mem.c 5.9 KB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204205206207208209210211212213214215216217218219220221222223224225
  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. #include <linux/module.h>
  33. #include <rdma/ib_umem.h>
  34. #include <rdma/ib_umem_odp.h>
  35. #include "mlx5_ib.h"
  36. /* @umem: umem object to scan
  37. * @addr: ib virtual address requested by the user
  38. * @count: number of PAGE_SIZE pages covered by umem
  39. * @shift: page shift for the compound pages found in the region
  40. * @ncont: number of compund pages
  41. * @order: log2 of the number of compound pages
  42. */
  43. void mlx5_ib_cont_pages(struct ib_umem *umem, u64 addr, int *count, int *shift,
  44. int *ncont, int *order)
  45. {
  46. unsigned long tmp;
  47. unsigned long m;
  48. int i, k;
  49. u64 base = 0;
  50. int p = 0;
  51. int skip;
  52. int mask;
  53. u64 len;
  54. u64 pfn;
  55. struct scatterlist *sg;
  56. int entry;
  57. unsigned long page_shift = ilog2(umem->page_size);
  58. /* With ODP we must always match OS page size. */
  59. if (umem->odp_data) {
  60. *count = ib_umem_page_count(umem);
  61. *shift = PAGE_SHIFT;
  62. *ncont = *count;
  63. if (order)
  64. *order = ilog2(roundup_pow_of_two(*count));
  65. return;
  66. }
  67. addr = addr >> page_shift;
  68. tmp = (unsigned long)addr;
  69. m = find_first_bit(&tmp, BITS_PER_LONG);
  70. skip = 1 << m;
  71. mask = skip - 1;
  72. i = 0;
  73. for_each_sg(umem->sg_head.sgl, sg, umem->nmap, entry) {
  74. len = sg_dma_len(sg) >> page_shift;
  75. pfn = sg_dma_address(sg) >> page_shift;
  76. for (k = 0; k < len; k++) {
  77. if (!(i & mask)) {
  78. tmp = (unsigned long)pfn;
  79. m = min_t(unsigned long, m, find_first_bit(&tmp, BITS_PER_LONG));
  80. skip = 1 << m;
  81. mask = skip - 1;
  82. base = pfn;
  83. p = 0;
  84. } else {
  85. if (base + p != pfn) {
  86. tmp = (unsigned long)p;
  87. m = find_first_bit(&tmp, BITS_PER_LONG);
  88. skip = 1 << m;
  89. mask = skip - 1;
  90. base = pfn;
  91. p = 0;
  92. }
  93. }
  94. p++;
  95. i++;
  96. }
  97. }
  98. if (i) {
  99. m = min_t(unsigned long, ilog2(roundup_pow_of_two(i)), m);
  100. if (order)
  101. *order = ilog2(roundup_pow_of_two(i) >> m);
  102. *ncont = DIV_ROUND_UP(i, (1 << m));
  103. } else {
  104. m = 0;
  105. if (order)
  106. *order = 0;
  107. *ncont = 0;
  108. }
  109. *shift = page_shift + m;
  110. *count = i;
  111. }
  112. #ifdef CONFIG_INFINIBAND_ON_DEMAND_PAGING
  113. static u64 umem_dma_to_mtt(dma_addr_t umem_dma)
  114. {
  115. u64 mtt_entry = umem_dma & ODP_DMA_ADDR_MASK;
  116. if (umem_dma & ODP_READ_ALLOWED_BIT)
  117. mtt_entry |= MLX5_IB_MTT_READ;
  118. if (umem_dma & ODP_WRITE_ALLOWED_BIT)
  119. mtt_entry |= MLX5_IB_MTT_WRITE;
  120. return mtt_entry;
  121. }
  122. #endif
  123. /*
  124. * Populate the given array with bus addresses from the umem.
  125. *
  126. * dev - mlx5_ib device
  127. * umem - umem to use to fill the pages
  128. * page_shift - determines the page size used in the resulting array
  129. * offset - offset into the umem to start from,
  130. * only implemented for ODP umems
  131. * num_pages - total number of pages to fill
  132. * pas - bus addresses array to fill
  133. * access_flags - access flags to set on all present pages.
  134. use enum mlx5_ib_mtt_access_flags for this.
  135. */
  136. void __mlx5_ib_populate_pas(struct mlx5_ib_dev *dev, struct ib_umem *umem,
  137. int page_shift, size_t offset, size_t num_pages,
  138. __be64 *pas, int access_flags)
  139. {
  140. unsigned long umem_page_shift = ilog2(umem->page_size);
  141. int shift = page_shift - umem_page_shift;
  142. int mask = (1 << shift) - 1;
  143. int i, k;
  144. u64 cur = 0;
  145. u64 base;
  146. int len;
  147. struct scatterlist *sg;
  148. int entry;
  149. #ifdef CONFIG_INFINIBAND_ON_DEMAND_PAGING
  150. const bool odp = umem->odp_data != NULL;
  151. if (odp) {
  152. WARN_ON(shift != 0);
  153. WARN_ON(access_flags != (MLX5_IB_MTT_READ | MLX5_IB_MTT_WRITE));
  154. for (i = 0; i < num_pages; ++i) {
  155. dma_addr_t pa = umem->odp_data->dma_list[offset + i];
  156. pas[i] = cpu_to_be64(umem_dma_to_mtt(pa));
  157. }
  158. return;
  159. }
  160. #endif
  161. i = 0;
  162. for_each_sg(umem->sg_head.sgl, sg, umem->nmap, entry) {
  163. len = sg_dma_len(sg) >> umem_page_shift;
  164. base = sg_dma_address(sg);
  165. for (k = 0; k < len; k++) {
  166. if (!(i & mask)) {
  167. cur = base + (k << umem_page_shift);
  168. cur |= access_flags;
  169. pas[i >> shift] = cpu_to_be64(cur);
  170. mlx5_ib_dbg(dev, "pas[%d] 0x%llx\n",
  171. i >> shift, be64_to_cpu(pas[i >> shift]));
  172. } else
  173. mlx5_ib_dbg(dev, "=====> 0x%llx\n",
  174. base + (k << umem_page_shift));
  175. i++;
  176. }
  177. }
  178. }
  179. void mlx5_ib_populate_pas(struct mlx5_ib_dev *dev, struct ib_umem *umem,
  180. int page_shift, __be64 *pas, int access_flags)
  181. {
  182. return __mlx5_ib_populate_pas(dev, umem, page_shift, 0,
  183. ib_umem_num_pages(umem), pas,
  184. access_flags);
  185. }
  186. int mlx5_ib_get_buf_offset(u64 addr, int page_shift, u32 *offset)
  187. {
  188. u64 page_size;
  189. u64 page_mask;
  190. u64 off_size;
  191. u64 off_mask;
  192. u64 buf_off;
  193. page_size = (u64)1 << page_shift;
  194. page_mask = page_size - 1;
  195. buf_off = addr & page_mask;
  196. off_size = page_size >> 6;
  197. off_mask = off_size - 1;
  198. if (buf_off & off_mask)
  199. return -EINVAL;
  200. *offset = buf_off >> ilog2(off_size);
  201. return 0;
  202. }