[linux-2.6-block.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 rvt_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 rvt_lkey_table *rkt = &dev->lk_table;

	spin_lock_irqsave(&rkt->lock, flags);

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

		tmr = rcu_access_pointer(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++;
	/*
	 * bits are capped in qib_verbs.c to insure enough bits
	 * for generation number
	 */
	mr->lkey = (r << (32 - ib_rvt_lkey_table_size)) |
		((((1 << (24 - ib_rvt_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 rvt_mregion *mr)
{
	unsigned long flags;
	u32 lkey = mr->lkey;
	u32 r;
	struct qib_ibdev *dev = to_idev(mr->pd->device);
	struct rvt_lkey_table *rkt = &dev->lk_table;

	spin_lock_irqsave(&rkt->lock, flags);
	if (!mr->lkey_published)
		goto out;
	if (lkey == 0)
		RCU_INIT_POINTER(dev->dma_mr, NULL);
	else {
		r = lkey >> (32 - ib_rvt_lkey_table_size);
		RCU_INIT_POINTER(rkt->table[r], NULL);
	}
	qib_put_mr(mr);
	mr->lkey_published = 0;
out:
	spin_unlock_irqrestore(&rkt->lock, flags);
}

/**
 * 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 rvt_qp *qp, struct rvt_sge *sge,
		u32 len, u64 vaddr, u32 rkey, int acc)
{
	struct rvt_lkey_table *rkt = &to_idev(qp->ibqp.device)->lk_table;
	struct rvt_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 rvt_pd *pd = ibpd_to_rvtpd(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_rvt_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 / RVT_SEGSZ;
		n = entries_spanned_by_off % RVT_SEGSZ;
	} else {
		m = 0;
		n = 0;
		while (off >= mr->map[m]->segs[n].length) {
			off -= mr->map[m]->segs[n].length;
			n++;
			if (n >= RVT_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;
}
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
7c2e11fe	49	int qib_alloc_lkey(struct rvt_mregion *mr, int dma_region)
f931551b RC	50	{
	51	unsigned long flags;
	52	u32 r;
	53	u32 n;
6a82649f MM	54	int ret = 0;
6a82649f MM	55	struct qib_ibdev *dev = to_idev(mr->pd->device);
7c2e11fe	56	struct rvt_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) {
7c2e11fe	62	struct rvt_mregion *tmr;
8aac4cc3	63
f3bdf344	64	tmr = rcu_access_pointer(dev->dma_mr);
8aac4cc3	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++;
d6f1c17e MM	89	/*
	90	* bits are capped in qib_verbs.c to insure enough bits
	91	* for generation number
	92	*/
7c2e11fe DD	93	mr->lkey = (r << (32 - ib_rvt_lkey_table_size)) \|
7c2e11fe DD	94	((((1 << (24 - ib_rvt_lkey_table_size)) - 1) & rkt->gen)
f931551b RC	95	<< 8);
	96	if (mr->lkey == 0) {
	97	mr->lkey \|= 1 << 8;
	98	rkt->gen++;
	99	}
6a82649f	100	qib_get_mr(mr);
8aac4cc3	101	rcu_assign_pointer(rkt->table[r], mr);
6a82649f MM	102	mr->lkey_published = 1;
6a82649f MM	103	success:
f931551b	104	spin_unlock_irqrestore(&rkt->lock, flags);
6a82649f	105	out:
f931551b	106	return ret;
6a82649f MM	107	bail:
	108	spin_unlock_irqrestore(&rkt->lock, flags);
	109	ret = -ENOMEM;
	110	goto out;
f931551b RC	111	}
	112
	113	/**
	114	* qib_free_lkey - free an lkey
6a82649f	115	* @mr: mr to free from tables
f931551b	116	*/
7c2e11fe	117	void qib_free_lkey(struct rvt_mregion *mr)
f931551b RC	118	{
	119	unsigned long flags;
	120	u32 lkey = mr->lkey;
	121	u32 r;
6a82649f	122	struct qib_ibdev *dev = to_idev(mr->pd->device);
7c2e11fe	123	struct rvt_lkey_table *rkt = &dev->lk_table;
6a82649f MM	124
	125	spin_lock_irqsave(&rkt->lock, flags);
	126	if (!mr->lkey_published)
	127	goto out;
8aac4cc3	128	if (lkey == 0)
590c3fec	129	RCU_INIT_POINTER(dev->dma_mr, NULL);
8aac4cc3	130	else {
7c2e11fe	131	r = lkey >> (32 - ib_rvt_lkey_table_size);
590c3fec	132	RCU_INIT_POINTER(rkt->table[r], NULL);
f931551b	133	}
8aac4cc3 MM	134	qib_put_mr(mr);
8aac4cc3 MM	135	mr->lkey_published = 0;
6a82649f	136	out:
8aac4cc3	137	spin_unlock_irqrestore(&rkt->lock, flags);
f931551b RC	138	}
f931551b RC	139
f931551b RC	140	/**
f931551b RC	141	* qib_rkey_ok - check the IB virtual address, length, and RKEY
8aac4cc3 MM	142	* @qp: qp for validation
8aac4cc3 MM	143	* @sge: SGE state
f931551b RC	144	* @len: length of data
	145	* @vaddr: virtual address to place data
	146	* @rkey: rkey to check
	147	* @acc: access flags
	148	*
	149	* Return 1 if successful, otherwise 0.
8aac4cc3 MM	150	*
8aac4cc3 MM	151	* increments the reference count upon success
f931551b	152	*/
7c2e11fe	153	int qib_rkey_ok(struct rvt_qp qp, struct rvt_sge sge,
f931551b RC	154	u32 len, u64 vaddr, u32 rkey, int acc)
f931551b RC	155	{
7c2e11fe DD	156	struct rvt_lkey_table *rkt = &to_idev(qp->ibqp.device)->lk_table;
7c2e11fe DD	157	struct rvt_mregion *mr;
f931551b RC	158	unsigned n, m;
f931551b RC	159	size_t off;
f931551b RC	160
	161	/*
	162	* We use RKEY == zero for kernel virtual addresses
	163	* (see qib_get_dma_mr and qib_dma.c).
	164	*/
8aac4cc3	165	rcu_read_lock();
f931551b	166	if (rkey == 0) {
f44728d6	167	struct rvt_pd *pd = ibpd_to_rvtpd(qp->ibqp.pd);
f931551b RC	168	struct qib_ibdev *dev = to_idev(pd->ibpd.device);
	169
	170	if (pd->user)
	171	goto bail;
8aac4cc3 MM	172	mr = rcu_dereference(dev->dma_mr);
	173	if (!mr)
	174	goto bail;
	175	if (unlikely(!atomic_inc_not_zero(&mr->refcount)))
f931551b	176	goto bail;
8aac4cc3	177	rcu_read_unlock();
4db62d47	178
8aac4cc3	179	sge->mr = mr;
f931551b RC	180	sge->vaddr = (void *) vaddr;
	181	sge->length = len;
	182	sge->sge_length = len;
	183	sge->m = 0;
	184	sge->n = 0;
	185	goto ok;
	186	}
	187
8aac4cc3	188	mr = rcu_dereference(
7c2e11fe	189	rkt->table[(rkey >> (32 - ib_rvt_lkey_table_size))]);
8aac4cc3	190	if (unlikely(!mr \|\| mr->lkey != rkey \|\| qp->ibqp.pd != mr->pd))
f931551b RC	191	goto bail;
	192
	193	off = vaddr - mr->iova;
	194	if (unlikely(vaddr < mr->iova \|\| off + len > mr->length \|\|
	195	(mr->access_flags & acc) == 0))
4db62d47	196	goto bail;
8aac4cc3 MM	197	if (unlikely(!atomic_inc_not_zero(&mr->refcount)))
	198	goto bail;
	199	rcu_read_unlock();
f931551b RC	200
f931551b RC	201	off += mr->offset;
2a600f14 MM	202	if (mr->page_shift) {
	203	/*
	204	page sizes are uniform power of 2 so no loop is necessary
	205	entries_spanned_by_off is the number of times the loop below
	206	would have executed.
	207	*/
	208	size_t entries_spanned_by_off;
	209
	210	entries_spanned_by_off = off >> mr->page_shift;
	211	off -= (entries_spanned_by_off << mr->page_shift);
7c2e11fe DD	212	m = entries_spanned_by_off / RVT_SEGSZ;
7c2e11fe DD	213	n = entries_spanned_by_off % RVT_SEGSZ;
2a600f14 MM	214	} else {
	215	m = 0;
	216	n = 0;
	217	while (off >= mr->map[m]->segs[n].length) {
	218	off -= mr->map[m]->segs[n].length;
	219	n++;
7c2e11fe	220	if (n >= RVT_SEGSZ) {
2a600f14 MM	221	m++;
	222	n = 0;
	223	}
f931551b RC	224	}
f931551b RC	225	}
f931551b RC	226	sge->mr = mr;
	227	sge->vaddr = mr->map[m]->segs[n].vaddr + off;
	228	sge->length = mr->map[m]->segs[n].length - off;
	229	sge->sge_length = len;
	230	sge->m = m;
	231	sge->n = n;
	232	ok:
4db62d47	233	return 1;
f931551b	234	bail:
8aac4cc3	235	rcu_read_unlock();
4db62d47	236	return 0;
f931551b RC	237	}
f931551b RC	238