dax.h 4.9 KB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167
  1. #ifndef _LINUX_DAX_H
  2. #define _LINUX_DAX_H
  3. #include <linux/fs.h>
  4. #include <linux/mm.h>
  5. #include <linux/radix-tree.h>
  6. #include <asm/pgtable.h>
  7. struct iomap_ops;
  8. struct dax_device;
  9. struct dax_operations {
  10. /*
  11. * direct_access: translate a device-relative
  12. * logical-page-offset into an absolute physical pfn. Return the
  13. * number of pages available for DAX at that pfn.
  14. */
  15. long (*direct_access)(struct dax_device *, pgoff_t, long,
  16. void **, pfn_t *);
  17. /* copy_from_iter: dax-driver override for default copy_from_iter */
  18. size_t (*copy_from_iter)(struct dax_device *, pgoff_t, void *, size_t,
  19. struct iov_iter *);
  20. /* flush: optional driver-specific cache management after writes */
  21. void (*flush)(struct dax_device *, pgoff_t, void *, size_t);
  22. };
  23. #if IS_ENABLED(CONFIG_DAX)
  24. struct dax_device *dax_get_by_host(const char *host);
  25. void put_dax(struct dax_device *dax_dev);
  26. #else
  27. static inline struct dax_device *dax_get_by_host(const char *host)
  28. {
  29. return NULL;
  30. }
  31. static inline void put_dax(struct dax_device *dax_dev)
  32. {
  33. }
  34. #endif
  35. int bdev_dax_pgoff(struct block_device *, sector_t, size_t, pgoff_t *pgoff);
  36. #if IS_ENABLED(CONFIG_FS_DAX)
  37. int __bdev_dax_supported(struct super_block *sb, int blocksize);
  38. static inline int bdev_dax_supported(struct super_block *sb, int blocksize)
  39. {
  40. return __bdev_dax_supported(sb, blocksize);
  41. }
  42. static inline struct dax_device *fs_dax_get_by_host(const char *host)
  43. {
  44. return dax_get_by_host(host);
  45. }
  46. static inline void fs_put_dax(struct dax_device *dax_dev)
  47. {
  48. put_dax(dax_dev);
  49. }
  50. #else
  51. static inline int bdev_dax_supported(struct super_block *sb, int blocksize)
  52. {
  53. return -EOPNOTSUPP;
  54. }
  55. static inline struct dax_device *fs_dax_get_by_host(const char *host)
  56. {
  57. return NULL;
  58. }
  59. static inline void fs_put_dax(struct dax_device *dax_dev)
  60. {
  61. }
  62. #endif
  63. int dax_read_lock(void);
  64. void dax_read_unlock(int id);
  65. struct dax_device *alloc_dax(void *private, const char *host,
  66. const struct dax_operations *ops);
  67. bool dax_alive(struct dax_device *dax_dev);
  68. void kill_dax(struct dax_device *dax_dev);
  69. void *dax_get_private(struct dax_device *dax_dev);
  70. long dax_direct_access(struct dax_device *dax_dev, pgoff_t pgoff, long nr_pages,
  71. void **kaddr, pfn_t *pfn);
  72. size_t dax_copy_from_iter(struct dax_device *dax_dev, pgoff_t pgoff, void *addr,
  73. size_t bytes, struct iov_iter *i);
  74. void dax_flush(struct dax_device *dax_dev, pgoff_t pgoff, void *addr,
  75. size_t size);
  76. /*
  77. * We use lowest available bit in exceptional entry for locking, one bit for
  78. * the entry size (PMD) and two more to tell us if the entry is a huge zero
  79. * page (HZP) or an empty entry that is just used for locking. In total four
  80. * special bits.
  81. *
  82. * If the PMD bit isn't set the entry has size PAGE_SIZE, and if the HZP and
  83. * EMPTY bits aren't set the entry is a normal DAX entry with a filesystem
  84. * block allocation.
  85. */
  86. #define RADIX_DAX_SHIFT (RADIX_TREE_EXCEPTIONAL_SHIFT + 4)
  87. #define RADIX_DAX_ENTRY_LOCK (1 << RADIX_TREE_EXCEPTIONAL_SHIFT)
  88. #define RADIX_DAX_PMD (1 << (RADIX_TREE_EXCEPTIONAL_SHIFT + 1))
  89. #define RADIX_DAX_HZP (1 << (RADIX_TREE_EXCEPTIONAL_SHIFT + 2))
  90. #define RADIX_DAX_EMPTY (1 << (RADIX_TREE_EXCEPTIONAL_SHIFT + 3))
  91. static inline unsigned long dax_radix_sector(void *entry)
  92. {
  93. return (unsigned long)entry >> RADIX_DAX_SHIFT;
  94. }
  95. static inline void *dax_radix_locked_entry(sector_t sector, unsigned long flags)
  96. {
  97. return (void *)(RADIX_TREE_EXCEPTIONAL_ENTRY | flags |
  98. ((unsigned long)sector << RADIX_DAX_SHIFT) |
  99. RADIX_DAX_ENTRY_LOCK);
  100. }
  101. ssize_t dax_iomap_rw(struct kiocb *iocb, struct iov_iter *iter,
  102. const struct iomap_ops *ops);
  103. int dax_iomap_fault(struct vm_fault *vmf, enum page_entry_size pe_size,
  104. const struct iomap_ops *ops);
  105. int dax_delete_mapping_entry(struct address_space *mapping, pgoff_t index);
  106. int dax_invalidate_mapping_entry_sync(struct address_space *mapping,
  107. pgoff_t index);
  108. void dax_wake_mapping_entry_waiter(struct address_space *mapping,
  109. pgoff_t index, void *entry, bool wake_all);
  110. #ifdef CONFIG_FS_DAX
  111. int __dax_zero_page_range(struct block_device *bdev,
  112. struct dax_device *dax_dev, sector_t sector,
  113. unsigned int offset, unsigned int length);
  114. #else
  115. static inline int __dax_zero_page_range(struct block_device *bdev,
  116. struct dax_device *dax_dev, sector_t sector,
  117. unsigned int offset, unsigned int length)
  118. {
  119. return -ENXIO;
  120. }
  121. #endif
  122. #ifdef CONFIG_FS_DAX_PMD
  123. static inline unsigned int dax_radix_order(void *entry)
  124. {
  125. if ((unsigned long)entry & RADIX_DAX_PMD)
  126. return PMD_SHIFT - PAGE_SHIFT;
  127. return 0;
  128. }
  129. #else
  130. static inline unsigned int dax_radix_order(void *entry)
  131. {
  132. return 0;
  133. }
  134. #endif
  135. int dax_pfn_mkwrite(struct vm_fault *vmf);
  136. static inline bool vma_is_dax(struct vm_area_struct *vma)
  137. {
  138. return vma->vm_file && IS_DAX(vma->vm_file->f_mapping->host);
  139. }
  140. static inline bool dax_mapping(struct address_space *mapping)
  141. {
  142. return mapping->host && IS_DAX(mapping->host);
  143. }
  144. struct writeback_control;
  145. int dax_writeback_mapping_range(struct address_space *mapping,
  146. struct block_device *bdev, struct writeback_control *wbc);
  147. #endif