Donate to e Foundation | Murena handsets with /e/OS | Own a part of Murena! Learn more

Commit cef1cce5 authored by Bryan O'Sullivan's avatar Bryan O'Sullivan Committed by Roland Dreier
Browse files

IB/ipath: misc infiniband code, part 1



Completion queues, local and remote memory keys, and memory region
support.

Signed-off-by: default avatarBryan O'Sullivan <bos@pathscale.com>
Signed-off-by: default avatarRoland Dreier <rolandd@cisco.com>
parent 97f9efbc
Loading
Loading
Loading
Loading
+295 −0
Original line number Diff line number Diff line
/*
 * 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>

#include "ipath_verbs.h"

/**
 * ipath_cq_enter - add a new entry to the completion queue
 * @cq: completion queue
 * @entry: work completion entry to add
 * @sig: true if @entry is a solicitated entry
 *
 * This may be called with one of the qp->s_lock or qp->r_rq.lock held.
 */
void ipath_cq_enter(struct ipath_cq *cq, struct ib_wc *entry, int solicited)
{
	unsigned long flags;
	u32 next;

	spin_lock_irqsave(&cq->lock, flags);

	if (cq->head == cq->ibcq.cqe)
		next = 0;
	else
		next = cq->head + 1;
	if (unlikely(next == cq->tail)) {
		spin_unlock_irqrestore(&cq->lock, flags);
		if (cq->ibcq.event_handler) {
			struct ib_event ev;

			ev.device = cq->ibcq.device;
			ev.element.cq = &cq->ibcq;
			ev.event = IB_EVENT_CQ_ERR;
			cq->ibcq.event_handler(&ev, cq->ibcq.cq_context);
		}
		return;
	}
	cq->queue[cq->head] = *entry;
	cq->head = next;

	if (cq->notify == IB_CQ_NEXT_COMP ||
	    (cq->notify == IB_CQ_SOLICITED && solicited)) {
		cq->notify = IB_CQ_NONE;
		cq->triggered++;
		/*
		 * This will cause send_complete() to be called in
		 * another thread.
		 */
		tasklet_hi_schedule(&cq->comptask);
	}

	spin_unlock_irqrestore(&cq->lock, flags);

	if (entry->status != IB_WC_SUCCESS)
		to_idev(cq->ibcq.device)->n_wqe_errs++;
}

/**
 * ipath_poll_cq - poll for work completion entries
 * @ibcq: the completion queue to poll
 * @num_entries: the maximum number of entries to return
 * @entry: pointer to array where work completions are placed
 *
 * Returns the number of completion entries polled.
 *
 * This may be called from interrupt context.  Also called by ib_poll_cq()
 * in the generic verbs code.
 */
int ipath_poll_cq(struct ib_cq *ibcq, int num_entries, struct ib_wc *entry)
{
	struct ipath_cq *cq = to_icq(ibcq);
	unsigned long flags;
	int npolled;

	spin_lock_irqsave(&cq->lock, flags);

	for (npolled = 0; npolled < num_entries; ++npolled, ++entry) {
		if (cq->tail == cq->head)
			break;
		*entry = cq->queue[cq->tail];
		if (cq->tail == cq->ibcq.cqe)
			cq->tail = 0;
		else
			cq->tail++;
	}

	spin_unlock_irqrestore(&cq->lock, flags);

	return npolled;
}

static void send_complete(unsigned long data)
{
	struct ipath_cq *cq = (struct ipath_cq *)data;

	/*
	 * The completion handler will most likely rearm the notification
	 * and poll for all pending entries.  If a new completion entry
	 * is added while we are in this routine, tasklet_hi_schedule()
	 * won't call us again until we return so we check triggered to
	 * see if we need to call the handler again.
	 */
	for (;;) {
		u8 triggered = cq->triggered;

		cq->ibcq.comp_handler(&cq->ibcq, cq->ibcq.cq_context);

		if (cq->triggered == triggered)
			return;
	}
}

/**
 * ipath_create_cq - create a completion queue
 * @ibdev: the device this completion queue is attached to
 * @entries: the minimum size of the completion queue
 * @context: unused by the InfiniPath driver
 * @udata: unused by the InfiniPath driver
 *
 * Returns a pointer to the completion queue or negative errno values
 * for failure.
 *
 * Called by ib_create_cq() in the generic verbs code.
 */
struct ib_cq *ipath_create_cq(struct ib_device *ibdev, int entries,
			      struct ib_ucontext *context,
			      struct ib_udata *udata)
{
	struct ipath_cq *cq;
	struct ib_wc *wc;
	struct ib_cq *ret;

	/*
	 * Need to use vmalloc() if we want to support large #s of
	 * entries.
	 */
	cq = kmalloc(sizeof(*cq), GFP_KERNEL);
	if (!cq) {
		ret = ERR_PTR(-ENOMEM);
		goto bail;
	}

	/*
	 * Need to use vmalloc() if we want to support large #s of entries.
	 */
	wc = vmalloc(sizeof(*wc) * (entries + 1));
	if (!wc) {
		kfree(cq);
		ret = ERR_PTR(-ENOMEM);
		goto bail;
	}
	/*
	 * ib_create_cq() will initialize cq->ibcq except for cq->ibcq.cqe.
	 * The number of entries should be >= the number requested or return
	 * an error.
	 */
	cq->ibcq.cqe = entries;
	cq->notify = IB_CQ_NONE;
	cq->triggered = 0;
	spin_lock_init(&cq->lock);
	tasklet_init(&cq->comptask, send_complete, (unsigned long)cq);
	cq->head = 0;
	cq->tail = 0;
	cq->queue = wc;

	ret = &cq->ibcq;

bail:
	return ret;
}

/**
 * ipath_destroy_cq - destroy a completion queue
 * @ibcq: the completion queue to destroy.
 *
 * Returns 0 for success.
 *
 * Called by ib_destroy_cq() in the generic verbs code.
 */
int ipath_destroy_cq(struct ib_cq *ibcq)
{
	struct ipath_cq *cq = to_icq(ibcq);

	tasklet_kill(&cq->comptask);
	vfree(cq->queue);
	kfree(cq);

	return 0;
}

/**
 * ipath_req_notify_cq - change the notification type for a completion queue
 * @ibcq: the completion queue
 * @notify: the type of notification to request
 *
 * Returns 0 for success.
 *
 * This may be called from interrupt context.  Also called by
 * ib_req_notify_cq() in the generic verbs code.
 */
int ipath_req_notify_cq(struct ib_cq *ibcq, enum ib_cq_notify notify)
{
	struct ipath_cq *cq = to_icq(ibcq);
	unsigned long flags;

	spin_lock_irqsave(&cq->lock, flags);
	/*
	 * Don't change IB_CQ_NEXT_COMP to IB_CQ_SOLICITED but allow
	 * any other transitions.
	 */
	if (cq->notify != IB_CQ_NEXT_COMP)
		cq->notify = notify;
	spin_unlock_irqrestore(&cq->lock, flags);
	return 0;
}

int ipath_resize_cq(struct ib_cq *ibcq, int cqe, struct ib_udata *udata)
{
	struct ipath_cq *cq = to_icq(ibcq);
	struct ib_wc *wc, *old_wc;
	u32 n;
	int ret;

	/*
	 * Need to use vmalloc() if we want to support large #s of entries.
	 */
	wc = vmalloc(sizeof(*wc) * (cqe + 1));
	if (!wc) {
		ret = -ENOMEM;
		goto bail;
	}

	spin_lock_irq(&cq->lock);
	if (cq->head < cq->tail)
		n = cq->ibcq.cqe + 1 + cq->head - cq->tail;
	else
		n = cq->head - cq->tail;
	if (unlikely((u32)cqe < n)) {
		spin_unlock_irq(&cq->lock);
		vfree(wc);
		ret = -EOVERFLOW;
		goto bail;
	}
	for (n = 0; cq->tail != cq->head; n++) {
		wc[n] = cq->queue[cq->tail];
		if (cq->tail == cq->ibcq.cqe)
			cq->tail = 0;
		else
			cq->tail++;
	}
	cq->ibcq.cqe = cqe;
	cq->head = n;
	cq->tail = 0;
	old_wc = cq->queue;
	cq->queue = wc;
	spin_unlock_irq(&cq->lock);

	vfree(old_wc);

	ret = 0;

bail:
	return ret;
}
+236 −0
Original line number Diff line number Diff line
/*
 * 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 <asm/io.h>

#include "ipath_verbs.h"

/**
 * ipath_alloc_lkey - allocate an lkey
 * @rkt: lkey table in which to allocate the lkey
 * @mr: memory region that this lkey protects
 *
 * Returns 1 if successful, otherwise returns 0.
 */

int ipath_alloc_lkey(struct ipath_lkey_table *rkt, struct ipath_mregion *mr)
{
	unsigned long flags;
	u32 r;
	u32 n;
	int ret;

	spin_lock_irqsave(&rkt->lock, flags);

	/* Find the next available LKEY */
	r = n = rkt->next;
	for (;;) {
		if (rkt->table[r] == NULL)
			break;
		r = (r + 1) & (rkt->max - 1);
		if (r == n) {
			spin_unlock_irqrestore(&rkt->lock, flags);
			_VERBS_INFO("LKEY table full\n");
			ret = 0;
			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_ipath_lkey_table_size)) |
		((((1 << (24 - ib_ipath_lkey_table_size)) - 1) & rkt->gen)
		 << 8);
	if (mr->lkey == 0) {
		mr->lkey |= 1 << 8;
		rkt->gen++;
	}
	rkt->table[r] = mr;
	spin_unlock_irqrestore(&rkt->lock, flags);

	ret = 1;

bail:
	return ret;
}

/**
 * ipath_free_lkey - free an lkey
 * @rkt: table from which to free the lkey
 * @lkey: lkey id to free
 */
void ipath_free_lkey(struct ipath_lkey_table *rkt, u32 lkey)
{
	unsigned long flags;
	u32 r;

	if (lkey == 0)
		return;
	r = lkey >> (32 - ib_ipath_lkey_table_size);
	spin_lock_irqsave(&rkt->lock, flags);
	rkt->table[r] = NULL;
	spin_unlock_irqrestore(&rkt->lock, flags);
}

/**
 * ipath_lkey_ok - check IB SGE for validity and initialize
 * @rkt: table containing lkey to check SGE against
 * @isge: outgoing internal SGE
 * @sge: SGE to check
 * @acc: access flags
 *
 * Return 1 if valid and successful, otherwise returns 0.
 *
 * Check the IB SGE for validity and initialize our internal version
 * of it.
 */
int ipath_lkey_ok(struct ipath_lkey_table *rkt, struct ipath_sge *isge,
		  struct ib_sge *sge, int acc)
{
	struct ipath_mregion *mr;
	size_t off;
	int ret;

	/*
	 * We use LKEY == zero to mean a physical kmalloc() address.
	 * This is a bit of a hack since we rely on dma_map_single()
	 * being reversible by calling bus_to_virt().
	 */
	if (sge->lkey == 0) {
		isge->mr = NULL;
		isge->vaddr = bus_to_virt(sge->addr);
		isge->length = sge->length;
		isge->sge_length = sge->length;
		ret = 1;
		goto bail;
	}
	spin_lock(&rkt->lock);
	mr = rkt->table[(sge->lkey >> (32 - ib_ipath_lkey_table_size))];
	spin_unlock(&rkt->lock);
	if (unlikely(mr == NULL || mr->lkey != sge->lkey)) {
		ret = 0;
		goto bail;
	}

	off = sge->addr - mr->user_base;
	if (unlikely(sge->addr < mr->user_base ||
		     off + sge->length > mr->length ||
		     (mr->access_flags & acc) != acc)) {
		ret = 0;
		goto bail;
	}

	off += mr->offset;
	isge->mr = mr;
	isge->m = 0;
	isge->n = 0;
	while (off >= mr->map[isge->m]->segs[isge->n].length) {
		off -= mr->map[isge->m]->segs[isge->n].length;
		isge->n++;
		if (isge->n >= IPATH_SEGSZ) {
			isge->m++;
			isge->n = 0;
		}
	}
	isge->vaddr = mr->map[isge->m]->segs[isge->n].vaddr + off;
	isge->length = mr->map[isge->m]->segs[isge->n].length - off;
	isge->sge_length = sge->length;

	ret = 1;

bail:
	return ret;
}

/**
 * ipath_rkey_ok - check the IB virtual address, length, and RKEY
 * @dev: infiniband device
 * @ss: 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.
 *
 * The QP r_rq.lock should be held.
 */
int ipath_rkey_ok(struct ipath_ibdev *dev, struct ipath_sge_state *ss,
		  u32 len, u64 vaddr, u32 rkey, int acc)
{
	struct ipath_lkey_table *rkt = &dev->lk_table;
	struct ipath_sge *sge = &ss->sge;
	struct ipath_mregion *mr;
	size_t off;
	int ret;

	spin_lock(&rkt->lock);
	mr = rkt->table[(rkey >> (32 - ib_ipath_lkey_table_size))];
	spin_unlock(&rkt->lock);
	if (unlikely(mr == NULL || mr->lkey != rkey)) {
		ret = 0;
		goto bail;
	}

	off = vaddr - mr->iova;
	if (unlikely(vaddr < mr->iova || off + len > mr->length ||
		     (mr->access_flags & acc) == 0)) {
		ret = 0;
		goto bail;
	}

	off += mr->offset;
	sge->mr = mr;
	sge->m = 0;
	sge->n = 0;
	while (off >= mr->map[sge->m]->segs[sge->n].length) {
		off -= mr->map[sge->m]->segs[sge->n].length;
		sge->n++;
		if (sge->n >= IPATH_SEGSZ) {
			sge->m++;
			sge->n = 0;
		}
	}
	sge->vaddr = mr->map[sge->m]->segs[sge->n].vaddr + off;
	sge->length = mr->map[sge->m]->segs[sge->n].length - off;
	sge->sge_length = len;
	ss->sg_list = NULL;
	ss->num_sge = 1;

	ret = 1;

bail:
	return ret;
}
+383 −0
Original line number Diff line number Diff line
/*
 * 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 <rdma/ib_pack.h>
#include <rdma/ib_smi.h>

#include "ipath_verbs.h"

/**
 * ipath_get_dma_mr - get a DMA memory region
 * @pd: protection domain for this memory region
 * @acc: access flags
 *
 * Returns the memory region on success, otherwise returns an errno.
 */
struct ib_mr *ipath_get_dma_mr(struct ib_pd *pd, int acc)
{
	struct ipath_mr *mr;
	struct ib_mr *ret;

	mr = kzalloc(sizeof *mr, GFP_KERNEL);
	if (!mr) {
		ret = ERR_PTR(-ENOMEM);
		goto bail;
	}

	mr->mr.access_flags = acc;
	ret = &mr->ibmr;

bail:
	return ret;
}

static struct ipath_mr *alloc_mr(int count,
				 struct ipath_lkey_table *lk_table)
{
	struct ipath_mr *mr;
	int m, i = 0;

	/* Allocate struct plus pointers to first level page tables. */
	m = (count + IPATH_SEGSZ - 1) / IPATH_SEGSZ;
	mr = kmalloc(sizeof *mr + m * sizeof mr->mr.map[0], GFP_KERNEL);
	if (!mr)
		goto done;

	/* Allocate first level page tables. */
	for (; i < m; i++) {
		mr->mr.map[i] = kmalloc(sizeof *mr->mr.map[0], GFP_KERNEL);
		if (!mr->mr.map[i])
			goto bail;
	}
	mr->mr.mapsz = m;

	/*
	 * ib_reg_phys_mr() will initialize mr->ibmr except for
	 * lkey and rkey.
	 */
	if (!ipath_alloc_lkey(lk_table, &mr->mr))
		goto bail;
	mr->ibmr.rkey = mr->ibmr.lkey = mr->mr.lkey;

	goto done;

bail:
	while (i) {
		i--;
		kfree(mr->mr.map[i]);
	}
	kfree(mr);
	mr = NULL;

done:
	return mr;
}

/**
 * ipath_reg_phys_mr - register a physical memory region
 * @pd: protection domain for this memory region
 * @buffer_list: pointer to the list of physical buffers to register
 * @num_phys_buf: the number of physical buffers to register
 * @iova_start: the starting address passed over IB which maps to this MR
 *
 * Returns the memory region on success, otherwise returns an errno.
 */
struct ib_mr *ipath_reg_phys_mr(struct ib_pd *pd,
				struct ib_phys_buf *buffer_list,
				int num_phys_buf, int acc, u64 *iova_start)
{
	struct ipath_mr *mr;
	int n, m, i;
	struct ib_mr *ret;

	mr = alloc_mr(num_phys_buf, &to_idev(pd->device)->lk_table);
	if (mr == NULL) {
		ret = ERR_PTR(-ENOMEM);
		goto bail;
	}

	mr->mr.user_base = *iova_start;
	mr->mr.iova = *iova_start;
	mr->mr.length = 0;
	mr->mr.offset = 0;
	mr->mr.access_flags = acc;
	mr->mr.max_segs = num_phys_buf;

	m = 0;
	n = 0;
	for (i = 0; i < num_phys_buf; i++) {
		mr->mr.map[m]->segs[n].vaddr =
			phys_to_virt(buffer_list[i].addr);
		mr->mr.map[m]->segs[n].length = buffer_list[i].size;
		mr->mr.length += buffer_list[i].size;
		n++;
		if (n == IPATH_SEGSZ) {
			m++;
			n = 0;
		}
	}

	ret = &mr->ibmr;

bail:
	return ret;
}

/**
 * ipath_reg_user_mr - register a userspace memory region
 * @pd: protection domain for this memory region
 * @region: the user memory region
 * @mr_access_flags: access flags for this memory region
 * @udata: unused by the InfiniPath driver
 *
 * Returns the memory region on success, otherwise returns an errno.
 */
struct ib_mr *ipath_reg_user_mr(struct ib_pd *pd, struct ib_umem *region,
				int mr_access_flags, struct ib_udata *udata)
{
	struct ipath_mr *mr;
	struct ib_umem_chunk *chunk;
	int n, m, i;
	struct ib_mr *ret;

	n = 0;
	list_for_each_entry(chunk, &region->chunk_list, list)
		n += chunk->nents;

	mr = alloc_mr(n, &to_idev(pd->device)->lk_table);
	if (!mr) {
		ret = ERR_PTR(-ENOMEM);
		goto bail;
	}

	mr->mr.user_base = region->user_base;
	mr->mr.iova = region->virt_base;
	mr->mr.length = region->length;
	mr->mr.offset = region->offset;
	mr->mr.access_flags = mr_access_flags;
	mr->mr.max_segs = n;

	m = 0;
	n = 0;
	list_for_each_entry(chunk, &region->chunk_list, list) {
		for (i = 0; i < chunk->nmap; i++) {
			mr->mr.map[m]->segs[n].vaddr =
				page_address(chunk->page_list[i].page);
			mr->mr.map[m]->segs[n].length = region->page_size;
			n++;
			if (n == IPATH_SEGSZ) {
				m++;
				n = 0;
			}
		}
	}
	ret = &mr->ibmr;

bail:
	return ret;
}

/**
 * ipath_dereg_mr - unregister and free a memory region
 * @ibmr: the memory region to free
 *
 * Returns 0 on success.
 *
 * Note that this is called to free MRs created by ipath_get_dma_mr()
 * or ipath_reg_user_mr().
 */
int ipath_dereg_mr(struct ib_mr *ibmr)
{
	struct ipath_mr *mr = to_imr(ibmr);
	int i;

	ipath_free_lkey(&to_idev(ibmr->device)->lk_table, ibmr->lkey);
	i = mr->mr.mapsz;
	while (i) {
		i--;
		kfree(mr->mr.map[i]);
	}
	kfree(mr);
	return 0;
}

/**
 * ipath_alloc_fmr - allocate a fast memory region
 * @pd: the protection domain for this memory region
 * @mr_access_flags: access flags for this memory region
 * @fmr_attr: fast memory region attributes
 *
 * Returns the memory region on success, otherwise returns an errno.
 */
struct ib_fmr *ipath_alloc_fmr(struct ib_pd *pd, int mr_access_flags,
			       struct ib_fmr_attr *fmr_attr)
{
	struct ipath_fmr *fmr;
	int m, i = 0;
	struct ib_fmr *ret;

	/* Allocate struct plus pointers to first level page tables. */
	m = (fmr_attr->max_pages + IPATH_SEGSZ - 1) / IPATH_SEGSZ;
	fmr = kmalloc(sizeof *fmr + m * sizeof fmr->mr.map[0], GFP_KERNEL);
	if (!fmr)
		goto bail;

	/* Allocate first level page tables. */
	for (; i < m; i++) {
		fmr->mr.map[i] = kmalloc(sizeof *fmr->mr.map[0],
					 GFP_KERNEL);
		if (!fmr->mr.map[i])
			goto bail;
	}
	fmr->mr.mapsz = m;

	/*
	 * ib_alloc_fmr() will initialize fmr->ibfmr except for lkey &
	 * rkey.
	 */
	if (!ipath_alloc_lkey(&to_idev(pd->device)->lk_table, &fmr->mr))
		goto bail;
	fmr->ibfmr.rkey = fmr->ibfmr.lkey = fmr->mr.lkey;
	/*
	 * Resources are allocated but no valid mapping (RKEY can't be
	 * used).
	 */
	fmr->mr.user_base = 0;
	fmr->mr.iova = 0;
	fmr->mr.length = 0;
	fmr->mr.offset = 0;
	fmr->mr.access_flags = mr_access_flags;
	fmr->mr.max_segs = fmr_attr->max_pages;
	fmr->page_shift = fmr_attr->page_shift;

	ret = &fmr->ibfmr;
	goto done;

bail:
	while (i)
		kfree(fmr->mr.map[--i]);
	kfree(fmr);
	ret = ERR_PTR(-ENOMEM);

done:
	return ret;
}

/**
 * ipath_map_phys_fmr - set up a fast memory region
 * @ibmfr: the fast memory region to set up
 * @page_list: the list of pages to associate with the fast memory region
 * @list_len: the number of pages to associate with the fast memory region
 * @iova: the virtual address of the start of the fast memory region
 *
 * This may be called from interrupt context.
 */

int ipath_map_phys_fmr(struct ib_fmr *ibfmr, u64 * page_list,
		       int list_len, u64 iova)
{
	struct ipath_fmr *fmr = to_ifmr(ibfmr);
	struct ipath_lkey_table *rkt;
	unsigned long flags;
	int m, n, i;
	u32 ps;
	int ret;

	if (list_len > fmr->mr.max_segs) {
		ret = -EINVAL;
		goto bail;
	}
	rkt = &to_idev(ibfmr->device)->lk_table;
	spin_lock_irqsave(&rkt->lock, flags);
	fmr->mr.user_base = iova;
	fmr->mr.iova = iova;
	ps = 1 << fmr->page_shift;
	fmr->mr.length = list_len * ps;
	m = 0;
	n = 0;
	ps = 1 << fmr->page_shift;
	for (i = 0; i < list_len; i++) {
		fmr->mr.map[m]->segs[n].vaddr = phys_to_virt(page_list[i]);
		fmr->mr.map[m]->segs[n].length = ps;
		if (++n == IPATH_SEGSZ) {
			m++;
			n = 0;
		}
	}
	spin_unlock_irqrestore(&rkt->lock, flags);
	ret = 0;

bail:
	return ret;
}

/**
 * ipath_unmap_fmr - unmap fast memory regions
 * @fmr_list: the list of fast memory regions to unmap
 *
 * Returns 0 on success.
 */
int ipath_unmap_fmr(struct list_head *fmr_list)
{
	struct ipath_fmr *fmr;
	struct ipath_lkey_table *rkt;
	unsigned long flags;

	list_for_each_entry(fmr, fmr_list, ibfmr.list) {
		rkt = &to_idev(fmr->ibfmr.device)->lk_table;
		spin_lock_irqsave(&rkt->lock, flags);
		fmr->mr.user_base = 0;
		fmr->mr.iova = 0;
		fmr->mr.length = 0;
		spin_unlock_irqrestore(&rkt->lock, flags);
	}
	return 0;
}

/**
 * ipath_dealloc_fmr - deallocate a fast memory region
 * @ibfmr: the fast memory region to deallocate
 *
 * Returns 0 on success.
 */
int ipath_dealloc_fmr(struct ib_fmr *ibfmr)
{
	struct ipath_fmr *fmr = to_ifmr(ibfmr);
	int i;

	ipath_free_lkey(&to_idev(ibfmr->device)->lk_table, ibfmr->lkey);
	i = fmr->mr.mapsz;
	while (i)
		kfree(fmr->mr.map[--i]);
	kfree(fmr);
	return 0;
}
+273 −0

File added.

Preview size limit exceeded, changes collapsed.