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 * @count: number of PAGE_SIZE pages covered by umem 41 * @shift: page shift for the compound pages found in the region 42 * @ncont: number of compund pages 43 * @order: log2 of the number of compound pages 44 */ 45void mlx5_ib_cont_pages(struct ib_umem *umem, u64 addr, int *count, int *shift, 46 int *ncont, int *order) 47{ 48 unsigned long tmp; 49 unsigned long m; 50 int i, k; 51 u64 base = 0; 52 int p = 0; 53 int skip; 54 int mask; 55 u64 len; 56 u64 pfn; 57 struct scatterlist *sg; 58 int entry; 59 unsigned long page_shift = ilog2(umem->page_size); 60 61 /* With ODP we must always match OS page size. */ 62 if (umem->odp_data) { 63 *count = ib_umem_page_count(umem); 64 *shift = PAGE_SHIFT; 65 *ncont = *count; 66 if (order) 67 *order = ilog2(roundup_pow_of_two(*count)); 68 69 return; 70 } 71 72 addr = addr >> page_shift; 73 tmp = (unsigned long)addr; 74 m = find_first_bit(&tmp, sizeof(tmp)); 75 skip = 1 << m; 76 mask = skip - 1; 77 i = 0; 78 for_each_sg(umem->sg_head.sgl, sg, umem->nmap, entry) { 79 len = sg_dma_len(sg) >> page_shift; 80 pfn = sg_dma_address(sg) >> page_shift; 81 for (k = 0; k < len; k++) { 82 if (!(i & mask)) { 83 tmp = (unsigned long)pfn; 84 m = min_t(unsigned long, m, find_first_bit(&tmp, sizeof(tmp))); 85 skip = 1 << m; 86 mask = skip - 1; 87 base = pfn; 88 p = 0; 89 } else { 90 if (base + p != pfn) { 91 tmp = (unsigned long)p; 92 m = find_first_bit(&tmp, sizeof(tmp)); 93 skip = 1 << m; 94 mask = skip - 1; 95 base = pfn; 96 p = 0; 97 } 98 } 99 p++; 100 i++; 101 } 102 } 103 104 if (i) { 105 m = min_t(unsigned long, ilog2(roundup_pow_of_two(i)), m); 106 107 if (order) 108 *order = ilog2(roundup_pow_of_two(i) >> m); 109 110 *ncont = DIV_ROUND_UP(i, (1 << m)); 111 } else { 112 m = 0; 113 114 if (order) 115 *order = 0; 116 117 *ncont = 0; 118 } 119 *shift = page_shift + m; 120 *count = i; 121} 122 123#ifdef CONFIG_INFINIBAND_ON_DEMAND_PAGING 124static u64 umem_dma_to_mtt(dma_addr_t umem_dma) 125{ 126 u64 mtt_entry = umem_dma & ODP_DMA_ADDR_MASK; 127 128 if (umem_dma & ODP_READ_ALLOWED_BIT) 129 mtt_entry |= MLX5_IB_MTT_READ; 130 if (umem_dma & ODP_WRITE_ALLOWED_BIT) 131 mtt_entry |= MLX5_IB_MTT_WRITE; 132 133 return mtt_entry; 134} 135#endif 136 137/* 138 * Populate the given array with bus addresses from the umem. 139 * 140 * dev - mlx5_ib device 141 * umem - umem to use to fill the pages 142 * page_shift - determines the page size used in the resulting array 143 * offset - offset into the umem to start from, 144 * only implemented for ODP umems 145 * num_pages - total number of pages to fill 146 * pas - bus addresses array to fill 147 * access_flags - access flags to set on all present pages. 148 use enum mlx5_ib_mtt_access_flags for this. 149 */ 150void __mlx5_ib_populate_pas(struct mlx5_ib_dev *dev, struct ib_umem *umem, 151 int page_shift, size_t offset, size_t num_pages, 152 __be64 *pas, int access_flags) 153{ 154 unsigned long umem_page_shift = ilog2(umem->page_size); 155 int shift = page_shift - umem_page_shift; 156 int mask = (1 << shift) - 1; 157 int i, k; 158 u64 cur = 0; 159 u64 base; 160 int len; 161 struct scatterlist *sg; 162 int entry; 163#ifdef CONFIG_INFINIBAND_ON_DEMAND_PAGING 164 const bool odp = umem->odp_data != NULL; 165 166 if (odp) { 167 WARN_ON(shift != 0); 168 WARN_ON(access_flags != (MLX5_IB_MTT_READ | MLX5_IB_MTT_WRITE)); 169 170 for (i = 0; i < num_pages; ++i) { 171 dma_addr_t pa = umem->odp_data->dma_list[offset + i]; 172 173 pas[i] = cpu_to_be64(umem_dma_to_mtt(pa)); 174 } 175 return; 176 } 177#endif 178 179 i = 0; 180 for_each_sg(umem->sg_head.sgl, sg, umem->nmap, entry) { 181 len = sg_dma_len(sg) >> umem_page_shift; 182 base = sg_dma_address(sg); 183 for (k = 0; k < len; k++) { 184 if (!(i & mask)) { 185 cur = base + (k << umem_page_shift); 186 cur |= access_flags; 187 188 pas[i >> shift] = cpu_to_be64(cur); 189 mlx5_ib_dbg(dev, "pas[%d] 0x%llx\n", 190 i >> shift, be64_to_cpu(pas[i >> shift])); 191 } else 192 mlx5_ib_dbg(dev, "=====> 0x%llx\n", 193 base + (k << umem_page_shift)); 194 i++; 195 } 196 } 197} 198 199void mlx5_ib_populate_pas(struct mlx5_ib_dev *dev, struct ib_umem *umem, 200 int page_shift, __be64 *pas, int access_flags) 201{ 202 return __mlx5_ib_populate_pas(dev, umem, page_shift, 0, 203 ib_umem_num_pages(umem), pas, 204 access_flags); 205} 206int mlx5_ib_get_buf_offset(u64 addr, int page_shift, u32 *offset) 207{ 208 u64 page_size; 209 u64 page_mask; 210 u64 off_size; 211 u64 off_mask; 212 u64 buf_off; 213 214 page_size = (u64)1 << page_shift; 215 page_mask = page_size - 1; 216 buf_off = addr & page_mask; 217 off_size = page_size >> 6; 218 off_mask = off_size - 1; 219 220 if (buf_off & off_mask) 221 return -EINVAL; 222 223 *offset = buf_off >> ilog2(off_size); 224 return 0; 225} 226