[linux-2.6-block.git] / drivers / dma / iovlock.c

/*
 * Copyright(c) 2004 - 2006 Intel Corporation. All rights reserved.
 * Portions based on net/core/datagram.c and copyrighted by their authors.
 *
 * This program is free software; you can redistribute it and/or modify it
 * under the terms of the GNU General Public License as published by the Free
 * Software Foundation; either version 2 of the License, or (at your option)
 * any later version.
 *
 * This program is distributed in the hope that it will be useful, but WITHOUT
 * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
 * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License for
 * more details.
 *
 * You should have received a copy of the GNU General Public License along with
 * this program; if not, write to the Free Software Foundation, Inc., 59
 * Temple Place - Suite 330, Boston, MA  02111-1307, USA.
 *
 * The full GNU General Public License is included in this distribution in the
 * file called COPYING.
 */

/*
 * This code allows the net stack to make use of a DMA engine for
 * skb to iovec copies.
 */

#include <linux/dmaengine.h>
#include <linux/pagemap.h>
#include <net/tcp.h> /* for memcpy_toiovec */
#include <asm/io.h>
#include <asm/uaccess.h>

static int num_pages_spanned(struct iovec *iov)
{
	return
	((PAGE_ALIGN((unsigned long)iov->iov_base + iov->iov_len) -
	((unsigned long)iov->iov_base & PAGE_MASK)) >> PAGE_SHIFT);
}

/*
 * Pin down all the iovec pages needed for len bytes.
 * Return a struct dma_pinned_list to keep track of pages pinned down.
 *
 * We are allocating a single chunk of memory, and then carving it up into
 * 3 sections, the latter 2 whose size depends on the number of iovecs and the
 * total number of pages, respectively.
 */
struct dma_pinned_list *dma_pin_iovec_pages(struct iovec *iov, size_t len)
{
	struct dma_pinned_list *local_list;
	struct page **pages;
	int i;
	int ret;
	int nr_iovecs = 0;
	int iovec_len_used = 0;
	int iovec_pages_used = 0;

	/* don't pin down non-user-based iovecs */
	if (segment_eq(get_fs(), KERNEL_DS))
		return NULL;

	/* determine how many iovecs/pages there are, up front */
	do {
		iovec_len_used += iov[nr_iovecs].iov_len;
		iovec_pages_used += num_pages_spanned(&iov[nr_iovecs]);
		nr_iovecs++;
	} while (iovec_len_used < len);

	/* single kmalloc for pinned list, page_list[], and the page arrays */
	local_list = kmalloc(sizeof(*local_list)
		+ (nr_iovecs * sizeof (struct dma_page_list))
		+ (iovec_pages_used * sizeof (struct page*)), GFP_KERNEL);
	if (!local_list)
		goto out;

	/* list of pages starts right after the page list array */
	pages = (struct page **) &local_list->page_list[nr_iovecs];

	local_list->nr_iovecs = 0;

	for (i = 0; i < nr_iovecs; i++) {
		struct dma_page_list *page_list = &local_list->page_list[i];

		len -= iov[i].iov_len;

		if (!access_ok(VERIFY_WRITE, iov[i].iov_base, iov[i].iov_len))
			goto unpin;

		page_list->nr_pages = num_pages_spanned(&iov[i]);
		page_list->base_address = iov[i].iov_base;

		page_list->pages = pages;
		pages += page_list->nr_pages;

		/* pin pages down */
		down_read(&current->mm->mmap_sem);
		ret = get_user_pages(
			current,
			current->mm,
			(unsigned long) iov[i].iov_base,
			page_list->nr_pages,
			1,	/* write */
			0,	/* force */
			page_list->pages,
			NULL);
		up_read(&current->mm->mmap_sem);

		if (ret != page_list->nr_pages)
			goto unpin;

		local_list->nr_iovecs = i + 1;
	}

	return local_list;

unpin:
	dma_unpin_iovec_pages(local_list);
out:
	return NULL;
}

void dma_unpin_iovec_pages(struct dma_pinned_list *pinned_list)
{
	int i, j;

	if (!pinned_list)
		return;

	for (i = 0; i < pinned_list->nr_iovecs; i++) {
		struct dma_page_list *page_list = &pinned_list->page_list[i];
		for (j = 0; j < page_list->nr_pages; j++) {
			set_page_dirty_lock(page_list->pages[j]);
			page_cache_release(page_list->pages[j]);
		}
	}

	kfree(pinned_list);
}


/*
 * We have already pinned down the pages we will be using in the iovecs.
 * Each entry in iov array has corresponding entry in pinned_list->page_list.
 * Using array indexing to keep iov[] and page_list[] in sync.
 * Initial elements in iov array's iov->iov_len will be 0 if already copied into
 *   by another call.
 * iov array length remaining guaranteed to be bigger than len.
 */
dma_cookie_t dma_memcpy_to_iovec(struct dma_chan *chan, struct iovec *iov,
	struct dma_pinned_list *pinned_list, unsigned char *kdata, size_t len)
{
	int iov_byte_offset;
	int copy;
	dma_cookie_t dma_cookie = 0;
	int iovec_idx;
	int page_idx;

	if (!chan)
		return memcpy_toiovec(iov, kdata, len);

	iovec_idx = 0;
	while (iovec_idx < pinned_list->nr_iovecs) {
		struct dma_page_list *page_list;

		/* skip already used-up iovecs */
		while (!iov[iovec_idx].iov_len)
			iovec_idx++;

		page_list = &pinned_list->page_list[iovec_idx];

		iov_byte_offset = ((unsigned long)iov[iovec_idx].iov_base & ~PAGE_MASK);
		page_idx = (((unsigned long)iov[iovec_idx].iov_base & PAGE_MASK)
			 - ((unsigned long)page_list->base_address & PAGE_MASK)) >> PAGE_SHIFT;

		/* break up copies to not cross page boundary */
		while (iov[iovec_idx].iov_len) {
			copy = min_t(int, PAGE_SIZE - iov_byte_offset, len);
			copy = min_t(int, copy, iov[iovec_idx].iov_len);

			dma_cookie = dma_async_memcpy_buf_to_pg(chan,
					page_list->pages[page_idx],
					iov_byte_offset,
					kdata,
					copy);
			/* poll for a descriptor slot */
			if (unlikely(dma_cookie < 0)) {
				dma_async_issue_pending(chan);
				continue;
			}

			len -= copy;
			iov[iovec_idx].iov_len -= copy;
			iov[iovec_idx].iov_base += copy;

			if (!len)
				return dma_cookie;

			kdata += copy;
			iov_byte_offset = 0;
			page_idx++;
		}
		iovec_idx++;
	}

	/* really bad if we ever run out of iovecs */
	BUG();
	return -EFAULT;
}

dma_cookie_t dma_memcpy_pg_to_iovec(struct dma_chan *chan, struct iovec *iov,
	struct dma_pinned_list *pinned_list, struct page *page,
	unsigned int offset, size_t len)
{
	int iov_byte_offset;
	int copy;
	dma_cookie_t dma_cookie = 0;
	int iovec_idx;
	int page_idx;
	int err;

	/* this needs as-yet-unimplemented buf-to-buff, so punt. */
	/* TODO: use dma for this */
	if (!chan || !pinned_list) {
		u8 *vaddr = kmap(page);
		err = memcpy_toiovec(iov, vaddr + offset, len);
		kunmap(page);
		return err;
	}

	iovec_idx = 0;
	while (iovec_idx < pinned_list->nr_iovecs) {
		struct dma_page_list *page_list;

		/* skip already used-up iovecs */
		while (!iov[iovec_idx].iov_len)
			iovec_idx++;

		page_list = &pinned_list->page_list[iovec_idx];

		iov_byte_offset = ((unsigned long)iov[iovec_idx].iov_base & ~PAGE_MASK);
		page_idx = (((unsigned long)iov[iovec_idx].iov_base & PAGE_MASK)
			 - ((unsigned long)page_list->base_address & PAGE_MASK)) >> PAGE_SHIFT;

		/* break up copies to not cross page boundary */
		while (iov[iovec_idx].iov_len) {
			copy = min_t(int, PAGE_SIZE - iov_byte_offset, len);
			copy = min_t(int, copy, iov[iovec_idx].iov_len);

			dma_cookie = dma_async_memcpy_pg_to_pg(chan,
					page_list->pages[page_idx],
					iov_byte_offset,
					page,
					offset,
					copy);
			/* poll for a descriptor slot */
			if (unlikely(dma_cookie < 0)) {
				dma_async_issue_pending(chan);
				continue;
			}

			len -= copy;
			iov[iovec_idx].iov_len -= copy;
			iov[iovec_idx].iov_base += copy;

			if (!len)
				return dma_cookie;

			offset += copy;
			iov_byte_offset = 0;
			page_idx++;
		}
		iovec_idx++;
	}

	/* really bad if we ever run out of iovecs */
	BUG();
	return -EFAULT;
}
Commit	Line	Data
de5506e1 CL	1	/*
	2	* Copyright(c) 2004 - 2006 Intel Corporation. All rights reserved.
	3	* Portions based on net/core/datagram.c and copyrighted by their authors.
	4	*
	5	* This program is free software; you can redistribute it and/or modify it
	6	* under the terms of the GNU General Public License as published by the Free
	7	* Software Foundation; either version 2 of the License, or (at your option)
	8	* any later version.
	9	*
	10	* This program is distributed in the hope that it will be useful, but WITHOUT
	11	* ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
	12	* FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for
	13	* more details.
	14	*
	15	* You should have received a copy of the GNU General Public License along with
	16	* this program; if not, write to the Free Software Foundation, Inc., 59
	17	* Temple Place - Suite 330, Boston, MA 02111-1307, USA.
	18	*
	19	* The full GNU General Public License is included in this distribution in the
	20	* file called COPYING.
	21	*/
	22
	23	/*
	24	* This code allows the net stack to make use of a DMA engine for
	25	* skb to iovec copies.
	26	*/
	27
	28	#include <linux/dmaengine.h>
	29	#include <linux/pagemap.h>
	30	#include <net/tcp.h> /* for memcpy_toiovec */
	31	#include <asm/io.h>
	32	#include <asm/uaccess.h>
	33
56e0873b	34	static int num_pages_spanned(struct iovec *iov)
de5506e1 CL	35	{
	36	return
	37	((PAGE_ALIGN((unsigned long)iov->iov_base + iov->iov_len) -
	38	((unsigned long)iov->iov_base & PAGE_MASK)) >> PAGE_SHIFT);
	39	}
	40
	41	/*
	42	* Pin down all the iovec pages needed for len bytes.
	43	* Return a struct dma_pinned_list to keep track of pages pinned down.
	44	*
	45	* We are allocating a single chunk of memory, and then carving it up into
	46	* 3 sections, the latter 2 whose size depends on the number of iovecs and the
	47	* total number of pages, respectively.
	48	*/
	49	struct dma_pinned_list dma_pin_iovec_pages(struct iovec iov, size_t len)
	50	{
	51	struct dma_pinned_list *local_list;
	52	struct page **pages;
	53	int i;
	54	int ret;
	55	int nr_iovecs = 0;
	56	int iovec_len_used = 0;
	57	int iovec_pages_used = 0;
de5506e1 CL	58
	59	/* don't pin down non-user-based iovecs */
	60	if (segment_eq(get_fs(), KERNEL_DS))
	61	return NULL;
	62
	63	/* determine how many iovecs/pages there are, up front */
	64	do {
	65	iovec_len_used += iov[nr_iovecs].iov_len;
	66	iovec_pages_used += num_pages_spanned(&iov[nr_iovecs]);
	67	nr_iovecs++;
	68	} while (iovec_len_used < len);
	69
	70	/* single kmalloc for pinned list, page_list[], and the page arrays */
	71	local_list = kmalloc(sizeof(*local_list)
	72	+ (nr_iovecs * sizeof (struct dma_page_list))
	73	+ (iovec_pages_used * sizeof (struct page*)), GFP_KERNEL);
c2c0b4c5	74	if (!local_list)
de5506e1	75	goto out;
de5506e1 CL	76
	77	/* list of pages starts right after the page list array */
	78	pages = (struct page **) &local_list->page_list[nr_iovecs];
	79
c2c0b4c5 MS	80	local_list->nr_iovecs = 0;
c2c0b4c5 MS	81
de5506e1 CL	82	for (i = 0; i < nr_iovecs; i++) {
	83	struct dma_page_list *page_list = &local_list->page_list[i];
	84
	85	len -= iov[i].iov_len;
	86
c2c0b4c5	87	if (!access_ok(VERIFY_WRITE, iov[i].iov_base, iov[i].iov_len))
de5506e1	88	goto unpin;
de5506e1 CL	89
	90	page_list->nr_pages = num_pages_spanned(&iov[i]);
	91	page_list->base_address = iov[i].iov_base;
	92
	93	page_list->pages = pages;
	94	pages += page_list->nr_pages;
	95
	96	/* pin pages down */
	97	down_read(&current->mm->mmap_sem);
	98	ret = get_user_pages(
	99	current,
	100	current->mm,
	101	(unsigned long) iov[i].iov_base,
	102	page_list->nr_pages,
	103	1, /* write */
	104	0, /* force */
	105	page_list->pages,
	106	NULL);
	107	up_read(&current->mm->mmap_sem);
	108
c2c0b4c5	109	if (ret != page_list->nr_pages)
de5506e1	110	goto unpin;
de5506e1 CL	111
	112	local_list->nr_iovecs = i + 1;
	113	}
	114
	115	return local_list;
	116
	117	unpin:
	118	dma_unpin_iovec_pages(local_list);
	119	out:
c2c0b4c5	120	return NULL;
de5506e1 CL	121	}
	122
	123	void dma_unpin_iovec_pages(struct dma_pinned_list *pinned_list)
	124	{
	125	int i, j;
	126
	127	if (!pinned_list)
	128	return;
	129
	130	for (i = 0; i < pinned_list->nr_iovecs; i++) {
	131	struct dma_page_list *page_list = &pinned_list->page_list[i];
	132	for (j = 0; j < page_list->nr_pages; j++) {
	133	set_page_dirty_lock(page_list->pages[j]);
	134	page_cache_release(page_list->pages[j]);
	135	}
	136	}
	137
	138	kfree(pinned_list);
	139	}
	140
de5506e1 CL	141
	142	/*
	143	* We have already pinned down the pages we will be using in the iovecs.
	144	* Each entry in iov array has corresponding entry in pinned_list->page_list.
	145	* Using array indexing to keep iov[] and page_list[] in sync.
	146	* Initial elements in iov array's iov->iov_len will be 0 if already copied into
	147	* by another call.
	148	* iov array length remaining guaranteed to be bigger than len.
	149	*/
	150	dma_cookie_t dma_memcpy_to_iovec(struct dma_chan chan, struct iovec iov,
	151	struct dma_pinned_list pinned_list, unsigned char kdata, size_t len)
	152	{
	153	int iov_byte_offset;
	154	int copy;
	155	dma_cookie_t dma_cookie = 0;
	156	int iovec_idx;
	157	int page_idx;
	158
	159	if (!chan)
	160	return memcpy_toiovec(iov, kdata, len);
	161
de5506e1 CL	162	iovec_idx = 0;
	163	while (iovec_idx < pinned_list->nr_iovecs) {
	164	struct dma_page_list *page_list;
	165
	166	/* skip already used-up iovecs */
	167	while (!iov[iovec_idx].iov_len)
	168	iovec_idx++;
	169
	170	page_list = &pinned_list->page_list[iovec_idx];
	171
	172	iov_byte_offset = ((unsigned long)iov[iovec_idx].iov_base & ~PAGE_MASK);
	173	page_idx = (((unsigned long)iov[iovec_idx].iov_base & PAGE_MASK)
	174	- ((unsigned long)page_list->base_address & PAGE_MASK)) >> PAGE_SHIFT;
	175
	176	/* break up copies to not cross page boundary */
	177	while (iov[iovec_idx].iov_len) {
	178	copy = min_t(int, PAGE_SIZE - iov_byte_offset, len);
	179	copy = min_t(int, copy, iov[iovec_idx].iov_len);
	180
	181	dma_cookie = dma_async_memcpy_buf_to_pg(chan,
	182	page_list->pages[page_idx],
	183	iov_byte_offset,
	184	kdata,
	185	copy);
4b652f0d DW	186	/* poll for a descriptor slot */
	187	if (unlikely(dma_cookie < 0)) {
	188	dma_async_issue_pending(chan);
	189	continue;
	190	}
de5506e1 CL	191
	192	len -= copy;
	193	iov[iovec_idx].iov_len -= copy;
	194	iov[iovec_idx].iov_base += copy;
	195
	196	if (!len)
	197	return dma_cookie;
	198
	199	kdata += copy;
	200	iov_byte_offset = 0;
	201	page_idx++;
	202	}
	203	iovec_idx++;
	204	}
	205
	206	/* really bad if we ever run out of iovecs */
	207	BUG();
	208	return -EFAULT;
	209	}
	210
	211	dma_cookie_t dma_memcpy_pg_to_iovec(struct dma_chan chan, struct iovec iov,
	212	struct dma_pinned_list pinned_list, struct page page,
	213	unsigned int offset, size_t len)
	214	{
	215	int iov_byte_offset;
	216	int copy;
	217	dma_cookie_t dma_cookie = 0;
	218	int iovec_idx;
	219	int page_idx;
	220	int err;
	221
	222	/* this needs as-yet-unimplemented buf-to-buff, so punt. */
	223	/* TODO: use dma for this */
	224	if (!chan \|\| !pinned_list) {
	225	u8 *vaddr = kmap(page);
	226	err = memcpy_toiovec(iov, vaddr + offset, len);
	227	kunmap(page);
	228	return err;
	229	}
	230
	231	iovec_idx = 0;
	232	while (iovec_idx < pinned_list->nr_iovecs) {
	233	struct dma_page_list *page_list;
	234
	235	/* skip already used-up iovecs */
	236	while (!iov[iovec_idx].iov_len)
	237	iovec_idx++;
	238
	239	page_list = &pinned_list->page_list[iovec_idx];
	240
	241	iov_byte_offset = ((unsigned long)iov[iovec_idx].iov_base & ~PAGE_MASK);
	242	page_idx = (((unsigned long)iov[iovec_idx].iov_base & PAGE_MASK)
	243	- ((unsigned long)page_list->base_address & PAGE_MASK)) >> PAGE_SHIFT;
	244
	245	/* break up copies to not cross page boundary */
	246	while (iov[iovec_idx].iov_len) {
	247	copy = min_t(int, PAGE_SIZE - iov_byte_offset, len);
	248	copy = min_t(int, copy, iov[iovec_idx].iov_len);
	249
	250	dma_cookie = dma_async_memcpy_pg_to_pg(chan,
	251	page_list->pages[page_idx],
	252	iov_byte_offset,
	253	page,
	254	offset,
255	copy);
4b652f0d DW	256	/* poll for a descriptor slot */
	257	if (unlikely(dma_cookie < 0)) {
	258	dma_async_issue_pending(chan);
	259	continue;
	260	}
de5506e1 CL	261
	262	len -= copy;
	263	iov[iovec_idx].iov_len -= copy;
	264	iov[iovec_idx].iov_base += copy;
	265
	266	if (!len)
	267	return dma_cookie;
	268
	269	offset += copy;
	270	iov_byte_offset = 0;
	271	page_idx++;
	272	}
	273	iovec_idx++;
	274	}
	275
	276	/* really bad if we ever run out of iovecs */
	277	BUG();
	278	return -EFAULT;
	279	}