2 * Copyright (c) 2013-2015, Mellanox Technologies. All rights reserved.
4 * This software is available to you under a choice of one of two
5 * licenses. You may choose to be licensed under the terms of the GNU
6 * General Public License (GPL) Version 2, available from the file
7 * COPYING in the main directory of this source tree, or the
8 * OpenIB.org BSD license below:
10 * Redistribution and use in source and binary forms, with or
11 * without modification, are permitted provided that the following
14 * - Redistributions of source code must retain the above
15 * copyright notice, this list of conditions and the following
18 * - Redistributions in binary form must reproduce the above
19 * copyright notice, this list of conditions and the following
20 * disclaimer in the documentation and/or other materials
21 * provided with the distribution.
23 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
24 * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
25 * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
26 * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
27 * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
28 * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
29 * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
33 #include <linux/module.h>
34 #include <rdma/ib_umem.h>
35 #include <rdma/ib_umem_odp.h>
38 /* @umem: umem object to scan
39 * @addr: ib virtual address requested by the user
40 * @max_page_shift: high limit for page_shift - 0 means no limit
41 * @count: number of PAGE_SIZE pages covered by umem
42 * @shift: page shift for the compound pages found in the region
43 * @ncont: number of compund pages
44 * @order: log2 of the number of compound pages
46 void mlx5_ib_cont_pages(struct ib_umem
*umem
, u64 addr
,
47 unsigned long max_page_shift
,
48 int *count
, int *shift
,
49 int *ncont
, int *order
)
56 struct scatterlist
*sg
;
58 unsigned long page_shift
= umem
->page_shift
;
61 *ncont
= ib_umem_page_count(umem
);
62 *count
= *ncont
<< (page_shift
- PAGE_SHIFT
);
65 *order
= ilog2(roundup_pow_of_two(*ncont
));
70 addr
= addr
>> page_shift
;
71 tmp
= (unsigned long)addr
;
72 m
= find_first_bit(&tmp
, BITS_PER_LONG
);
74 m
= min_t(unsigned long, max_page_shift
- page_shift
, m
);
76 for_each_sg(umem
->sg_head
.sgl
, sg
, umem
->nmap
, entry
) {
77 len
= sg_dma_len(sg
) >> page_shift
;
78 pfn
= sg_dma_address(sg
) >> page_shift
;
79 if (base
+ p
!= pfn
) {
80 /* If either the offset or the new
81 * base are unaligned update m
83 tmp
= (unsigned long)(pfn
| p
);
84 if (!IS_ALIGNED(tmp
, 1 << m
))
85 m
= find_first_bit(&tmp
, BITS_PER_LONG
);
96 m
= min_t(unsigned long, ilog2(roundup_pow_of_two(i
)), m
);
99 *order
= ilog2(roundup_pow_of_two(i
) >> m
);
101 *ncont
= DIV_ROUND_UP(i
, (1 << m
));
110 *shift
= page_shift
+ m
;
114 #ifdef CONFIG_INFINIBAND_ON_DEMAND_PAGING
115 static u64
umem_dma_to_mtt(dma_addr_t umem_dma
)
117 u64 mtt_entry
= umem_dma
& ODP_DMA_ADDR_MASK
;
119 if (umem_dma
& ODP_READ_ALLOWED_BIT
)
120 mtt_entry
|= MLX5_IB_MTT_READ
;
121 if (umem_dma
& ODP_WRITE_ALLOWED_BIT
)
122 mtt_entry
|= MLX5_IB_MTT_WRITE
;
129 * Populate the given array with bus addresses from the umem.
131 * dev - mlx5_ib device
132 * umem - umem to use to fill the pages
133 * page_shift - determines the page size used in the resulting array
134 * offset - offset into the umem to start from,
135 * only implemented for ODP umems
136 * num_pages - total number of pages to fill
137 * pas - bus addresses array to fill
138 * access_flags - access flags to set on all present pages.
139 use enum mlx5_ib_mtt_access_flags for this.
141 void __mlx5_ib_populate_pas(struct mlx5_ib_dev
*dev
, struct ib_umem
*umem
,
142 int page_shift
, size_t offset
, size_t num_pages
,
143 __be64
*pas
, int access_flags
)
145 unsigned long umem_page_shift
= umem
->page_shift
;
146 int shift
= page_shift
- umem_page_shift
;
147 int mask
= (1 << shift
) - 1;
152 struct scatterlist
*sg
;
154 #ifdef CONFIG_INFINIBAND_ON_DEMAND_PAGING
155 const bool odp
= umem
->odp_data
!= NULL
;
159 WARN_ON(access_flags
!= (MLX5_IB_MTT_READ
| MLX5_IB_MTT_WRITE
));
161 for (i
= 0; i
< num_pages
; ++i
) {
162 dma_addr_t pa
= umem
->odp_data
->dma_list
[offset
+ i
];
164 pas
[i
] = cpu_to_be64(umem_dma_to_mtt(pa
));
171 for_each_sg(umem
->sg_head
.sgl
, sg
, umem
->nmap
, entry
) {
172 len
= sg_dma_len(sg
) >> umem_page_shift
;
173 base
= sg_dma_address(sg
);
175 /* Skip elements below offset */
176 if (i
+ len
< offset
<< shift
) {
181 /* Skip pages below offset */
182 if (i
< offset
<< shift
) {
183 k
= (offset
<< shift
) - i
;
189 for (; k
< len
; k
++) {
191 cur
= base
+ (k
<< umem_page_shift
);
193 idx
= (i
>> shift
) - offset
;
195 pas
[idx
] = cpu_to_be64(cur
);
196 mlx5_ib_dbg(dev
, "pas[%d] 0x%llx\n",
197 i
>> shift
, be64_to_cpu(pas
[idx
]));
201 /* Stop after num_pages reached */
202 if (i
>> shift
>= offset
+ num_pages
)
208 void mlx5_ib_populate_pas(struct mlx5_ib_dev
*dev
, struct ib_umem
*umem
,
209 int page_shift
, __be64
*pas
, int access_flags
)
211 return __mlx5_ib_populate_pas(dev
, umem
, page_shift
, 0,
212 ib_umem_num_pages(umem
), pas
,
215 int mlx5_ib_get_buf_offset(u64 addr
, int page_shift
, u32
*offset
)
223 page_size
= (u64
)1 << page_shift
;
224 page_mask
= page_size
- 1;
225 buf_off
= addr
& page_mask
;
226 off_size
= page_size
>> 6;
227 off_mask
= off_size
- 1;
229 if (buf_off
& off_mask
)
232 *offset
= buf_off
>> ilog2(off_size
);