[linux-2.6-block.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 void
do_async_xor(struct dma_async_tx_descriptor *tx, struct dma_device *device,
	struct dma_chan *chan, struct page *dest, struct page **src_list,
	unsigned int offset, unsigned 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)
{
	dma_addr_t dma_addr;
	int i;

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

	dma_addr = dma_map_page(device->dev, dest, offset, len,
				DMA_FROM_DEVICE);
	tx->tx_set_dest(dma_addr, tx, 0);

	for (i = 0; i < src_cnt; i++) {
		dma_addr = dma_map_page(device->dev, src_list[i],
			offset, len, DMA_TO_DEVICE);
		tx->tx_set_src(dma_addr, tx, i);
	}

	async_tx_submit(chan, tx, flags, depend_tx, cb_fn, cb_param);
}

static void
do_sync_xor(struct page *dest, struct page **src_list, unsigned int offset,
	unsigned 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)
{
	void *_dest;
	int i;

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

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

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

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

	xor_blocks(src_cnt, len, _dest,
		(void **) src_list);

	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);
	struct dma_device *device = chan ? chan->device : NULL;
	struct dma_async_tx_descriptor *tx = NULL;
	dma_async_tx_callback _cb_fn;
	void *_cb_param;
	unsigned long local_flags;
	int xor_src_cnt;
	int i = 0, src_off = 0, int_en;

	BUG_ON(src_cnt <= 1);

	while (src_cnt) {
		local_flags = flags;
		if (device) { /* run the xor asynchronously */
			xor_src_cnt = min(src_cnt, device->max_xor);
			/* if we are submitting additional xors
			 * only set the callback on the last transaction
			 */
			if (src_cnt > xor_src_cnt) {
				local_flags &= ~ASYNC_TX_ACK;
				_cb_fn = NULL;
				_cb_param = NULL;
			} else {
				_cb_fn = cb_fn;
				_cb_param = cb_param;
			}

			int_en = _cb_fn ? 1 : 0;

			tx = device->device_prep_dma_xor(
				chan, xor_src_cnt, len, int_en);

			if (tx) {
				do_async_xor(tx, device, chan, dest,
				&src_list[src_off], offset, xor_src_cnt, len,
				local_flags, depend_tx, _cb_fn,
				_cb_param);
			} else /* fall through */
				goto xor_sync;
		} else { /* run the xor synchronously */
xor_sync:
			/* in the sync case the dest is an implied source
			 * (assumes the dest is at the src_off index)
			 */
			if (flags & ASYNC_TX_XOR_DROP_DST) {
				src_cnt--;
				src_off++;
			}

			/* process up to 'MAX_XOR_BLOCKS' sources */
			xor_src_cnt = min(src_cnt, MAX_XOR_BLOCKS);

			/* if we are submitting additional xors
			 * only set the callback on the last transaction
			 */
			if (src_cnt > xor_src_cnt) {
				local_flags &= ~ASYNC_TX_ACK;
				_cb_fn = NULL;
				_cb_param = NULL;
			} else {
				_cb_fn = cb_fn;
				_cb_param = cb_param;
			}

			/* 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(depend_tx->ack);
				if (dma_wait_for_async_tx(depend_tx) ==
					DMA_ERROR)
					panic("%s: DMA_ERROR waiting for "
						"depend_tx\n",
						__FUNCTION__);
			}

			do_sync_xor(dest, &src_list[src_off], offset,
				xor_src_cnt, len, local_flags, depend_tx,
				_cb_fn, _cb_param);
		}

		/* the previous tx is hidden from the client,
		 * so ack it
		 */
		if (i && depend_tx)
			async_tx_ack(depend_tx);

		depend_tx = tx;

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

			/* unconditionally preserve the destination */
			flags &= ~ASYNC_TX_XOR_ZERO_DST;

			/* use the intermediate result a source, but remember
			 * it's dropped, because it's implied, in the sync case
			 */
			src_list[--src_off] = dest;
			src_cnt++;
			flags |= ASYNC_TX_XOR_DROP_DST;
		} else
			src_cnt = 0;
		i++;
	}

	return tx;
}
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);
	struct dma_device *device = chan ? chan->device : NULL;
	int int_en = cb_fn ? 1 : 0;
	struct dma_async_tx_descriptor *tx = device ?
		device->device_prep_dma_zero_sum(chan, src_cnt, len, result,
			int_en) : NULL;
	int i;

	BUG_ON(src_cnt <= 1);

	if (tx) {
		dma_addr_t dma_addr;

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

		for (i = 0; i < src_cnt; i++) {
			dma_addr = dma_map_page(device->dev, src_list[i],
				offset, len, DMA_TO_DEVICE);
			tx->tx_set_src(dma_addr, tx, i);
		}

		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", __FUNCTION__, 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",
					__FUNCTION__);
			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)
{
	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	*/
	37	static __always_inline void
9bc89cd8 DW	38	do_async_xor(struct dma_async_tx_descriptor tx, struct dma_device device,
	39	struct dma_chan chan, struct page dest, struct page **src_list,
	40	unsigned int offset, unsigned int src_cnt, size_t len,
	41	enum async_tx_flags flags, struct dma_async_tx_descriptor *depend_tx,
	42	dma_async_tx_callback cb_fn, void *cb_param)
	43	{
	44	dma_addr_t dma_addr;
9bc89cd8 DW	45	int i;
	46
	47	pr_debug("%s: len: %zu\n", __FUNCTION__, len);
	48
d909b347 DW	49	dma_addr = dma_map_page(device->dev, dest, offset, len,
d909b347 DW	50	DMA_FROM_DEVICE);
9bc89cd8 DW	51	tx->tx_set_dest(dma_addr, tx, 0);
9bc89cd8 DW	52
9bc89cd8 DW	53	for (i = 0; i < src_cnt; i++) {
9bc89cd8 DW	54	dma_addr = dma_map_page(device->dev, src_list[i],
d909b347	55	offset, len, DMA_TO_DEVICE);
9bc89cd8 DW	56	tx->tx_set_src(dma_addr, tx, i);
	57	}
	58
	59	async_tx_submit(chan, tx, flags, depend_tx, cb_fn, cb_param);
	60	}
	61
	62	static void
	63	do_sync_xor(struct page dest, struct page *src_list, unsigned int offset,
	64	unsigned int src_cnt, size_t len, enum async_tx_flags flags,
	65	struct dma_async_tx_descriptor *depend_tx,
	66	dma_async_tx_callback cb_fn, void *cb_param)
	67	{
	68	void *_dest;
	69	int i;
	70
	71	pr_debug("%s: len: %zu\n", __FUNCTION__, len);
	72
	73	/* reuse the 'src_list' array to convert to buffer pointers */
	74	for (i = 0; i < src_cnt; i++)
	75	src_list[i] = (struct page *)
	76	(page_address(src_list[i]) + offset);
	77
	78	/* set destination address */
	79	_dest = page_address(dest) + offset;
	80
	81	if (flags & ASYNC_TX_XOR_ZERO_DST)
	82	memset(_dest, 0, len);
	83
	84	xor_blocks(src_cnt, len, _dest,
	85	(void **) src_list);
	86
	87	async_tx_sync_epilog(flags, depend_tx, cb_fn, cb_param);
	88	}
	89
	90	/**
	91	* async_xor - attempt to xor a set of blocks with a dma engine.
	92	* xor_blocks always uses the dest as a source so the ASYNC_TX_XOR_ZERO_DST
	93	* flag must be set to not include dest data in the calculation. The
	94	* assumption with dma eninges is that they only use the destination
	95	* buffer as a source when it is explicity specified in the source list.
	96	* @dest: destination page
	97	* @src_list: array of source pages (if the dest is also a source it must be
	98	* at index zero). The contents of this array may be overwritten.
	99	* @offset: offset in pages to start transaction
	100	* @src_cnt: number of source pages
	101	* @len: length in bytes
	102	* @flags: ASYNC_TX_XOR_ZERO_DST, ASYNC_TX_XOR_DROP_DEST,
d909b347	103	* ASYNC_TX_ACK, ASYNC_TX_DEP_ACK
9bc89cd8 DW	104	* @depend_tx: xor depends on the result of this transaction.
	105	* @cb_fn: function to call when the xor completes
	106	* @cb_param: parameter to pass to the callback routine
	107	*/
	108	struct dma_async_tx_descriptor *
	109	async_xor(struct page dest, struct page *src_list, unsigned int offset,
	110	int src_cnt, size_t len, enum async_tx_flags flags,
	111	struct dma_async_tx_descriptor *depend_tx,
	112	dma_async_tx_callback cb_fn, void *cb_param)
	113	{
	114	struct dma_chan *chan = async_tx_find_channel(depend_tx, DMA_XOR);
	115	struct dma_device *device = chan ? chan->device : NULL;
	116	struct dma_async_tx_descriptor *tx = NULL;
	117	dma_async_tx_callback _cb_fn;
	118	void *_cb_param;
	119	unsigned long local_flags;
	120	int xor_src_cnt;
	121	int i = 0, src_off = 0, int_en;
	122
	123	BUG_ON(src_cnt <= 1);
	124
	125	while (src_cnt) {
	126	local_flags = flags;
	127	if (device) { /* run the xor asynchronously */
	128	xor_src_cnt = min(src_cnt, device->max_xor);
	129	/* if we are submitting additional xors
	130	* only set the callback on the last transaction
	131	*/
	132	if (src_cnt > xor_src_cnt) {
	133	local_flags &= ~ASYNC_TX_ACK;
	134	_cb_fn = NULL;
	135	_cb_param = NULL;
	136	} else {
	137	_cb_fn = cb_fn;
	138	_cb_param = cb_param;
	139	}
	140
	141	int_en = _cb_fn ? 1 : 0;
	142
	143	tx = device->device_prep_dma_xor(
	144	chan, xor_src_cnt, len, int_en);
	145
	146	if (tx) {
	147	do_async_xor(tx, device, chan, dest,
	148	&src_list[src_off], offset, xor_src_cnt, len,
	149	local_flags, depend_tx, _cb_fn,
	150	_cb_param);
	151	} else /* fall through */
	152	goto xor_sync;
	153	} else { /* run the xor synchronously */
	154	xor_sync:
	155	/* in the sync case the dest is an implied source
	156	* (assumes the dest is at the src_off index)
	157	*/
	158	if (flags & ASYNC_TX_XOR_DROP_DST) {
	159	src_cnt--;
	160	src_off++;
	161	}
	162
	163	/* process up to 'MAX_XOR_BLOCKS' sources */
	164	xor_src_cnt = min(src_cnt, MAX_XOR_BLOCKS);
	165
	166	/* if we are submitting additional xors
	167	* only set the callback on the last transaction
168	*/
169	if (src_cnt > xor_src_cnt) {
170	local_flags &= ~ASYNC_TX_ACK;
171	_cb_fn = NULL;
172	_cb_param = NULL;
173	} else {
174	_cb_fn = cb_fn;
175	_cb_param = cb_param;
176	}
177
178	/* wait for any prerequisite operations */
179	if (depend_tx) {
180	/* if ack is already set then we cannot be sure
181	* we are referring to the correct operation
182	*/
183	BUG_ON(depend_tx->ack);
184	if (dma_wait_for_async_tx(depend_tx) ==
185	DMA_ERROR)
186	panic("%s: DMA_ERROR waiting for "
187	"depend_tx\n",
188	__FUNCTION__);
189	}
190
191	do_sync_xor(dest, &src_list[src_off], offset,
192	xor_src_cnt, len, local_flags, depend_tx,
193	_cb_fn, _cb_param);
194	}
195
196	/* the previous tx is hidden from the client,
197	* so ack it
198	*/
199	if (i && depend_tx)
200	async_tx_ack(depend_tx);
201
202	depend_tx = tx;
203
204	if (src_cnt > xor_src_cnt) {
205	/* drop completed sources */
206	src_cnt -= xor_src_cnt;
207	src_off += xor_src_cnt;
208
209	/* unconditionally preserve the destination */
210	flags &= ~ASYNC_TX_XOR_ZERO_DST;
211
212	/* use the intermediate result a source, but remember
213	* it's dropped, because it's implied, in the sync case
214	*/
215	src_list[--src_off] = dest;
216	src_cnt++;
217	flags \|= ASYNC_TX_XOR_DROP_DST;
218	} else
219	src_cnt = 0;
220	i++;
221	}
222
223	return tx;
224	}
225	EXPORT_SYMBOL_GPL(async_xor);
226
227	static int page_is_zero(struct page *p, unsigned int offset, size_t len)
228	{
229	char *a = page_address(p) + offset;
230	return (((u32 ) a) == 0 &&
231	memcmp(a, a + 4, len - 4) == 0);
232	}
233
234	/**
235	* async_xor_zero_sum - attempt a xor parity check with a dma engine.
236	* @dest: destination page used if the xor is performed synchronously
237	* @src_list: array of source pages. The dest page must be listed as a source
238	* at index zero. The contents of this array may be overwritten.
239	* @offset: offset in pages to start transaction
240	* @src_cnt: number of source pages
241	* @len: length in bytes
242	* @result: 0 if sum == 0 else non-zero
d909b347	243	* @flags: ASYNC_TX_ACK, ASYNC_TX_DEP_ACK
9bc89cd8 DW	244	* @depend_tx: xor depends on the result of this transaction.
	245	* @cb_fn: function to call when the xor completes
	246	* @cb_param: parameter to pass to the callback routine
	247	*/
	248	struct dma_async_tx_descriptor *
	249	async_xor_zero_sum(struct page dest, struct page *src_list,
	250	unsigned int offset, int src_cnt, size_t len,
	251	u32 *result, enum async_tx_flags flags,
	252	struct dma_async_tx_descriptor *depend_tx,
	253	dma_async_tx_callback cb_fn, void *cb_param)
	254	{
	255	struct dma_chan *chan = async_tx_find_channel(depend_tx, DMA_ZERO_SUM);
	256	struct dma_device *device = chan ? chan->device : NULL;
	257	int int_en = cb_fn ? 1 : 0;
	258	struct dma_async_tx_descriptor *tx = device ?
	259	device->device_prep_dma_zero_sum(chan, src_cnt, len, result,
	260	int_en) : NULL;
	261	int i;
	262
	263	BUG_ON(src_cnt <= 1);
	264
	265	if (tx) {
	266	dma_addr_t dma_addr;
9bc89cd8 DW	267
	268	pr_debug("%s: (async) len: %zu\n", __FUNCTION__, len);
	269
9bc89cd8 DW	270	for (i = 0; i < src_cnt; i++) {
9bc89cd8 DW	271	dma_addr = dma_map_page(device->dev, src_list[i],
d909b347	272	offset, len, DMA_TO_DEVICE);
9bc89cd8 DW	273	tx->tx_set_src(dma_addr, tx, i);
	274	}
	275
	276	async_tx_submit(chan, tx, flags, depend_tx, cb_fn, cb_param);
	277	} else {
	278	unsigned long xor_flags = flags;
	279
	280	pr_debug("%s: (sync) len: %zu\n", __FUNCTION__, len);
	281
	282	xor_flags \|= ASYNC_TX_XOR_DROP_DST;
	283	xor_flags &= ~ASYNC_TX_ACK;
	284
	285	tx = async_xor(dest, src_list, offset, src_cnt, len, xor_flags,
	286	depend_tx, NULL, NULL);
	287
	288	if (tx) {
	289	if (dma_wait_for_async_tx(tx) == DMA_ERROR)
	290	panic("%s: DMA_ERROR waiting for tx\n",
	291	__FUNCTION__);
	292	async_tx_ack(tx);
	293	}
	294
	295	*result = page_is_zero(dest, offset, len) ? 0 : 1;
	296
	297	tx = NULL;
	298
	299	async_tx_sync_epilog(flags, depend_tx, cb_fn, cb_param);
	300	}
	301
	302	return tx;
	303	}
	304	EXPORT_SYMBOL_GPL(async_xor_zero_sum);
	305
	306	static int __init async_xor_init(void)
	307	{
	308	return 0;
	309	}
	310
	311	static void __exit async_xor_exit(void)
	312	{
	313	do { } while (0);
	314	}
	315
	316	module_init(async_xor_init);
	317	module_exit(async_xor_exit);
	318
	319	MODULE_AUTHOR("Intel Corporation");
	320	MODULE_DESCRIPTION("asynchronous xor/xor-zero-sum api");
	321	MODULE_LICENSE("GPL");