibqp.pd != mr->pd))
f931551b RC	285	goto bail;
	286
	287	off = vaddr - mr->iova;
	288	if (unlikely(vaddr < mr->iova \|\| off + len > mr->length \|\|
	289	(mr->access_flags & acc) == 0))
4db62d47	290	goto bail;
8aac4cc3 MM	291	if (unlikely(!atomic_inc_not_zero(&mr->refcount)))
	292	goto bail;
	293	rcu_read_unlock();
f931551b RC	294
f931551b RC	295	off += mr->offset;
2a600f14 MM	296	if (mr->page_shift) {
	297	/*
	298	page sizes are uniform power of 2 so no loop is necessary
	299	entries_spanned_by_off is the number of times the loop below
	300	would have executed.
	301	*/
	302	size_t entries_spanned_by_off;
	303
	304	entries_spanned_by_off = off >> mr->page_shift;
	305	off -= (entries_spanned_by_off << mr->page_shift);
	306	m = entries_spanned_by_off/QIB_SEGSZ;
	307	n = entries_spanned_by_off%QIB_SEGSZ;
	308	} else {
	309	m = 0;
	310	n = 0;
	311	while (off >= mr->map[m]->segs[n].length) {
	312	off -= mr->map[m]->segs[n].length;
	313	n++;
	314	if (n >= QIB_SEGSZ) {
	315	m++;
	316	n = 0;
	317	}
f931551b RC	318	}
f931551b RC	319	}
f931551b RC	320	sge->mr = mr;
	321	sge->vaddr = mr->map[m]->segs[n].vaddr + off;
	322	sge->length = mr->map[m]->segs[n].length - off;
	323	sge->sge_length = len;
	324	sge->m = m;
	325	sge->n = n;
	326	ok:
4db62d47	327	return 1;
f931551b	328	bail:
8aac4cc3	329	rcu_read_unlock();
4db62d47	330	return 0;
f931551b RC	331	}
	332
	333	/*
	334	* Initialize the memory region specified by the work reqeust.
	335	*/
	336	int qib_fast_reg_mr(struct qib_qp qp, struct ib_send_wr wr)
	337	{
	338	struct qib_lkey_table *rkt = &to_idev(qp->ibqp.device)->lk_table;
	339	struct qib_pd *pd = to_ipd(qp->ibqp.pd);
	340	struct qib_mregion *mr;
	341	u32 rkey = wr->wr.fast_reg.rkey;
	342	unsigned i, n, m;
	343	int ret = -EINVAL;
	344	unsigned long flags;
	345	u64 *page_list;
	346	size_t ps;
	347
	348	spin_lock_irqsave(&rkt->lock, flags);
	349	if (pd->user \|\| rkey == 0)
	350	goto bail;
	351
7e230177 MM	352	mr = rcu_dereference_protected(
	353	rkt->table[(rkey >> (32 - ib_qib_lkey_table_size))],
	354	lockdep_is_held(&rkt->lock));
f931551b RC	355	if (unlikely(mr == NULL \|\| qp->ibqp.pd != mr->pd))
	356	goto bail;
	357
	358	if (wr->wr.fast_reg.page_list_len > mr->max_segs)
	359	goto bail;
	360
	361	ps = 1UL << wr->wr.fast_reg.page_shift;
	362	if (wr->wr.fast_reg.length > ps * wr->wr.fast_reg.page_list_len)
	363	goto bail;
	364
	365	mr->user_base = wr->wr.fast_reg.iova_start;
	366	mr->iova = wr->wr.fast_reg.iova_start;
	367	mr->lkey = rkey;
	368	mr->length = wr->wr.fast_reg.length;
	369	mr->access_flags = wr->wr.fast_reg.access_flags;
	370	page_list = wr->wr.fast_reg.page_list->page_list;
	371	m = 0;
	372	n = 0;
	373	for (i = 0; i < wr->wr.fast_reg.page_list_len; i++) {
	374	mr->map[m]->segs[n].vaddr = (void *) page_list[i];
	375	mr->map[m]->segs[n].length = ps;
	376	if (++n == QIB_SEGSZ) {
	377	m++;
	378	n = 0;
	379	}
	380	}
	381
	382	ret = 0;
	383	bail:
	384	spin_unlock_irqrestore(&rkt->lock, flags);
	385	return ret;
	386	}