1 /*-
2 * Copyright (c) 2013-2020, Mellanox Technologies, Ltd. All rights reserved.
3 *
4 * Redistribution and use in source and binary forms, with or without
5 * modification, are permitted provided that the following conditions
6 * are met:
7 * 1. Redistributions of source code must retain the above copyright
8 * notice, this list of conditions and the following disclaimer.
9 * 2. Redistributions in binary form must reproduce the above copyright
10 * notice, this list of conditions and the following disclaimer in the
11 * documentation and/or other materials provided with the distribution.
12 *
13 * THIS SOFTWARE IS PROVIDED BY AUTHOR AND CONTRIBUTORS `AS IS' AND
14 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
15 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
16 * ARE DISCLAIMED. IN NO EVENT SHALL AUTHOR OR CONTRIBUTORS BE LIABLE
17 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
18 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
19 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
20 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
21 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
22 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
23 * SUCH DAMAGE.
24 */
25
26 #include "opt_rss.h"
27 #include "opt_ratelimit.h"
28
29 #include <linux/module.h>
30 #include <rdma/ib_umem.h>
31 #include <rdma/ib_umem_odp.h>
32 #include <dev/mlx5/mlx5_ib/mlx5_ib.h>
33
34 /* @umem: umem object to scan
35 * @addr: ib virtual address requested by the user
36 * @max_page_shift: high limit for page_shift - 0 means no limit
37 * @count: number of PAGE_SIZE pages covered by umem
38 * @shift: page shift for the compound pages found in the region
39 * @ncont: number of compund pages
40 * @order: log2 of the number of compound pages
41 */
mlx5_ib_cont_pages(struct ib_umem * umem,u64 addr,unsigned long max_page_shift,int * count,int * shift,int * ncont,int * order)42 void mlx5_ib_cont_pages(struct ib_umem *umem, u64 addr,
43 unsigned long max_page_shift,
44 int *count, int *shift,
45 int *ncont, int *order)
46 {
47 unsigned long tmp;
48 unsigned long m;
49 u64 base = ~0, p = 0;
50 u64 len, pfn;
51 int i = 0;
52 struct scatterlist *sg;
53 int entry;
54
55 addr = addr >> PAGE_SHIFT;
56 tmp = (unsigned long)addr;
57 m = find_first_bit(&tmp, BITS_PER_LONG);
58 if (max_page_shift)
59 m = min_t(unsigned long, max_page_shift - PAGE_SHIFT, m);
60
61 for_each_sg(umem->sg_head.sgl, sg, umem->nmap, entry) {
62 len = sg_dma_len(sg) >> PAGE_SHIFT;
63 pfn = sg_dma_address(sg) >> PAGE_SHIFT;
64 if (base + p != pfn) {
65 /* If either the offset or the new
66 * base are unaligned update m
67 */
68 tmp = (unsigned long)(pfn | p);
69 if (!IS_ALIGNED(tmp, 1 << m))
70 m = find_first_bit(&tmp, BITS_PER_LONG);
71
72 base = pfn;
73 p = 0;
74 }
75
76 p += len;
77 i += len;
78 }
79
80 if (i) {
81 m = min_t(unsigned long, order_base_2(i), m);
82
83 if (order)
84 *order = order_base_2(i) - m;
85
86 *ncont = DIV_ROUND_UP(i, (1 << m));
87 } else {
88 m = 0;
89
90 if (order)
91 *order = 0;
92
93 *ncont = 0;
94 }
95 *shift = PAGE_SHIFT + m;
96 *count = i;
97 }
98
99 #ifdef CONFIG_INFINIBAND_ON_DEMAND_PAGING
umem_dma_to_mtt(dma_addr_t umem_dma)100 static u64 umem_dma_to_mtt(dma_addr_t umem_dma)
101 {
102 u64 mtt_entry = umem_dma & ODP_DMA_ADDR_MASK;
103
104 if (umem_dma & ODP_READ_ALLOWED_BIT)
105 mtt_entry |= MLX5_IB_MTT_READ;
106 if (umem_dma & ODP_WRITE_ALLOWED_BIT)
107 mtt_entry |= MLX5_IB_MTT_WRITE;
108
109 return mtt_entry;
110 }
111 #endif
112
113 /*
114 * Populate the given array with bus addresses from the umem.
115 *
116 * dev - mlx5_ib device
117 * umem - umem to use to fill the pages
118 * page_shift - determines the page size used in the resulting array
119 * offset - offset into the umem to start from,
120 * only implemented for ODP umems
121 * num_pages - total number of pages to fill
122 * pas - bus addresses array to fill
123 * access_flags - access flags to set on all present pages.
124 use enum mlx5_ib_mtt_access_flags for this.
125 */
__mlx5_ib_populate_pas(struct mlx5_ib_dev * dev,struct ib_umem * umem,int page_shift,size_t offset,size_t num_pages,__be64 * pas,int access_flags)126 void __mlx5_ib_populate_pas(struct mlx5_ib_dev *dev, struct ib_umem *umem,
127 int page_shift, size_t offset, size_t num_pages,
128 __be64 *pas, int access_flags)
129 {
130 unsigned long umem_page_shift = ilog2(umem->page_size);
131 int shift = page_shift - umem_page_shift;
132 int mask = (1 << shift) - 1;
133 int i, k;
134 u64 cur = 0;
135 u64 base;
136 int len;
137 struct scatterlist *sg;
138 int entry;
139 #ifdef CONFIG_INFINIBAND_ON_DEMAND_PAGING
140 const bool odp = umem->odp_data != NULL;
141
142 if (odp) {
143 WARN_ON(shift != 0);
144 WARN_ON(access_flags != (MLX5_IB_MTT_READ | MLX5_IB_MTT_WRITE));
145
146 for (i = 0; i < num_pages; ++i) {
147 dma_addr_t pa = umem->odp_data->dma_list[offset + i];
148
149 pas[i] = cpu_to_be64(umem_dma_to_mtt(pa));
150 }
151 return;
152 }
153 #endif
154
155 i = 0;
156 for_each_sg(umem->sg_head.sgl, sg, umem->nmap, entry) {
157 len = sg_dma_len(sg) >> umem_page_shift;
158 base = sg_dma_address(sg);
159 for (k = 0; k < len; k++) {
160 if (!(i & mask)) {
161 cur = base + (k << umem_page_shift);
162 cur |= access_flags;
163
164 pas[i >> shift] = cpu_to_be64(cur);
165 mlx5_ib_dbg(dev, "pas[%d] 0x%llx\n",
166 i >> shift, (long long)be64_to_cpu(pas[i >> shift]));
167 } else
168 mlx5_ib_dbg(dev, "=====> 0x%llx\n",
169 (long long)(base + (k << umem_page_shift)));
170 i++;
171 }
172 }
173 }
174
mlx5_ib_populate_pas(struct mlx5_ib_dev * dev,struct ib_umem * umem,int page_shift,__be64 * pas,int access_flags)175 void mlx5_ib_populate_pas(struct mlx5_ib_dev *dev, struct ib_umem *umem,
176 int page_shift, __be64 *pas, int access_flags)
177 {
178 return __mlx5_ib_populate_pas(dev, umem, page_shift, 0,
179 ib_umem_num_pages(umem), pas,
180 access_flags);
181 }
mlx5_ib_get_buf_offset(u64 addr,int page_shift,u32 * offset)182 int mlx5_ib_get_buf_offset(u64 addr, int page_shift, u32 *offset)
183 {
184 u64 page_size;
185 u64 page_mask;
186 u64 off_size;
187 u64 off_mask;
188 u64 buf_off;
189
190 page_size = (u64)1 << page_shift;
191 page_mask = page_size - 1;
192 buf_off = addr & page_mask;
193 off_size = page_size >> 6;
194 off_mask = off_size - 1;
195
196 if (buf_off & off_mask)
197 return -EINVAL;
198
199 *offset = buf_off >> ilog2(off_size);
200 return 0;
201 }
202