1 /*
2  * copy offload engine support
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/highmem.h>
28 #include <linux/module.h>
29 #include <linux/mm.h>
30 #include <linux/dma-mapping.h>
31 #include <linux/async_tx.h>
32 
33 /**
34  * async_memcpy - attempt to copy memory with a dma engine.
35  * @dest: destination page
36  * @src: src page
37  * @dest_offset: offset into 'dest' to start transaction
38  * @src_offset: offset into 'src' to start transaction
39  * @len: length in bytes
40  * @submit: submission / completion modifiers
41  *
42  * honored flags: ASYNC_TX_ACK
43  */
44 struct dma_async_tx_descriptor *
async_memcpy(struct page * dest,struct page * src,unsigned int dest_offset,unsigned int src_offset,size_t len,struct async_submit_ctl * submit)45 async_memcpy(struct page *dest, struct page *src, unsigned int dest_offset,
46 	     unsigned int src_offset, size_t len,
47 	     struct async_submit_ctl *submit)
48 {
49 	struct dma_chan *chan = async_tx_find_channel(submit, DMA_MEMCPY,
50 						      &dest, 1, &src, 1, len);
51 	struct dma_device *device = chan ? chan->device : NULL;
52 	struct dma_async_tx_descriptor *tx = NULL;
53 	struct dmaengine_unmap_data *unmap = NULL;
54 
55 	if (device)
56 		unmap = dmaengine_get_unmap_data(device->dev, 2, GFP_NOWAIT);
57 
58 	if (unmap && is_dma_copy_aligned(device, src_offset, dest_offset, len)) {
59 		unsigned long dma_prep_flags = 0;
60 
61 		if (submit->cb_fn)
62 			dma_prep_flags |= DMA_PREP_INTERRUPT;
63 		if (submit->flags & ASYNC_TX_FENCE)
64 			dma_prep_flags |= DMA_PREP_FENCE;
65 
66 		unmap->to_cnt = 1;
67 		unmap->addr[0] = dma_map_page(device->dev, src, src_offset, len,
68 					      DMA_TO_DEVICE);
69 		unmap->from_cnt = 1;
70 		unmap->addr[1] = dma_map_page(device->dev, dest, dest_offset, len,
71 					      DMA_FROM_DEVICE);
72 		unmap->len = len;
73 
74 		tx = device->device_prep_dma_memcpy(chan, unmap->addr[1],
75 						    unmap->addr[0], len,
76 						    dma_prep_flags);
77 	}
78 
79 	if (tx) {
80 		pr_debug("%s: (async) len: %zu\n", __func__, len);
81 
82 		dma_set_unmap(tx, unmap);
83 		async_tx_submit(chan, tx, submit);
84 	} else {
85 		void *dest_buf, *src_buf;
86 		pr_debug("%s: (sync) len: %zu\n", __func__, len);
87 
88 		/* wait for any prerequisite operations */
89 		async_tx_quiesce(&submit->depend_tx);
90 
91 		dest_buf = kmap_atomic(dest) + dest_offset;
92 		src_buf = kmap_atomic(src) + src_offset;
93 
94 		memcpy(dest_buf, src_buf, len);
95 
96 		kunmap_atomic(src_buf);
97 		kunmap_atomic(dest_buf);
98 
99 		async_tx_sync_epilog(submit);
100 	}
101 
102 	dmaengine_unmap_put(unmap);
103 
104 	return tx;
105 }
106 EXPORT_SYMBOL_GPL(async_memcpy);
107 
108 MODULE_AUTHOR("Intel Corporation");
109 MODULE_DESCRIPTION("asynchronous memcpy api");
110 MODULE_LICENSE("GPL");
111