1 /* 2 * Copyright (c) 2013-2015, Mellanox Technologies. All rights reserved. 3 * 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: 9 * 10 * Redistribution and use in source and binary forms, with or 11 * without modification, are permitted provided that the following 12 * conditions are met: 13 * 14 * - Redistributions of source code must retain the above 15 * copyright notice, this list of conditions and the following 16 * disclaimer. 17 * 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. 22 * 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 30 * SOFTWARE. 31 */ 32 33 #include <linux/module.h> 34 #include <rdma/ib_umem.h> 35 #include <rdma/ib_umem_odp.h> 36 #include "mlx5_ib.h" 37 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 45 */ 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) 50 { 51 unsigned long tmp; 52 unsigned long m; 53 u64 base = ~0, p = 0; 54 u64 len, pfn; 55 int i = 0; 56 struct scatterlist *sg; 57 int entry; 58 unsigned long page_shift = umem->page_shift; 59 60 if (umem->is_odp) { 61 *ncont = ib_umem_page_count(umem); 62 *count = *ncont << (page_shift - PAGE_SHIFT); 63 *shift = page_shift; 64 if (order) 65 *order = ilog2(roundup_pow_of_two(*ncont)); 66 67 return; 68 } 69 70 addr = addr >> page_shift; 71 tmp = (unsigned long)addr; 72 m = find_first_bit(&tmp, BITS_PER_LONG); 73 if (max_page_shift) 74 m = min_t(unsigned long, max_page_shift - page_shift, m); 75 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 82 */ 83 tmp = (unsigned long)(pfn | p); 84 if (!IS_ALIGNED(tmp, 1 << m)) 85 m = find_first_bit(&tmp, BITS_PER_LONG); 86 87 base = pfn; 88 p = 0; 89 } 90 91 p += len; 92 i += len; 93 } 94 95 if (i) { 96 m = min_t(unsigned long, ilog2(roundup_pow_of_two(i)), m); 97 98 if (order) 99 *order = ilog2(roundup_pow_of_two(i) >> m); 100 101 *ncont = DIV_ROUND_UP(i, (1 << m)); 102 } else { 103 m = 0; 104 105 if (order) 106 *order = 0; 107 108 *ncont = 0; 109 } 110 *shift = page_shift + m; 111 *count = i; 112 } 113 114 static u64 umem_dma_to_mtt(dma_addr_t umem_dma) 115 { 116 u64 mtt_entry = umem_dma & ODP_DMA_ADDR_MASK; 117 118 if (umem_dma & ODP_READ_ALLOWED_BIT) 119 mtt_entry |= MLX5_IB_MTT_READ; 120 if (umem_dma & ODP_WRITE_ALLOWED_BIT) 121 mtt_entry |= MLX5_IB_MTT_WRITE; 122 123 return mtt_entry; 124 } 125 126 /* 127 * Populate the given array with bus addresses from the umem. 128 * 129 * dev - mlx5_ib device 130 * umem - umem to use to fill the pages 131 * page_shift - determines the page size used in the resulting array 132 * offset - offset into the umem to start from, 133 * only implemented for ODP umems 134 * num_pages - total number of pages to fill 135 * pas - bus addresses array to fill 136 * access_flags - access flags to set on all present pages. 137 use enum mlx5_ib_mtt_access_flags for this. 138 */ 139 void __mlx5_ib_populate_pas(struct mlx5_ib_dev *dev, struct ib_umem *umem, 140 int page_shift, size_t offset, size_t num_pages, 141 __be64 *pas, int access_flags) 142 { 143 unsigned long umem_page_shift = umem->page_shift; 144 int shift = page_shift - umem_page_shift; 145 int mask = (1 << shift) - 1; 146 int i, k, idx; 147 u64 cur = 0; 148 u64 base; 149 int len; 150 struct scatterlist *sg; 151 int entry; 152 153 if (umem->is_odp) { 154 WARN_ON(shift != 0); 155 WARN_ON(access_flags != (MLX5_IB_MTT_READ | MLX5_IB_MTT_WRITE)); 156 157 for (i = 0; i < num_pages; ++i) { 158 dma_addr_t pa = 159 to_ib_umem_odp(umem)->dma_list[offset + i]; 160 161 pas[i] = cpu_to_be64(umem_dma_to_mtt(pa)); 162 } 163 return; 164 } 165 166 i = 0; 167 for_each_sg(umem->sg_head.sgl, sg, umem->nmap, entry) { 168 len = sg_dma_len(sg) >> umem_page_shift; 169 base = sg_dma_address(sg); 170 171 /* Skip elements below offset */ 172 if (i + len < offset << shift) { 173 i += len; 174 continue; 175 } 176 177 /* Skip pages below offset */ 178 if (i < offset << shift) { 179 k = (offset << shift) - i; 180 i = offset << shift; 181 } else { 182 k = 0; 183 } 184 185 for (; k < len; k++) { 186 if (!(i & mask)) { 187 cur = base + (k << umem_page_shift); 188 cur |= access_flags; 189 idx = (i >> shift) - offset; 190 191 pas[idx] = cpu_to_be64(cur); 192 mlx5_ib_dbg(dev, "pas[%d] 0x%llx\n", 193 i >> shift, be64_to_cpu(pas[idx])); 194 } 195 i++; 196 197 /* Stop after num_pages reached */ 198 if (i >> shift >= offset + num_pages) 199 return; 200 } 201 } 202 } 203 204 void mlx5_ib_populate_pas(struct mlx5_ib_dev *dev, struct ib_umem *umem, 205 int page_shift, __be64 *pas, int access_flags) 206 { 207 return __mlx5_ib_populate_pas(dev, umem, page_shift, 0, 208 ib_umem_num_pages(umem), pas, 209 access_flags); 210 } 211 int mlx5_ib_get_buf_offset(u64 addr, int page_shift, u32 *offset) 212 { 213 u64 page_size; 214 u64 page_mask; 215 u64 off_size; 216 u64 off_mask; 217 u64 buf_off; 218 219 page_size = (u64)1 << page_shift; 220 page_mask = page_size - 1; 221 buf_off = addr & page_mask; 222 off_size = page_size >> 6; 223 off_mask = off_size - 1; 224 225 if (buf_off & off_mask) 226 return -EINVAL; 227 228 *offset = buf_off >> ilog2(off_size); 229 return 0; 230 } 231