[mirror_ubuntu-zesty-kernel.git] / crypto / async_tx / async_xor.c

/*
 * xor offload engine api
 *
 * Copyright © 2006, Intel Corporation.
 *
 *      Dan Williams <dan.j.williams@intel.com>
 *
 *      with architecture considerations by:
 *      Neil Brown <neilb@suse.de>
 *      Jeff Garzik <jeff@garzik.org>
 *
 * This program is free software; you can redistribute it and/or modify it
 * under the terms and conditions of the GNU General Public License,
 * version 2, as published by the Free Software Foundation.
 *
 * This program is distributed in the hope 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.,
 * 51 Franklin St - Fifth Floor, Boston, MA 02110-1301 USA.
 *
 */
#include <linux/kernel.h>
#include <linux/interrupt.h>
#include <linux/mm.h>
#include <linux/dma-mapping.h>
#include <linux/raid/xor.h>
#include <linux/async_tx.h>

/* do_async_xor - dma map the pages and perform the xor with an engine.
 * 	This routine is marked __always_inline so it can be compiled away
 * 	when CONFIG_DMA_ENGINE=n
 */
static __always_inline struct dma_async_tx_descriptor *
do_async_xor(struct dma_chan *chan, struct page *dest, struct page **src_list,
	     unsigned int offset, int src_cnt, size_t len,
	     enum async_tx_flags flags,
	     struct dma_async_tx_descriptor *depend_tx,
	     dma_async_tx_callback cb_fn, void *cb_param)
{
	struct dma_device *dma = chan->device;
	dma_addr_t *dma_src = (dma_addr_t *) src_list;
	struct dma_async_tx_descriptor *tx = NULL;
	int src_off = 0;
	int i;
	dma_async_tx_callback _cb_fn;
	void *_cb_param;
	enum async_tx_flags async_flags;
	enum dma_ctrl_flags dma_flags;
	int xor_src_cnt;
	dma_addr_t dma_dest;

	dma_dest = dma_map_page(dma->dev, dest, offset, len, DMA_FROM_DEVICE);
	for (i = 0; i < src_cnt; i++)
		dma_src[i] = dma_map_page(dma->dev, src_list[i], offset,
					  len, DMA_TO_DEVICE);

	while (src_cnt) {
		async_flags = flags;
		dma_flags = 0;
		xor_src_cnt = min(src_cnt, dma->max_xor);
		/* if we are submitting additional xors, leave the chain open,
		 * clear the callback parameters, and leave the destination
		 * buffer mapped
		 */
		if (src_cnt > xor_src_cnt) {
			async_flags &= ~ASYNC_TX_ACK;
			dma_flags = DMA_COMPL_SKIP_DEST_UNMAP;
			_cb_fn = NULL;
			_cb_param = NULL;
		} else {
			_cb_fn = cb_fn;
			_cb_param = cb_param;
		}
		if (_cb_fn)
			dma_flags |= DMA_PREP_INTERRUPT;

		/* Since we have clobbered the src_list we are committed
		 * to doing this asynchronously.  Drivers force forward progress
		 * in case they can not provide a descriptor
		 */
		tx = dma->device_prep_dma_xor(chan, dma_dest, &dma_src[src_off],
					      xor_src_cnt, len, dma_flags);

		if (unlikely(!tx && depend_tx))
			dma_wait_for_async_tx(depend_tx);

		/* spin wait for the preceeding transactions to complete */
		while (unlikely(!tx))
			tx = dma->device_prep_dma_xor(chan, dma_dest,
						      &dma_src[src_off],
						      xor_src_cnt, len,
						      dma_flags);

		async_tx_submit(chan, tx, async_flags, depend_tx, _cb_fn,
				_cb_param);

		depend_tx = tx;
		flags |= ASYNC_TX_DEP_ACK;

		if (src_cnt > xor_src_cnt) {
			/* drop completed sources */
			src_cnt -= xor_src_cnt;
			src_off += xor_src_cnt;

			/* use the intermediate result a source */
			dma_src[--src_off] = dma_dest;
			src_cnt++;
		} else
			break;
	}

	return tx;
}

static void
do_sync_xor(struct page *dest, struct page **src_list, unsigned int offset,
	    int src_cnt, size_t len, enum async_tx_flags flags,
	    struct dma_async_tx_descriptor *depend_tx,
	    dma_async_tx_callback cb_fn, void *cb_param)
{
	int i;
	int xor_src_cnt;
	int src_off = 0;
	void *dest_buf;
	void **srcs = (void **) src_list;

	/* reuse the 'src_list' array to convert to buffer pointers */
	for (i = 0; i < src_cnt; i++)
		srcs[i] = page_address(src_list[i]) + offset;

	/* set destination address */
	dest_buf = page_address(dest) + offset;

	if (flags & ASYNC_TX_XOR_ZERO_DST)
		memset(dest_buf, 0, len);

	while (src_cnt > 0) {
		/* process up to 'MAX_XOR_BLOCKS' sources */
		xor_src_cnt = min(src_cnt, MAX_XOR_BLOCKS);
		xor_blocks(xor_src_cnt, len, dest_buf, &srcs[src_off]);

		/* drop completed sources */
		src_cnt -= xor_src_cnt;
		src_off += xor_src_cnt;
	}

	async_tx_sync_epilog(flags, depend_tx, cb_fn, cb_param);
}

/**
 * async_xor - attempt to xor a set of blocks with a dma engine.
 *	xor_blocks always uses the dest as a source so the ASYNC_TX_XOR_ZERO_DST
 *	flag must be set to not include dest data in the calculation.  The
 *	assumption with dma eninges is that they only use the destination
 *	buffer as a source when it is explicity specified in the source list.
 * @dest: destination page
 * @src_list: array of source pages (if the dest is also a source it must be
 *	at index zero).  The contents of this array may be overwritten.
 * @offset: offset in pages to start transaction
 * @src_cnt: number of source pages
 * @len: length in bytes
 * @flags: ASYNC_TX_XOR_ZERO_DST, ASYNC_TX_XOR_DROP_DEST,
 *	ASYNC_TX_ACK, ASYNC_TX_DEP_ACK
 * @depend_tx: xor depends on the result of this transaction.
 * @cb_fn: function to call when the xor completes
 * @cb_param: parameter to pass to the callback routine
 */
struct dma_async_tx_descriptor *
async_xor(struct page *dest, struct page **src_list, unsigned int offset,
	int src_cnt, size_t len, enum async_tx_flags flags,
	struct dma_async_tx_descriptor *depend_tx,
	dma_async_tx_callback cb_fn, void *cb_param)
{
	struct dma_chan *chan = async_tx_find_channel(depend_tx, DMA_XOR,
						      &dest, 1, src_list,
						      src_cnt, len);
	BUG_ON(src_cnt <= 1);

	if (chan) {
		/* run the xor asynchronously */
		pr_debug("%s (async): len: %zu\n", __func__, len);

		return do_async_xor(chan, dest, src_list, offset, src_cnt, len,
				    flags, depend_tx, cb_fn, cb_param);
	} else {
		/* run the xor synchronously */
		pr_debug("%s (sync): len: %zu\n", __func__, len);

		/* in the sync case the dest is an implied source
		 * (assumes the dest is the first source)
		 */
		if (flags & ASYNC_TX_XOR_DROP_DST) {
			src_cnt--;
			src_list++;
		}

		/* wait for any prerequisite operations */
		if (depend_tx) {
			/* if ack is already set then we cannot be sure
			 * we are referring to the correct operation
			 */
			BUG_ON(async_tx_test_ack(depend_tx));
			if (dma_wait_for_async_tx(depend_tx) == DMA_ERROR)
				panic("%s: DMA_ERROR waiting for depend_tx\n",
					__func__);
		}

		do_sync_xor(dest, src_list, offset, src_cnt, len,
			    flags, depend_tx, cb_fn, cb_param);

		return NULL;
	}
}
EXPORT_SYMBOL_GPL(async_xor);

static int page_is_zero(struct page *p, unsigned int offset, size_t len)
{
	char *a = page_address(p) + offset;
	return ((*(u32 *) a) == 0 &&
		memcmp(a, a + 4, len - 4) == 0);
}

/**
 * async_xor_zero_sum - attempt a xor parity check with a dma engine.
 * @dest: destination page used if the xor is performed synchronously
 * @src_list: array of source pages.  The dest page must be listed as a source
 * 	at index zero.  The contents of this array may be overwritten.
 * @offset: offset in pages to start transaction
 * @src_cnt: number of source pages
 * @len: length in bytes
 * @result: 0 if sum == 0 else non-zero
 * @flags: ASYNC_TX_ACK, ASYNC_TX_DEP_ACK
 * @depend_tx: xor depends on the result of this transaction.
 * @cb_fn: function to call when the xor completes
 * @cb_param: parameter to pass to the callback routine
 */
struct dma_async_tx_descriptor *
async_xor_zero_sum(struct page *dest, struct page **src_list,
	unsigned int offset, int src_cnt, size_t len,
	u32 *result, enum async_tx_flags flags,
	struct dma_async_tx_descriptor *depend_tx,
	dma_async_tx_callback cb_fn, void *cb_param)
{
	struct dma_chan *chan = async_tx_find_channel(depend_tx, DMA_ZERO_SUM,
						      &dest, 1, src_list,
						      src_cnt, len);
	struct dma_device *device = chan ? chan->device : NULL;
	struct dma_async_tx_descriptor *tx = NULL;

	BUG_ON(src_cnt <= 1);

	if (device && src_cnt <= device->max_xor) {
		dma_addr_t *dma_src = (dma_addr_t *) src_list;
		unsigned long dma_prep_flags = cb_fn ? DMA_PREP_INTERRUPT : 0;
		int i;

		pr_debug("%s: (async) len: %zu\n", __func__, len);

		for (i = 0; i < src_cnt; i++)
			dma_src[i] = dma_map_page(device->dev, src_list[i],
						  offset, len, DMA_TO_DEVICE);

		tx = device->device_prep_dma_zero_sum(chan, dma_src, src_cnt,
						      len, result,
						      dma_prep_flags);
		if (!tx) {
			if (depend_tx)
				dma_wait_for_async_tx(depend_tx);

			while (!tx)
				tx = device->device_prep_dma_zero_sum(chan,
					dma_src, src_cnt, len, result,
					dma_prep_flags);
		}

		async_tx_submit(chan, tx, flags, depend_tx, cb_fn, cb_param);
	} else {
		unsigned long xor_flags = flags;

		pr_debug("%s: (sync) len: %zu\n", __func__, len);

		xor_flags |= ASYNC_TX_XOR_DROP_DST;
		xor_flags &= ~ASYNC_TX_ACK;

		tx = async_xor(dest, src_list, offset, src_cnt, len, xor_flags,
			depend_tx, NULL, NULL);

		if (tx) {
			if (dma_wait_for_async_tx(tx) == DMA_ERROR)
				panic("%s: DMA_ERROR waiting for tx\n",
					__func__);
			async_tx_ack(tx);
		}

		*result = page_is_zero(dest, offset, len) ? 0 : 1;

		tx = NULL;

		async_tx_sync_epilog(flags, depend_tx, cb_fn, cb_param);
	}

	return tx;
}
EXPORT_SYMBOL_GPL(async_xor_zero_sum);

static int __init async_xor_init(void)
{
	#ifdef CONFIG_DMA_ENGINE
	/* To conserve stack space the input src_list (array of page pointers)
	 * is reused to hold the array of dma addresses passed to the driver.
	 * This conversion is only possible when dma_addr_t is less than the
	 * the size of a pointer.  HIGHMEM64G is known to violate this
	 * assumption.
	 */
	BUILD_BUG_ON(sizeof(dma_addr_t) > sizeof(struct page *));
	#endif

	return 0;
}

static void __exit async_xor_exit(void)
{
	do { } while (0);
}

module_init(async_xor_init);
module_exit(async_xor_exit);

MODULE_AUTHOR("Intel Corporation");
MODULE_DESCRIPTION("asynchronous xor/xor-zero-sum api");
MODULE_LICENSE("GPL");
Commit	Line	Data
9bc89cd8 DW	1	/*
	2	* xor offload engine api
	3	*
	4	* Copyright © 2006, Intel Corporation.
	5	*
	6	* Dan Williams <dan.j.williams@intel.com>
	7	*
	8	* with architecture considerations by:
	9	* Neil Brown <neilb@suse.de>
	10	* Jeff Garzik <jeff@garzik.org>
	11	*
	12	* This program is free software; you can redistribute it and/or modify it
	13	* under the terms and conditions of the GNU General Public License,
	14	* version 2, as published by the Free Software Foundation.
	15	*
	16	* This program is distributed in the hope it will be useful, but WITHOUT
	17	* ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
	18	* FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for
	19	* more details.
	20	*
	21	* You should have received a copy of the GNU General Public License along with
	22	* this program; if not, write to the Free Software Foundation, Inc.,
	23	* 51 Franklin St - Fifth Floor, Boston, MA 02110-1301 USA.
	24	*
	25	*/
	26	#include <linux/kernel.h>
	27	#include <linux/interrupt.h>
	28	#include <linux/mm.h>
	29	#include <linux/dma-mapping.h>
	30	#include <linux/raid/xor.h>
	31	#include <linux/async_tx.h>
	32
1367a3d3 DW	33	/* do_async_xor - dma map the pages and perform the xor with an engine.
	34	* This routine is marked __always_inline so it can be compiled away
	35	* when CONFIG_DMA_ENGINE=n
	36	*/
0036731c	37	static __always_inline struct dma_async_tx_descriptor *
1e55db2d DW	38	do_async_xor(struct dma_chan chan, struct page dest, struct page **src_list,
	39	unsigned int offset, int src_cnt, size_t len,
	40	enum async_tx_flags flags,
	41	struct dma_async_tx_descriptor *depend_tx,
	42	dma_async_tx_callback cb_fn, void *cb_param)
9bc89cd8	43	{
1e55db2d	44	struct dma_device *dma = chan->device;
0036731c	45	dma_addr_t dma_src = (dma_addr_t ) src_list;
1e55db2d DW	46	struct dma_async_tx_descriptor *tx = NULL;
1e55db2d DW	47	int src_off = 0;
9bc89cd8	48	int i;
1e55db2d DW	49	dma_async_tx_callback _cb_fn;
	50	void *_cb_param;
	51	enum async_tx_flags async_flags;
	52	enum dma_ctrl_flags dma_flags;
	53	int xor_src_cnt;
	54	dma_addr_t dma_dest;
9bc89cd8	55
1e55db2d	56	dma_dest = dma_map_page(dma->dev, dest, offset, len, DMA_FROM_DEVICE);
0036731c	57	for (i = 0; i < src_cnt; i++)
1e55db2d	58	dma_src[i] = dma_map_page(dma->dev, src_list[i], offset,
0036731c DW	59	len, DMA_TO_DEVICE);
0036731c DW	60
1e55db2d DW	61	while (src_cnt) {
	62	async_flags = flags;
	63	dma_flags = 0;
	64	xor_src_cnt = min(src_cnt, dma->max_xor);
	65	/* if we are submitting additional xors, leave the chain open,
	66	* clear the callback parameters, and leave the destination
	67	* buffer mapped
	68	*/
	69	if (src_cnt > xor_src_cnt) {
	70	async_flags &= ~ASYNC_TX_ACK;
	71	dma_flags = DMA_COMPL_SKIP_DEST_UNMAP;
	72	_cb_fn = NULL;
	73	_cb_param = NULL;
	74	} else {
	75	_cb_fn = cb_fn;
	76	_cb_param = cb_param;
	77	}
	78	if (_cb_fn)
	79	dma_flags \|= DMA_PREP_INTERRUPT;
	80
	81	/* Since we have clobbered the src_list we are committed
	82	* to doing this asynchronously. Drivers force forward progress
	83	* in case they can not provide a descriptor
	84	*/
	85	tx = dma->device_prep_dma_xor(chan, dma_dest, &dma_src[src_off],
	86	xor_src_cnt, len, dma_flags);
	87
	88	if (unlikely(!tx && depend_tx))
0036731c DW	89	dma_wait_for_async_tx(depend_tx);
0036731c DW	90
1e55db2d DW	91	/* spin wait for the preceeding transactions to complete */
	92	while (unlikely(!tx))
	93	tx = dma->device_prep_dma_xor(chan, dma_dest,
	94	&dma_src[src_off],
	95	xor_src_cnt, len,
	96	dma_flags);
9bc89cd8	97
1e55db2d DW	98	async_tx_submit(chan, tx, async_flags, depend_tx, _cb_fn,
	99	_cb_param);
	100
	101	depend_tx = tx;
	102	flags \|= ASYNC_TX_DEP_ACK;
	103
	104	if (src_cnt > xor_src_cnt) {
	105	/* drop completed sources */
	106	src_cnt -= xor_src_cnt;
	107	src_off += xor_src_cnt;
	108
	109	/* use the intermediate result a source */
	110	dma_src[--src_off] = dma_dest;
	111	src_cnt++;
	112	} else
	113	break;
	114	}
0036731c DW	115
0036731c DW	116	return tx;
9bc89cd8 DW	117	}
	118
	119	static void
	120	do_sync_xor(struct page dest, struct page *src_list, unsigned int offset,
1e55db2d DW	121	int src_cnt, size_t len, enum async_tx_flags flags,
	122	struct dma_async_tx_descriptor *depend_tx,
	123	dma_async_tx_callback cb_fn, void *cb_param)
9bc89cd8	124	{
9bc89cd8	125	int i;
1e55db2d DW	126	int xor_src_cnt;
	127	int src_off = 0;
	128	void *dest_buf;
	129	void srcs = (void ) src_list;
9bc89cd8 DW	130
	131	/* reuse the 'src_list' array to convert to buffer pointers */
	132	for (i = 0; i < src_cnt; i++)
1e55db2d	133	srcs[i] = page_address(src_list[i]) + offset;
9bc89cd8 DW	134
9bc89cd8 DW	135	/* set destination address */
1e55db2d	136	dest_buf = page_address(dest) + offset;
9bc89cd8 DW	137
9bc89cd8 DW	138	if (flags & ASYNC_TX_XOR_ZERO_DST)
1e55db2d	139	memset(dest_buf, 0, len);
9bc89cd8	140
1e55db2d DW	141	while (src_cnt > 0) {
	142	/* process up to 'MAX_XOR_BLOCKS' sources */
	143	xor_src_cnt = min(src_cnt, MAX_XOR_BLOCKS);
	144	xor_blocks(xor_src_cnt, len, dest_buf, &srcs[src_off]);
	145
	146	/* drop completed sources */
	147	src_cnt -= xor_src_cnt;
	148	src_off += xor_src_cnt;
	149	}
9bc89cd8 DW	150
	151	async_tx_sync_epilog(flags, depend_tx, cb_fn, cb_param);
	152	}
	153
	154	/**
	155	* async_xor - attempt to xor a set of blocks with a dma engine.
	156	* xor_blocks always uses the dest as a source so the ASYNC_TX_XOR_ZERO_DST
	157	* flag must be set to not include dest data in the calculation. The
	158	* assumption with dma eninges is that they only use the destination
	159	* buffer as a source when it is explicity specified in the source list.
	160	* @dest: destination page
	161	* @src_list: array of source pages (if the dest is also a source it must be
	162	* at index zero). The contents of this array may be overwritten.
	163	* @offset: offset in pages to start transaction
	164	* @src_cnt: number of source pages
	165	* @len: length in bytes
	166	* @flags: ASYNC_TX_XOR_ZERO_DST, ASYNC_TX_XOR_DROP_DEST,
d909b347	167	* ASYNC_TX_ACK, ASYNC_TX_DEP_ACK
9bc89cd8 DW	168	* @depend_tx: xor depends on the result of this transaction.
	169	* @cb_fn: function to call when the xor completes
	170	* @cb_param: parameter to pass to the callback routine
	171	*/
	172	struct dma_async_tx_descriptor *
	173	async_xor(struct page dest, struct page *src_list, unsigned int offset,
	174	int src_cnt, size_t len, enum async_tx_flags flags,
	175	struct dma_async_tx_descriptor *depend_tx,
	176	dma_async_tx_callback cb_fn, void *cb_param)
	177	{
47437b2c DW	178	struct dma_chan *chan = async_tx_find_channel(depend_tx, DMA_XOR,
	179	&dest, 1, src_list,
	180	src_cnt, len);
9bc89cd8 DW	181	BUG_ON(src_cnt <= 1);
9bc89cd8 DW	182
1e55db2d DW	183	if (chan) {
	184	/* run the xor asynchronously */
	185	pr_debug("%s (async): len: %zu\n", __func__, len);
9bc89cd8	186
1e55db2d DW	187	return do_async_xor(chan, dest, src_list, offset, src_cnt, len,
	188	flags, depend_tx, cb_fn, cb_param);
	189	} else {
	190	/* run the xor synchronously */
	191	pr_debug("%s (sync): len: %zu\n", __func__, len);
9bc89cd8	192
1e55db2d DW	193	/* in the sync case the dest is an implied source
1e55db2d DW	194	* (assumes the dest is the first source)
9bc89cd8	195	*/
1e55db2d DW	196	if (flags & ASYNC_TX_XOR_DROP_DST) {
	197	src_cnt--;
	198	src_list++;
	199	}
9bc89cd8	200
1e55db2d DW	201	/* wait for any prerequisite operations */
	202	if (depend_tx) {
	203	/* if ack is already set then we cannot be sure
	204	* we are referring to the correct operation
	205	*/
	206	BUG_ON(async_tx_test_ack(depend_tx));
	207	if (dma_wait_for_async_tx(depend_tx) == DMA_ERROR)
	208	panic("%s: DMA_ERROR waiting for depend_tx\n",
	209	__func__);
	210	}
9bc89cd8	211
1e55db2d DW	212	do_sync_xor(dest, src_list, offset, src_cnt, len,
1e55db2d DW	213	flags, depend_tx, cb_fn, cb_param);
9bc89cd8	214
1e55db2d	215	return NULL;
9bc89cd8	216	}
9bc89cd8 DW	217	}
	218	EXPORT_SYMBOL_GPL(async_xor);
	219
	220	static int page_is_zero(struct page *p, unsigned int offset, size_t len)
	221	{
	222	char *a = page_address(p) + offset;
	223	return (((u32 ) a) == 0 &&
	224	memcmp(a, a + 4, len - 4) == 0);
	225	}
	226
	227	/**
	228	* async_xor_zero_sum - attempt a xor parity check with a dma engine.
	229	* @dest: destination page used if the xor is performed synchronously
	230	* @src_list: array of source pages. The dest page must be listed as a source
	231	* at index zero. The contents of this array may be overwritten.
	232	* @offset: offset in pages to start transaction
	233	* @src_cnt: number of source pages
	234	* @len: length in bytes
	235	* @result: 0 if sum == 0 else non-zero
d909b347	236	* @flags: ASYNC_TX_ACK, ASYNC_TX_DEP_ACK
9bc89cd8 DW	237	* @depend_tx: xor depends on the result of this transaction.
	238	* @cb_fn: function to call when the xor completes
	239	* @cb_param: parameter to pass to the callback routine
	240	*/
	241	struct dma_async_tx_descriptor *
	242	async_xor_zero_sum(struct page dest, struct page *src_list,
	243	unsigned int offset, int src_cnt, size_t len,
	244	u32 *result, enum async_tx_flags flags,
	245	struct dma_async_tx_descriptor *depend_tx,
	246	dma_async_tx_callback cb_fn, void *cb_param)
	247	{
47437b2c DW	248	struct dma_chan *chan = async_tx_find_channel(depend_tx, DMA_ZERO_SUM,
	249	&dest, 1, src_list,
	250	src_cnt, len);
9bc89cd8	251	struct dma_device *device = chan ? chan->device : NULL;
0036731c	252	struct dma_async_tx_descriptor *tx = NULL;
9bc89cd8 DW	253
	254	BUG_ON(src_cnt <= 1);
	255
8d8002f6	256	if (device && src_cnt <= device->max_xor) {
0036731c	257	dma_addr_t dma_src = (dma_addr_t ) src_list;
d4c56f97	258	unsigned long dma_prep_flags = cb_fn ? DMA_PREP_INTERRUPT : 0;
0036731c	259	int i;
9bc89cd8	260
3280ab3e	261	pr_debug("%s: (async) len: %zu\n", __func__, len);
9bc89cd8	262
0036731c DW	263	for (i = 0; i < src_cnt; i++)
	264	dma_src[i] = dma_map_page(device->dev, src_list[i],
	265	offset, len, DMA_TO_DEVICE);
	266
	267	tx = device->device_prep_dma_zero_sum(chan, dma_src, src_cnt,
	268	len, result,
d4c56f97	269	dma_prep_flags);
0036731c DW	270	if (!tx) {
	271	if (depend_tx)
	272	dma_wait_for_async_tx(depend_tx);
	273
	274	while (!tx)
	275	tx = device->device_prep_dma_zero_sum(chan,
	276	dma_src, src_cnt, len, result,
d4c56f97	277	dma_prep_flags);
9bc89cd8 DW	278	}
	279
	280	async_tx_submit(chan, tx, flags, depend_tx, cb_fn, cb_param);
	281	} else {
	282	unsigned long xor_flags = flags;
	283
3280ab3e	284	pr_debug("%s: (sync) len: %zu\n", __func__, len);
9bc89cd8 DW	285
	286	xor_flags \|= ASYNC_TX_XOR_DROP_DST;
	287	xor_flags &= ~ASYNC_TX_ACK;
	288
	289	tx = async_xor(dest, src_list, offset, src_cnt, len, xor_flags,
	290	depend_tx, NULL, NULL);
	291
	292	if (tx) {
	293	if (dma_wait_for_async_tx(tx) == DMA_ERROR)
	294	panic("%s: DMA_ERROR waiting for tx\n",
3280ab3e	295	__func__);
9bc89cd8 DW	296	async_tx_ack(tx);
	297	}
	298
	299	*result = page_is_zero(dest, offset, len) ? 0 : 1;
	300
	301	tx = NULL;
	302
	303	async_tx_sync_epilog(flags, depend_tx, cb_fn, cb_param);
	304	}
	305
	306	return tx;
	307	}
	308	EXPORT_SYMBOL_GPL(async_xor_zero_sum);
	309
	310	static int __init async_xor_init(void)
	311	{
0036731c DW	312	#ifdef CONFIG_DMA_ENGINE
	313	/* To conserve stack space the input src_list (array of page pointers)
	314	* is reused to hold the array of dma addresses passed to the driver.
	315	* This conversion is only possible when dma_addr_t is less than the
	316	* the size of a pointer. HIGHMEM64G is known to violate this
	317	* assumption.
	318	*/
	319	BUILD_BUG_ON(sizeof(dma_addr_t) > sizeof(struct page *));
	320	#endif
	321
9bc89cd8 DW	322	return 0;
	323	}
	324
	325	static void __exit async_xor_exit(void)
	326	{
	327	do { } while (0);
	328	}
	329
	330	module_init(async_xor_init);
	331	module_exit(async_xor_exit);
	332
	333	MODULE_AUTHOR("Intel Corporation");
	334	MODULE_DESCRIPTION("asynchronous xor/xor-zero-sum api");
	335	MODULE_LICENSE("GPL");